THE ROLE OF MITOCHONDRIA, OXIDATIVE STRESS AND ALTERED CALCIUM HOMEOSTASIS IN AMYOTROPHIC LATERAL SCLEROSIS: FROM CURRENT DEVELOPMENTS IN THE LABORATORY TO CLINICAL TREATMENTS

EDITED BY: Manoj Kumar Jaiswal PUBLISHED IN: Frontiers in Cellular Neuroscience

#### *Frontiers Copyright Statement*

*© Copyright 2007-2017 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-146-3 DOI 10.3389/978-2-88945-146-3

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## **THE ROLE OF MITOCHONDRIA, OXIDATIVE STRESS AND ALTERED CALCIUM HOMEOSTASIS IN AMYOTROPHIC LATERAL SCLEROSIS: FROM CURRENT DEVELOPMENTS IN THE LABORATORY TO CLINICAL TREATMENTS**

Topic Editor: **Manoj Kumar Jaiswal,** Columbia University, USA

Graphical Illustrations of Lau Gehrig portrait mixed with Motor Neurons, Glia, Astrocytes, Bipolar Neurons and other Cell types. Copyright: sunshine-91 , modified from "Vecteezy.com".

https://www.vecteezy.com/vector-art/115898-neuron-vector

Cover page: Graphical Illustrations of Human, ALS patients, Motor Neurons, Glia, Astrocytes, Bipolar Neurons and other Cell types. Copyright: dumbmichael, modified from "Vecteezy.com".

https://www.vecteezy.com/vector-art/115473-free-neuron-vectors

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive, devastating and fatal disease characterized by selective loss of upper and lower motor neurons of the cerebral cortex, brainstem, spinal cord and muscle atrophy. In spite of many years of research, the pathogenesis of ALS is still not well understood. ALS is a multifaceted genetic disease, in which genetic susceptibility to motor neuron death interacts with environmental factors and there is still no cure for this deleterious disease. At present, there is only one FDA approved drug, Riluzole which according to past studies only modestly slows the progression of the disease, and improves survival by up to three months. The suffering of the ALS patients, and their families is enormous and the economic burden is colossal. There is therefore a pressing need for new therapies.

Different molecular pathways and pathological mechanisms have been implicated in ALS. According to past studies, altered calcium homeostasis, abnormal mitochondrial function, protein misfolding, axonal transport defects, excessive production of extracellular superoxide radicals, glutamate-mediated excitotoxicity, inflammatory events, and activation of oxidative stress pathways within the mitochondria and endoplasmic reticulum can act as major contributor that eventually leads to loss of connection between muscle and nerve ultimately resulting to ALS. However, the detailed molecular and cellular pathophysiological mechanisms and origin and temporal progression of the disease still remained elusive. Ongoing research and future advances will likely advance our improve understanding about various involved pathological mechanism ultimately leading to discoveries of new therapeutic cures. Importantly, clinical biomarkers of disease onset and progression are thus also urgently needed to support the development of the new therapeutic agents and novel preventive and curative strategies. Effective translation from pre-clinical to clinical studies will further require extensive knowledge regarding drug activity, bioavailability and efficacy in both the pre-clinical and clinical setting, and proof of biological activity in the target tissue.

During the last decades, the development of new therapeutic molecules, advance neuroimaging tools, patient derived induced stem cells and new precision medicine approaches to study ALS has significantly improved our understanding of disease. In particular, new genetic tools, neuroimaging methods, cellular probes, biomarker study and molecular techniques that achieve high spatiotemporal resolution have revealed new details about the disease onset and its progression. In our effort to provide the interested reader, clinician and researchers a comprehensive summaries and new findings in this field of ALS research, hereby we have created this electronic book which comprises of twenty seven chapters having various reviews, perspective and original research articles. All these chapters and articles in this book not only summarize the cutting-edge techniques, approaches, cell and animal models to study ALS but also provide unprecedented coverage of the current developments and new hypothesis emerging in ALS research. Some examples are novel genetic and cell culture based models, mitochondria-mediated therapy, oxidative stress and ROS mechanism, development of stem cells and mechanism-based therapies as well as novel biomarkers for designing and testing effective therapeutic strategies that can benefit ALS patients who are at the earlier stages in the disease.

I am extremely grateful to all the contributors to this book and want to thank them for their phenomenal efforts.

Manoj Kumar Jaiswal, Ph.D. February 5, 2017 New York, NY

**Citation:** Jaiswal, M. K., ed. (2017). The Role of Mitochondria, Oxidative Stress and Altered Calcium Homeostasis in Amyotrophic Lateral Sclerosis: From Current Developments in the Laboratory to Clinical Treatments. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-146-3

# Table of Contents


*126 Activation of the endoplasmic reticulum stress response in skeletal muscle of G93A\*SOD1 amyotrophic lateral sclerosis mice*

Dapeng Chen, Yan Wang and Eva R. Chin


Gabriella Dobrowolny, Camilla Bernardini, Martina Martini, Mirko Baranzini, Marta Barba and Antonio Musarò


Katharina A. Quinlan, Jonathan B. Lamano, Julienne Samuels and C. J. Heckman


Ellya Bukharaeva, Anastasia Shakirzyanova, Venera Khuzakhmetova, Guzel Sitdikova and Rashid Giniatullin

*236 NSC-34 Motor Neuron-Like Cells Are Unsuitable as Experimental Model for Glutamate-Mediated Excitotoxicity*

Blandine Madji Hounoum, Patrick Vourc'h, Romain Felix, Philippe Corcia, Franck Patin, Maxime Guéguinou, Marie Potier-Cartereau, Christophe Vandier, Cédric Raoul, Christian R. Andres, Sylvie Mavel and Hélène Blasco

*248 Adducin at the Neuromuscular Junction in Amyotrophic Lateral Sclerosis: Hanging on for Dear Life*

Charles Krieger, Simon Ji Hau Wang, Soo Hyun Yoo and Nicholas Harden

*260 Gene expression profiling for human iPS-derived motor neurons from sporadic ALS patients reveals a strong association between mitochondrial functions and neurodegeneration*

Chrystian J. Alves, Rafael Dariolli, Frederico M. Jorge, Matheus R. Monteiro, Jessica R. Maximino, Roberto S. Martins, Bryan E. Strauss, José E. Krieger, Dagoberto Callegaro and Gerson Chadi

*285 ALS Patient Stem Cells for Unveiling Disease Signatures of Motoneuron Susceptibility: Perspectives on the Deadly Mitochondria, ER Stress and Calcium Triad*

Anjoscha Kaus and Dhruv Sareen

*311 Dysregulated expression of death, stress and mitochondrion related genes in the sciatic nerve of presymptomatic SOD1G93A mouse model of Amyotrophic Lateral Sclerosis*

Chrystian J. Alves, Jessica R. Maximino and Gerson Chadi

*331 Promise and Pitfalls of Mitochondrial Replacement for Prevention and Cure of Heritable Neurodegenerative Diseases Caused by Deleterious Mutations in Mitochondrial DNA*

Ananta Paine and Manoj Kumar Jaiswal

## Calcium, mitochondria, and the pathogenesis of ALS: the good, the bad, and the ugly

#### *Manoj Kumar Jaiswal\**

*Department of Anatomy, Physiology, and Genetics, School of Medicine, USUHS, Bethesda, MD, USA \*Correspondence: manoj.jaiswal.ctr@usuhs.edu*

*Edited by:*

*Egidio D'Angelo, University of Pavia, Italy*

**Keywords: amyotrophic lateral sclerosis, calcium homeostasis dysregulation, mitochondrial dysfunction, neurodegeneration, calcium buffer, selective vulnerability, motor neuron disease**

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease characterized by the selective and progressive loss of upper and lower motor neurons (MNs) of the brainstem, spinal cord and cerebral cortex. Key discoveries in the field of ALS have exponentially increased since it was first described in 1869 by Jean Martin Charcot, and the identification in 1993 of the first ALS-causing mutations in the gene encoding the well-studied antioxidant enzyme Cu, Zn superoxide dismutase (SOD1). The etiology of sporadic ALS largely remains unknown and the mechanisms of MN degeneration are still being investigated. The only FDA drug approved for the treatment of ALS, riluzole only modestly benefits patients, but multiple drugs are currently in the development pipeline and in human ALS clinical trials. It has become obvious that the lack of understanding of the precise mechanisms of motor neuron degeneration presents a major obstacle in the development of effective therapies for ALS. Altered calcium homeostasis and calcium signaling pathway activation is one potential mechanism that accounts for at least three major and interrelated toxic pathways: oxidative stress, mitochondrial dysfunction and neuroinflammation in neurodegenerative diseases such as ALS.

The aim of this special topic was to determine the role of cellular Ca2<sup>+</sup> homeostasis and mitochondrial signaling pathways in selectively vulnerable MNs in ALS. In fact, recent evidence suggests that abnormalities in cellular Ca2<sup>+</sup> signaling are common features in the pathogenesis of a range of neurodegenerative disorders, including ALS. It is well known that Ca2<sup>+</sup> is one of the most relevant intracellular messengers, being essential in neuronal development, synaptic transmission and plasticity, as well as in the regulation of various metabolic pathways in the brain level. In subtypes of ALS associated with the SOD1 mutation and the sporadic disease, there have been several reports indicating that the involvement of mitochondria in pathogenesis includes the generation of intracellular free radical species, ultrastructral changes in mitochondrial morphology, swollen and vacuolar mitochondria and increased activity of complex I, III and IV in the Upper and lower MNS, frontal cortex and spinal cord (von Lewinski and Keller, 2005). Impaired spinal cord and vulnerable individual spinal MNs have also been reported. It has been proposed that this damage triggers the functional decline of MNs and the onset of pathology in ALS. The loss of mitochondrial membrane potential, excitotoxic stimulation of AMPA/kainite receptors and age related MN injury reported by many groups may contribute to ALS pathogenesis.

Further evidence for the involvement of disruption of intracellular Ca2<sup>+</sup> homeostasis arises from several studies of cellular and experimental animal models of ALS in which Ca2<sup>+</sup> binding proteins such as calbindin-D28K and parvalbumin in MN populations are lost early (hypoglossal, spinal, and lower cranial MNs). These findings are in good agreement with a quantitative comparison of Ca2<sup>+</sup> homeostasis where low cytosolic Ca2<sup>+</sup> buffering capacity acts as an important risk factor for degeneration and in contrast an increase in cytosolic Ca2<sup>+</sup> buffering capacity could protect vulnerable MNs from degeneration both *in-vitro* and *invivo*. Several lines of evidence support altered Ca2<sup>+</sup> homeostasis leading to MN degeneration in ALS, notably disturbance of glutamate neurotransmission and subsequent glutamate triggered Ca2<sup>+</sup> entry, increased extracellular glutamate levels probably due to reduced glial glutamate uptake caused by oxidative damage to EAAT2, study of fALS in cell lines and animal mouse models where potential mechanism for Ca2+disruption is inhibition of glial glutamate transport by mSOD1 similar to those proposed for sALS. Evidence also suggests that ROS generated in MNs can cross the plasma membrane and cause oxidative disruption of glutamate transporters in neighbouring astrocytes. However, in ALS pathology, the nature of dysfunction is highly controversial after recent findings where non-cell autonomous effect of glia on MNs in an embryonic stem cell-based ALS model and astrocytes expressing ALS-linked mSOD1 that release factors selectively toxic to MNs was shown. The data currently available so far leave several important questions unanswered; these include: (1) does the presence of mSOD1 cause morphological abnormalities of mitochondria when expressed at physiological levels? (2) Does the mitochondrial Ca2<sup>+</sup> sequestration source specificity and spatiotemporal properties of [Ca2+]i signaling varies at physiological levels? (3) What are the functional consequences of any changes in mitochondrial function on Ca2<sup>+</sup> homeostasis in the presence of mutant SOD1 gene?

Combining the lessons from multiple animal models as well as cell culture model used to determine pathogenic mechanisms in ALS, the central insight is that selective vulnerability of MNs likely arises from a combination of several mechanisms; two of them, mitochondrial dysfunction and Ca2<sup>+</sup> homeostasis are prominent (Jaiswal and Keller, 2009; Jaiswal et al., 2009). Documenting the involvement and importance of mitochondrial dysfunction and Ca2<sup>+</sup> homeostasis we believe that MN possess large number of voltage and ligand gated Ca2<sup>+</sup> channels that, when activated, cause rapid Ca2<sup>+</sup> influx, which, in part because of relatively weak cytosolic Ca2<sup>+</sup> buffering, results in mitochondrial Ca2<sup>+</sup> overload and strong ROS generation. Alternatively, the extrusion mechanisms, characterized by the plasma membrane calcium ATPase, Na+/Ca2<sup>+</sup> exchanger and accumulation by intracellular organelles contribute to the removal of calcium ions. Furthermore, their selective loss goes a long way explaining the oxidative damage, mitochondrial abnormalities and apoptotic contributions observed in ALS MNs. Chronic mitochondrial membrane depolarisation due to Ca2<sup>+</sup> entry can cause the release of pro-apoptotic proteins and activate enzymes involved in apoptotic pathways.

In the last few years, attention was drawn to the role of mitochondria as an efficient regulator of cytosolic calcium signals. Further studies with mitochondria targeted calcium probes indicate a rapid, dramatic increase in free intramitochondrial calcium. This uptake by mitochondria has an immense effect on the metabolic state of the cell as it can up regulate the activity of the enzymes in oxidative metabolism. The malformations in mitochondrial structure and massive vacuoles derived from degenerating mitochondria found in post mortem human ALS samples further strengthen the proposition. Although the exact molecular mechanism is still not known, we hypothesize that vulnerability to ALS is a consequence of specific physiological features, particularly highly specialized Ca2<sup>+</sup> homeostasis, continuous activity-dependent Ca2+cycling and the predominant role of mitochondria in buffering Ca2<sup>+</sup> transients. Definitive evidence for this will have to wait for further studies. Nevertheless, if Ca2<sup>+</sup> homeostasis and mitochondrial defects were found to play a role in disease pathogenesis, this would have broad ranging implications for therapy. Further studies will therefore be of considerable interest in attempting to understand the pathogenesis of cell death in ALS.

#### **ACKNOWLEDGMENTS**

The author would like to thank those individuals that were involved in conducting the unpublished studies discussed in this editorial. Thanks to Derek Holman for his insightful comments.

### **REFERENCES**


*Received: 19 August 2013; accepted: 10 October 2013; published online: 31 October 2013.*

*Citation: Jaiswal MK (2013) Calcium, mitochondria, and the pathogenesis of ALS: the good, the bad, and the ugly. Front. Cell. Neurosci. 7:199. doi: 10.3389/fncel.2013.00199 This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2013 Jaiswal. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The ER mitochondria calcium cycle and ER stress response as therapeutic targets in amyotrophic lateral sclerosis

#### *Vedrana Tadic\*,Tino Prell, Janin Lautenschlaeger and Julian Grosskreutz*

Hans Berger Department of Neurology, Jena University Hospital, Jena, Germany

#### *Edited by:*

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### *Reviewed by:*

Tibor Kristian, University of Maryland School of Medicine, USA Alexej Verkhratsky, University of Manchester, UK Pavle R. Andjus, University of Belgrade, Serbia Anthony Robert White, The University of Melbourne, Australia

#### *\*Correspondence:*

Vedrana Tadic, Hans Berger Department of Neurology, Jena University Hospital, Erlanger Allee 101, 07747 Jena, Germany e-mail: vedrana.tadic@ med.uni-jena.de

#### **INTRODUCTION**

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by progressive degeneration of the upper (spasticity, dysphagia, dysarthria) and lower motor neurons (atrophy, fasciculations). Approximately 90% of ALS patients have sporadic ALS (sALS) which is the most prevalent form and about 10% have the inherited or familial form of ALS (fALS). The latter form is believed to be due to several genes including *SOD1*,

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by progressive loss of upper and lower motor neurons. Although the etiology remains unclear, disturbances in calcium homoeostasis and protein folding are essential features of neurodegeneration in this disorder. Here, we review recent research findings on the interaction between endoplasmic reticulum (ER) and mitochondria, and its effect on calcium signaling and oxidative stress. We further provide insights into studies, providing evidence that structures of the ER mitochondria calcium cycle serve as a promising targets for therapeutic approaches for treatment of ALS.

**Keywords: amyotrophic lateral sclerosis, ER stress, protein misfolding, calcium dysregulation, SOD1, TDP-43, mitochondria, oxidative stress**

> *TARDBP, FUS, OPTN,* and *VCP*. In addition, a hexanucleotide (GGGGCC) repeat expansion in the first intron of the *C9ORF72* gene (DeJesus-Hernandez et al., 2011; Renton et al., 2011) has lately been demonstrated as being associated with ALS. However, the etiology of the disease is still unclear, although recent studies indicate that calcium (Ca2+) disturbances, ER stress, and mitochondrial dysfunction are involved in the pathogenesis of ALS (Grosskreutz et al., 2010; Lautenschlaeger et al., 2012). Other mechanisms possibly involved in ALS-related pathophysiology comprise: oxidative stress, protein aggregation, dysregulated endosomal trafficking, impaired axonal transport, neuroinflammation, and dysregulated transcription and RNA processing (Ferraiuolo et al., 2011). Several properties of motor neurons make them more vulnerable than other neuronal groups. Motor neurons express high levels of Ca2+–permeable α-amino-5-methyl-3 hydroxisoxazolone-4-propionate (AMPA) receptors that lack the GluR2 subunit which makes them more vulnerable to excitotoxicity and dysregulation of intracellular Ca2<sup>+</sup> homeostasis (Williams et al., 1997). Also, low levels of Ca2<sup>+</sup> -buffering proteins contributes greatly to this vulnerability (Ince et al., 1993). Because of high metabolic demands, motor neurons are largely dependent on optimal mitochondrial function, a robust cytoskeleton and an axonal transport mechanism. Despite all the above facts, there remain numerous unanswered questions in ALS related to selectivity and specificity of the cellular targets of motor neuron degeneration and cell-specific aspects of mitochondrial Ca2<sup>+</sup> signaling. This review focuses on crosstalk between ER, mitochondria, oxidative stress and calcium.

**Abbreviations:** ALS, amyotrophic lateral sclerosis; AMPA, α-amino-5-methyl-3-hydroxisoxazolone-4-propionate; ANT, adenine nucleotide translocator; AP-1, activator protein 1; ARE, antioxidant response element; ATF6, basic leucine-zipper transcription factor 6; Bax/Bak, Bcl-2-associated X protein; Bcl-2, B-cell lymphoma 2 protein; BIK, Bcl-2 interacting killer protein; *C9ORF72*, chromosome 9 open reading frame 72; CHOP, transcription factor C/EBP homologous protein; eIF2α, eukaryotic initiation factor-2; ER, endoplasmic reticulum; ERMCC, endoplasmic reticulum mitochondria calcium cycle; fALS, familial amyotrophic lateral sclerosis; FUS/TSL, fused in sarcoma/translated in liposarcoma; HIF-1α, hypoxia-induced factor; Hsf1, heat shock transcription factor 1; InsP3, inositol 1,4,5-trisphosphate; IP3R, inositol 1,4,5-triphosphate receptor-gated channel; IRE1, inositol-requiring enzyme 1; mNCE, mitochondrial sodium calcium exchanger; mPTP, mitochondrial permeability transition pore; mUP, mitochondrial uniporter; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; NO, nitric oxide; Nrf2, erythroid 2-related-factor 2; *OPTN*, optineurin; PDI, protein disulfide isomerase; PERK, the double-stranded RNA-activated protein kinase (PKR)-like ER kinase; PLCδ1, phospolipase C delta 1; ROS, reactive oxygen species; RyR, ryanodine receptors; sALS, sporadic amyotrophic lateral sclerosis; SERCA, sarco/endoplasmic reticulum Ca2<sup>+</sup> ATPase; SOD1, Cu/Zn superoxide dismutase type 1; SR, sarcoplasmic reticulum; *TARDBP*, TAR DNA binding protein; TCTP, translationally controlled tumor protein; TDP-43, transactive response DNA binding protein 43 kDa; UBQLN2, ubiquilin-2; UCH-L1, ubiquitin carboxy-terminal hydrolase L1; UPR, unfolded protein response; *VAPB*, vesicle-associated membrane protein (VAMP)-associated protein B; *VCP*, valosin-containing protein; VDAC, voltage-dependent anion channel.

#### **ALS GENES AND ENCODED PROTEINS – ROLE IN PATHOPHYISOLOGY** *Cu/Zn superoxide dismutase 1 (SOD 1)*

About 20% of fALS patients carry a mutation in the *SOD1* gene. Indeed, more than 170 different SOD1 mutations have been described in ALS families (http://alsod.iop.kcl.ac.uk/). Generally, *SOD1* mutations have not been linked to decreased SOD activity (Kostrominova, 2010; Fischer et al., 2011), instead mutant SOD1 likely acts through a combination of several mechanisms, including protein misfolding, mitochondrial dysfunction, oxidative damage, cytoskeletal abnormalities, and defective axonal transport, excitotoxicity, in addition to inadequate growth factor signaling and inflammation (Cozzolino et al., 2008).

#### *Fused in sarcoma (FUS) and TAR DNA-binding protein (TDP-43)*

Mutations in the gene encoding fused in sarcoma/translocated in liporsarcoma (FUS/TLS or FUS) are linked to 4% of fALS cases. Mutations in *TARDBP* (which encodes TDP-43) account for 4% of fALS and a smaller percentage of sALS. The toxicity of TDP-43 and FUS/TSL proteins is linked to their altered intracellular localization. Both TDP-43 and FUS/TLS are mainly localized in cell nuclei where they control gene transcription and pre-mRNA (Buratti et al., 2001; Winton et al., 2008; Kwiatkowski et al., 2009). In the presence of mutations or stress, these proteins accumulate in the cytosol (Liu-Yesucevitz et al., 2010; Dewey et al., 2012; Daigle et al., 2013; Walker et al., 2013; Watanabe et al., 2013). FUS triggers ER stress and causes fragmentation of the Golgi apparatus in patients with fALS (Farg et al., 2013). In NSC34 transfected cells, wild type human TDP-43 caused ER stress (Suzuki et al., 2011). In the same cell line, mutant TDP-43 induced mitochondrial dysfunction and probably caused oxidative stress (Duan et al., 2010). Experiments with TDP-43 mutation in zebrafish resulted in impairment of neuromuscular junctions (Armstrong and Drapeau, 2013).

#### *Chromosome 9 open reading frame 72 (C9ORF72)*

Large expansions of a non-coding GGGGCC-repeat in the first intron of the *C9ORF72* gene are accountable for 40% of fALS. C9ORF72 hexanucleotide repeats form highly stable RNA Gquadruplexes, which probably influence RNA transcription, splicing, translation and transport (Fratta et al., 2012). C9ORF72 pathology is characterized by intracellular inclusions, however the major proteins forming these inclusions have not yet been elucidated (Mori et al., 2013).

#### *Other genes*

Mutations in the valosin–containing protein (*VCP*) are responsible for 1–2% of fALS cases. In mice, overexpression of mutant VCP produces ubiquitin- and TDP-43-positive inclusions, suggesting that TDP-43 plays a role in VCP-induced disease (Rodriguez-Ortiz et al., 2013). Mutant VCP also impact mitochondria, such as via a decrease in ATP production related to mitochondrial uncoupling (Bartolome et al., 2013).

Another mutated gene found in patients with ALS comprises *OPTN* that encodes the protein optineurin which regulates membrane trafficking, protein secretion, cell division and host defense against pathogens (Kachaner et al., 2012). Wild-type optineurin suppresses nuclear factor-kappa B (NF-κB) activity, but the ALS-causing mutant optineurin is unable to suppress NF-κB activity. Therefore, there is an indication that inappropriate NF-κB activation is the pathogenic mechanism underlying optineurin mutation-related ALS (Akizuki et al., 2013). In two patients carrying mutation in *OPTN*, was shown that loss of function rather than proteinopathy itself resulted in the formation of TDP-43 inclusions in neuronal and glial cytoplasm, and Golgi apparatus fragmentation (Kamada et al., 2014).

Vesicle-associated membrane protein (VAMP)-associated protein B (VAPB) is usually ER-resident and is integral to its structure, protein transport, lipid metabolism, and the UPR. VAPB toxicity is probably mediated by impaired Ca2<sup>+</sup> homeostasis and ER stress (Langou et al., 2010; De Vos et al., 2012; Morotz et al., 2012).

#### **THE ERMCC AND CALCIUM DISTURBANCES IN ALS**

The ER and mitochondria form a highly dynamic interconnected network that is involved in the generation of Ca2<sup>+</sup> signals. Ca2<sup>+</sup> release from ER is controlled by ryanodine receptors (RyRs, Ca2+ gated Ca2<sup>+</sup> channels) (Meissner, 2002; Lanner et al., 2010), the inositol 1,4,5-triphosphate receptor-gated channels (IP3Rs), and the translocon (Taylor and Tovey, 2010). Restocking of the ER with Ca2<sup>+</sup> is executed by the sarco/endoplasmic reticulum Ca2<sup>+</sup> ATPase (SERCA; Wuytack et al., 2002; Verkhratsky, 2005; Lipskaia et al., 2009). Ultimately, the plasma membrane Na+/Ca2<sup>+</sup> exchanger and Ca2<sup>+</sup> ATPase remove Ca2<sup>+</sup> from the cell (Rhodes and Sanderson, 2009; **Figure 1**).

Mitochondria take up Ca2<sup>+</sup> via a Ca2+-sensitive electrogenic carrier, the mitochondrial uniporter (mUP) which is gated by cytosolic Ca2<sup>+</sup> in a biphasic-dependent manner (Gunter and Sheu, 2009). Ca2<sup>+</sup> uptake into mitochondria is facilitated by Ca2+/calmodulin. However, sustained cytosolic Ca2<sup>+</sup> levels inactivate the uniporter, preventing further Ca2<sup>+</sup> uptake (Moreau et al., 2006). Accumulated Ca2<sup>+</sup> in the mitochondria can slowly be ejected back into the cytosol through Na+/Ca2<sup>+</sup> and 2H+/Ca2<sup>+</sup> exchangers (Pivovarova and Andrews, 2010; **Figure 1**). Once intramitochondrial Ca2<sup>+</sup> rises above a certain threshold, the voltage- and Ca2+-dependent high-conductance channel in the inner membrane, known as the mitochondrial permeability transition pore (mPTP), opens, leading to cell death either by apoptosis or necrosis (Leung and Halestrap, 2008; Martin, 2010b). Mitochondria contain similar low Ca2<sup>+</sup> levels as resting cells, but accumulate a considerable amount during stimulated Ca2<sup>+</sup> entry, which affects numerous cellular processes such as cellular energy metabolism, synaptic transmission and excitability, intracellular signaling, generation of ROS, and activation of apoptosis (Chinopoulos and Adam-Vizi, 2010; Starkov, 2010).

Several studies have previously investigated abnormalities of Ca2<sup>+</sup> homeostasis, ER and mitochondria as well as excitotoxicity in motor neurons in ALS (Grosskreutz et al., 2010; Lautenschlager et al., 2013). Based on the models described by Berridge (2002), a persistent shift of Ca2<sup>+</sup> from the ER to mitochondria (i.e., through Ca2+induced Ca2<sup>+</sup> release via RyR and mitochondrial uptake through mUP) was postulated. This could be triggered by the physiological activity of AMPA receptors together with a pathologically increased Ca2+-permeability (Grosskreutz et al., 2010). This in turn, leads to a depletion of Ca2<sup>+</sup> levels in the ER,

resulting in protein folding dysfunction and chronic mitochondrial Ca2<sup>+</sup> overload. Both protein misfolding and Ca2<sup>+</sup> overload can then induce apoptosis through Bcl-2 dependent mechanisms (Grosskreutz et al., 2010). Since Ca2<sup>+</sup> appears to be shuttled back and forth between the ER and the mitochondrial compartment, the process has been termed the ER–mitochondria Ca2<sup>+</sup> cycle (ERMCC, **Figure 1**; Grosskreutz et al., 2010).

#### **IMPACT OF ER STRESS ON MITOCHONDRIA**

Recent studies indicate that ER stress is involved in the pathogenesis of familial and sporadic ALS (Ilieva et al., 2007; Atkin et al., 2008; Walker, 2010; Lautenschlaeger et al., 2012; Prell et al., 2012). ER stress occurs when ER Ca2<sup>+</sup> content is depleted (Verkhratsky, 2005) and misfolded proteins accumulate in the ER. To cope with ER stress, cells activate the unfolded protein response (UPR; **Figure 2**). The UPR mediates the (1) upregulation of genes encoding ER-resident chaperones, (2) down-regulation of general protein synthesis in order to reduce the ER protein load, and, (3) degradation of misfolded proteins by the proteasome (Kozutsumi et al., 1988; Yoshida et al., 2001; Dudek et al., 2009; **Figure 2**). On a cellular level, ER stress is transduced by three proximal sensors of the UPR: the double-stranded RNA-activated protein kinase

(PKR)-like ER kinase (PERK), the basic leucine-zipper transcription factor 6 (ATF6) and the inositol requiring enzyme 1 (IRE1) (Prell et al., 2013; **Figure 2**). When protein misfolding can no longer be compensated for, the prolonged UPR triggers apoptosis by the caspase pathways (Nakagawa et al., 2000; **Figure 2**).

ER stress can affect mitochondria because both organelles are functionally and morphologically connected by several pathways (Vannuvel et al., 2013). In particular, the contact between ER and mitochondria is essential for coordination of the Ca2<sup>+</sup> transfer (Rowland andVoeltz, 2012). The proteins B-cell lymphoma 2 (Bcl-2), Bcl-2-associated X protein (Bax/Bak) and the Bcl-2-interacting killer (BIK) can enhance Ca2<sup>+</sup> transfer from ER to mitochondria and the ensuing Ca2<sup>+</sup> accumulation activates apoptosis via cytochrome C (Germain et al., 2002; Nutt et al., 2002; Fannjiang et al.,2004;Kong et al.,2005). The PERK/ATF4 pathway can induce Lon protease that controls the assembly and/or the degradation of cytochrome C (Margineantu et al., 2002; Venkatesh et al., 2012). Bid, which is a pro-apoptotic BH3-only protein can be cleaved upon ER stress, which subsequently activates caspase-2 or caspase-8 leading to apoptosis (Upton et al., 2008; Uchibayashi et al., 2011). Recruitment of the dynamin-related protein 1 mediates the fission of the outer mitochondrial membrane (Breckenridge et al., 2003).

2012).

Grp78 dissociates from IRE1, PERK and ATF6 resulting in the activation and up-regulation of these three UPR sensor proteins. IRE1 splices XBP-1 mRNA (sXBP-1) which up-regulates the expression of UPR relevant genes. ATF6 is transported to Golgi apparatus, where it is cleaved. The cytosolic domain of

#### **THERAPEUTIC STRATEGIES FOCUSED ON ER STRESS**

Several therapeutic strategies aim to target the ER (**Table 1**; **Figure 4**). For instance, Salubrinal is a substance that reduces ER stress by activating the UPR. UPR activation is mediated via phosphorylation of elF2α and activation of PERK (Boyce et al., 2005). Salubrinal prevented neuronal cell death triggered by several ER stress inducers (Smith et al., 2005; Reijonen et al., 2008). Moreover, Salubrinal, Guanabenz and Phenazine have all been shown to reduce ER stress in worms and zebrafish expressing mutated TDP-43 (Vaccaro et al., 2013). In SOD1G93A mice, Salubrinal decreased muscle strength loss and extended survival (Saxena et al., 2009).

Another approach to target the ER is to encourage the natural cellular protein-folding machinery via activation of the heat shock transcription factor 1 (Hsf1). Hsf1 is the master activator of chaperone protein gene expression (Neef et al., 2011). Overexpression of human molecular chaperone hHSJ1a

*in vivo* mediated late-stage neuroprotection in the SOD1G93A mouse model, probably through a combination of chaperone, co-chaperone and pro-ubiquitylation activity on SOD1 (Novoselov et al., 2013). Arimoclomol, a hydroxylamine derivate and a co-inducer of the heat shock response delayed muscle denervation in the SOD1G93A mouse followed by a rise in expression of the heat shock protein 70 (Kalmar et al., 2012). The therapeutic potential of this drug is under investigation in a phase II/III clinical trial for ALS patients with SOD1 mutations (http://www.clinicaltrials.gov/ct2/show/NCT00706147? term=arimoclomol&rank=1). Other available pharmaceuticals that up-regulate heat shock response and that may be used to treat ALS include Celastrol and 17-AAG (Kalmar et al., 2014).

Bcl-2 family members are down-regulated, while pro-apoptotic Bcl-2 family members are up-regulated. Picture modified from (Lautenschlaeger et al.,

Accumulation of misfolded proteins may also be targeted by small molecule regulators of autophagy such as antipsychotics (fluspirilene, trifluoperazine, pimozide) and calcium-channel


modulators (nicardipine, niguldipine, amiodarone; Sarkar et al., 2007; Zhang et al., 2007).

The Sigma-1 receptors have also gained attention in the recent past. The receptors are a chaperone proteins residing at the mitochondrion-associated ER membrane, where they affect mitochondrial Ca2<sup>+</sup> influx by stabilizing IP3R and acting as interorganelle signaling modulators of Ca2<sup>+</sup> homeostatasis, ER stress and apoptosis. The Sigma-1 receptor agonist PRE-084 prevented neurons loss in SOD1G93A transgenic mice, probably by the activation of protein kinase C and reducing microglia activation (Mancuso et al., 2012). Neuroprotective effects of PRE-084 have also been demonstrated in the wobbler mouse model not linked to SOD1 mutation that is characterized by progressive neural atrophy shortly after birth (Peviani et al., 2014). Another agonist of Sigma-1 receptor SA4503 prevented SOD1G93A–induced neurotoxicity in NSC34 cells and extended survival of SOD1G93A mice (Ono et al., 2014). Pharmacological manipulation of the Sigma-1 receptor may increase availability of growth factors, as well as the modulation of astrocytosis and of macrophage/microglia as part of the mechanism involved in Sigma 1 receptor-mediated neuroprotection (Peviani et al., 2014).

Treatment with Geldanamycin, an inducer of heat shock response, successfully blocked protein aggregation but not Ca2<sup>+</sup> dysregulation or loss of mitochondrial membrane potential (-Ψ) in murine motor neurons expressing human SOD1G93A (Tradewell et al., 2011). This implies chaperone-based therapies would possibly require co-therapy targeting other important mechanisms of toxicity.

### **CROSS-TALK BETWEEN CALCIUM, MITOCHONDRIA, AND REACTIVE OXYGEN SPECIES SIGNALING**

#### **MITOCHONDRIAL DYSFUNCTION**

Mitochondria are central for energy metabolism and have been well studied in relation to ALS pathogenesis (von Lewinski and Keller, 2005; Cozzolino and Carri, 2012; Jaiswal, 2013). The levels of mutated mitochondrial DNA (mtDNA) were higher in ALS patients but the amount of mtDNA was reduced compared to controls. This reduction correlated well with a decrease of a mitochondrial marker, citrate synthase activity and with the activities of respiratory chain complexes I + III, II + III, and IV, suggesting a loss of mitochondria in ALS spinal cords (Wiedemann et al., 2002). Activity of cytochrome C oxidase in mitochondria is reduced in the spinal cord of sALS patients (Borthwick et al., 1999) and ALS spinal neurons show varied and reduced mtDNA gene copy numbers and increased mtDNA gene deletions (Keeney and Bennett, 2010). Oxidative stress, protein nitration and aggregation, and excitotoxicity participate in the process of motor neuron degeneration caused by mutated SOD1 (Martin et al., 2007). One of the pathological hallmarks of ALS is aggregation of ubiquitinated proteins in motor neurons (Stieber et al., 2000; **Figure 3**). SOD1, FUS, TDP-43, OPTN, and UBQLN2 have been identified asforming aggregates.Whether there is a causal relationship between misfolded proteins and mitochondrial dysfunction for novel mutations is still largely unknown, but there is considerable body of literature describing SOD1 and more recently TDP-43. Mutant SOD1 forms insoluble aggregates in mitochondria at the surface of the outer membrane (**Figure 3**). Further, there

is direct connection between mutated SOD1 and impaired mitochondrial function (Liu et al., 2004; Pasinelli et al., 2004; Pickles et al., 2013). Bcl-2 has been identified as an interacting partner of mutated SOD1 because SOD1 induces mitochondrial morphological changes and impairs mitochondrial membrane integrity only in the presence of Bcl-2. This leads to the release of cytochrome C, ultimately leading to cell death (Pedrini et al., 2010).

Degenerating mitochondrial vacuoles have been reported in presymptomatic mice expressing mutant SOD1 in previous studies (Wong et al., 1995; Kong and Xu, 1998). However, mitochondrial disturbances are not restricted to SOD1 mutations. In patients with ALS, dense conglomerates of mitochondria have been found in the anterior horn of lumbar and spinal cord and proximal axons (Hirano et al., 1984; Sasaki and Iwata, 1996). It has been demonstrated that neuronal Ca2+, mitochondrial volume and a number of synaptic vesicles are increased in ALS patients (Siklos et al., 1996). In addition, overexpression of TDP-43 causes mitochondrial dysfunction and induces mitophagy (Hong et al., 2012) and oxidative injury in NSC34 cell line (Duan et al., 2010; Lu et al., 2012). In a yeast model, TDP-43 aggregates around mitochondria and there is an inverse correlation between respiratory activity and toxicity of the mutant protein (Braun et al., 2011). Overexpression of wild-type TDP-43 resulted in reduced mitochondrial length and density in neurites of primary motor neurons and conversely, suppression of TDP-43 resulted in significantly increased mitochondrial length and density in neurites (Wang et al., 2013). Neuronal mitochondrial transport and morphological abnormalities occur *in vivo* in SOD1 and TDP-43 ALS (**Figure 3**) mouse models but show differences in temporal and spatial manifestation. This implies that different molecular mechanisms may be involved (Magrane et al., 2013).

#### **MITOCHONDRIAL CALCIUM DYNAMICS**

#### *Ca***2<sup>+</sup>** *regulation of mitochondrial metabolism*

Ca2<sup>+</sup> plays a central role in cell signaling at numerous levels. The tricarboxylic acid cycle consists of a series of reactions that produce energy through the breakdown of proteins, fatty acids and carbohydrates. Ca2<sup>+</sup> within mitochondria regulates the most important task of the organelle: ATP production by oxidative phosphorylation. The physiological increase of mitochondrial Ca2<sup>+</sup> stimulates the adenine nucleotide transporter (Mildaziene et al., 1995) and synthesis of ATP complex V (Das and Harris, 1990). Moreover, mitochondrial Ca2<sup>+</sup> increase activates three matrix dehydrogenases: isocitrate dehydrogenase, α-ketogluterate dehydrogenase and pyruvate dehydrogenase (McCormack and Denton, 1979, 1993; McCormack et al., 1990). All three dehydrogenases enhance the reaction rate of many of the steps in the tricarboxylic acid cycle and therefore increase flux throughout the pathway, raising ATP production (Jouaville et al., 1999). Furthermore, it has been shown that motor neurons have an insufficient mitochondrial capacity to buffer large Ca2<sup>+</sup> elevations which is partly due to a reduced mitochondrial density per volume compared to non-motor neurons (Grosskreutz et al., 2007). Mitochondrial disfunction and impaired Ca2<sup>+</sup> homeostasis largely contribute to selective vulnerability of motor neurons (Jaiswal et al., 2009; Jaiswal and Keller, 2009).

#### *Ca***2<sup>+</sup>** *overload and activation of permeability transition pore*

One main mediator of mitochondrial function or dysfunction in neurons is the mPTP, which is a Ca2<sup>+</sup> dependent highconductance channel in the inner membrane of mitochondria (Brenner and Moulin, 2012). The mPTP comprises of the voltage-dependent anion channel (VDAC), the adenine nucleotide translocator (ANT) and cyclophilin D. Since VDAC and ANT are not essential for functioning of mPTP regulator (Juhaszova et al., 2008), the soluble matrix protein cyclophilin D received special attention (Giorgio et al., 2010). The mPTP opening is promoted by binding of cyclophilin D to the inner mitochondrial membrane (Di Lisa and Bernardi, 2009) and is favored by Ca2<sup>+</sup> overload, ROS, inorganic phosphate and mitochondrial depolarization (Crompton, 1999; Brustovetsky et al., 2002; Bernardi et al., 2006). Binding of cyclophilin D to the inner mitochondrial membrane can be prevented by Cyclosporine A. Opening the mPTP causes a release of cytochrome C, which subsequently leads to apoptotic cell death (Liu et al., 1996; Crompton et al., 1998). High concentrations of cyclophylin D were found in swollen mitochondria in the SOD1 animal model. Modifying mPTP through different genetic and pharmacological manipulations has been shown to be protective in animal models of

ALS. Genetic ablation of cyclophilin D delayed disease onset and extended the lifespan in the ALS mouse model (Martin et al., 2009). However, in another study, deleting cyclophilin D in the SOD1 mouse model did not lead to prolongation of survival, although it improved mitochondrial buffering capacity and attenuated mitochondrial damage (Parone et al., 2013). Therefore the role of cyclophylin D as a potential therapeutic is not fully understood.

#### *Therapeutic strategies focused on mitochondrial Ca***2<sup>+</sup>** *dynamics*

Because there is growing evidence for mitochondrial dysfunction in ALS, mitochondria are promising therapeutic targets (**Table 2**; **Figure 4**). However, studies targeting mitochondria have failed so far. The mPTP modulator Olesoxime (TRO19622) had a neuroprotective effect in motor neuron cell culture and in ALS rodents (Bordet et al., 2007; Martin, 2010a; Sunyach et al., 2012), but failed in a phase III clinical trial (http://www.als.net/ALS-Research/Olesoxime/ALS-Topics/). Further, Dexpramipexole, which reduces mPTP opening and increases cellular energy supply, did not have significant effects on survival and disease progression in a recent clinical trial (Cudkowicz et al., 2013).


**Table 2 |**

**Substances**

 **targeting** 

**mitochondria.**

Minocycline and creatine, compounds that improve mitochondrial function have also failed in human trials (Shefner et al., 2004; Gordon et al., 2007). Other medications targeting mitochondria such as coenzyme Q and cylosporin A have been studied, but all trials in humans were negative (Appel et al., 1988; Kaufmann et al., 2009). Uridine, a pyrimidine nucleoside, extended survival in SOD1G93A mice, probably by improving bioenergetic effects, increasing ATP levels, and enhancing glycolytic energy production (Amante et al., 2010). However, it has not been tested in ALS patients yet.

Melatonin recently gained interest because of its ability to decrease cytochrome C release and caspase-3 activation. It delayed disease onset in the SOD1G93A mice model. Besides its mitochondria stabilization effects in ALS, melatonin attenuated the activation of astrocytes and microglia (Zhang et al., 2013).

CGP37157, which is able to cross blood-brain barrier (Gonzalez-Lafuente et al., 2012), blocks the mNCE. It showed protective effects against kainate induced excitotoxicity in SOD1G93A mice motor neurons (Lautenschlager et al.,2013), restored calcium levels in SOD1G37R N2 cells (Coussee et al., 2011) and protected rat hippocampal slices upon veratridine-induced sodium and calcium overload (Gonzalez-Lafuente et al., 2012).

The transport of ADP, ATP and inorganic phosphates across mitochondrial membranes is regulated by the VDAC at the outer mitochondrial membrane. VDAC is regulated by Bcl-2 and both can form toxic complexes with mutated SOD1 (Pasinelli et al., 2004; Arbel and Shoshan-Barmatz, 2010; Israelson et al., 2010; Pedrini et al., 2010). Interactions between VDAC1, Bcl-2 and mutated SOD1 inhibits the conductance of VDAC1, leading to cell death (Israelson et al., 2010; Pedrini et al., 2010). It was demonstrated that small SOD1-like therapeutic peptides specifically block the formation in symptomatic SOD1G93A mice by restoring mitochondrial ADP permeability (Tan et al., 2013).

Human TDP-43 caused mitochondrial morphologic abnormality and decrease of mitochondrial complex I activity, and mitochondrial transmembrane potential in human TDP-43 stably tranfected NSC-34 cells. Dimethoxy curcumin was able to ameliorate mitochondrial dysfunction in the same experiment, which makes this drug interesting as a potential therapeutic for TDP-43 linked ALS (Lu et al., 2012).

#### **CALCIUM AND MITOCHONDRIAL ROS**

Mitochondria are the main sites of ROS formation as by-products of ATP production (Coyle and Puttfarcken, 1993; Brand, 2010). However, mitochondrial Ca2<sup>+</sup> overload and abnormal oxidative phosphorylation increase ROS production and oxidative stress (Carriedo et al., 2000; Mattiazzi et al., 2002; Murphy, 2009; **Figure 3**). Many of the stated mitochondrial respiratory abnormalities have been linked to reduced activity of mitochondrial complexes I and IV (Mattiazzi et al., 2002; Rizzardini et al., 2006; Son et al., 2008; Coussee et al., 2011).

#### *Reactive oxygen species and oxidative stress*

Moderate levels of ROS and reactive nitrogen species (RNS) promote cellular proliferation, regulation and survival. Typical ROS are free radical species such as superoxide (O•− <sup>2</sup> ), hydroxyl radicals (•OH) and non-radical species like hydrogen peroxide (H2O2). The respiratory chain complexes I and III are the primary mitochondrial sources of O•− <sup>2</sup> Oxidative stress has been implicated as a pathological mechanism of both fALS and sALS (Ferrante et al., 1997). ROS has been detected in the spinal cord and cerebrospinal fluid of sALS patients (Tohgi et al., 1999). Increased H2O2 and oxidative damage to protein and DNA were observed in mutated SOD1 transgenic mice (Liu et al., 1999). Many ALS causing genes and genes modifiers are known to influence ROS production (Carter et al., 2009). SOD1 mutation induces oxidative modifications of several proteins in ALS: SOD1, translationally controlled tumor protein (TCTP), ubiquitin carboxyl-terminal hydrolase-L1 (UCH-L1) and probably alphaB-crystallin. These oxidative modifications lead to structural alteration and a decline of protein activity (Poon et al., 2005). ROS also directly influences transcriptional factors such as NF-kB, activator protein 1 (AP-1), and HIF-1α which are all involved in the regulation of gene expression and maintaining cellular homeostasis (Haddad, 2002).

ROS dysregulation of these factors is observed in ALS pathology (Iaccarino et al., 2011; Moreau et al., 2011) as the involvement of protein disulfide isomerase (PDI) family members plays an important role in oxidative folding of human secretory proteins (Rutkevich et al., 2010; **Figure 2**). PDI's enzymatic activity can be inactivated by oxidation and *S*-nitrosylation of their active site thiol groups. In motor neurons of patients with ALS, PDI was widely distributed and aggregated (Atkin et al., 2008). Therefore it was assumed that PDI is inactivated due to S-nitrosylation in the affected neurons, which causes protein misfolding in ALS (Honjo et al., 2011). Studies of genetics, model organisms, and patient's tissue samples support PDI upregulation triggered by ER stress and post-translational inhibition of PDI due to *S*-nitrosylation (Atkin et al., 2008; Walker and Atkin, 2011).

#### *Therapeutic strategies focused on reactive oxygen species*

Design of novel antioxidant strategies to selectively target oxidative stress and redox imbalance might be an important approach (**Table 3**; **Figure 4**). However, the antioxidant treatment therapy in ALS has not been successful so far (Traynor et al., 2006; Bedlack et al., 2007; Beghi et al., 2011). SOD1 is tightly connected with the nuclear erythroid 2-related-factor 2 (Nrf2). Nrf2 is a transcriptional factor and main expression regulator of many antioxidant/detoxification genes via its interaction with the antioxidant response element (ARE). Because it helps neuronal cells to cope with toxic effect of oxidative stress, pharmacological targeting of Nrf2/ARE pathway was proposed as a tool against neurodegeneration in ALS (Petri et al., 2012; Milani et al., 2013; **Figure 4**). S(+9)-Apomorphine, a CNS penetrating activator of the Nrf2/ARE pathway was able to reduce pathological oxidative stress and improved survival in fibroblasts of ALS patients, and also slowed disease progression in SOD1G93mice (Mead et al., 2013). On the other hand, Guo et al. (2013) reported a rather modest impact of Nrf2 on the course of disease in SOD1G93A mice.

Beneficial effects have been reported in NSC-34 cells for melatonin, a hormone which acts against oxidative and nitrosative stress-induced damage (Weishaupt et al., 2006). In several

studies, treatment with melatonin prolonged survival in the SOD1G93A mice (Weishaupt et al., 2006; Zhang et al., 2013), however, Dardiotis et al. (2013) showed contrary results in the same mouse model, possibly because melatonin exacerbated neurodegeneration.

In patients with ALS, high doses of melatonin were well tolerated (Jacob et al., 2002; Weishaupt et al., 2006) and it was reported that circulating serum protein carbonyls, which are oxidative stress markers, were decreased in melatonin treated ALS patients (Weishaupt et al., 2006).

Diacetylbis(*N*(4)-methylthiosemicarbazonato) copper(II), that inhibits the action of peroxynitrite on SOD1 and ensues nitration of cellular proteins, significantly delays onset of paralysis, prolongs lifespan and prevents accumulation of TDP-43 in the spinal cord of SOD1G93A mice. Therefore, it represents a potential neuroprotective agent targeting multiple disease pathways in ALS (Soon et al., 2011). In the spinal cord of ALS patients, metallothioneins (Zn modulators and anti-oxidant reaction inducers), were seriously reduced (Hozumi et al., 2008). It was demonstrated that metallothionein-III prevents the loss of motor neurons and prolongs the life span of ALS mice (Hashimoto et al., 2011; Hozumi, 2013).

#### **CALCIUM HOMEOSTASIS AND ROS**

Ca2<sup>+</sup> plays a role in ROS production and, vice versa, the redox state can modulate Ca2<sup>+</sup> signaling. Components of ROS homeostasis are regulated by Ca2+-dependent pathways. Ca2<sup>+</sup> stimulates NO synthesis, inhibits complex IV, and leads to ROS production at the complex III (Feissner et al., 2009). Depending on the targeted protein, the type and concentration of ROS and the duration of exposure, interactions between Ca2<sup>+</sup> and ROS signaling can be stimulating or inhibiting.

#### *Ryanodine receptors are stimulated by oxidation*

Ryanodine receptors (RyR) belong to a class of intracellular Ca2<sup>+</sup> channels and are important mediators of Ca2<sup>+</sup> induced Ca2<sup>+</sup> release in excitable cells such as muscles and neurons (Fabiato, 1983; McPherson et al., 1991). RyR are opened by Ca2<sup>+</sup> itself which may induce propagated Ca2+release on the ER surface. Ca2<sup>+</sup> entry through AMPA receptors caused RyR mediated Ca2+-induced Ca2<sup>+</sup> release from the ER in embryonic motor neurons in co-culture with neonatal Schwan cell (Jahn et al., 2006). In embryonic motor neurons, Ca2+-induced Ca2<sup>+</sup> release was shown to contribute greatly to AMPA receptor stimulation induced Ca2+-induced Ca2<sup>+</sup> release through RyR and Ca2<sup>+</sup> dysregulation (Grosskreutz et al., 2004). RyR form tetramers in the

sarcoplasmic reticulum (SR) and ER membranes (Xu et al., 1998; Fill and Copello, 2002). The reversible oxidation of endogenous SH groups opens the channel and releases Ca2<sup>+</sup> from SR (Abramson and Salama, 1989; Xu et al., 1998). Since sulfhydryl oxidation of reactive thiols is involved in the gating of the Ca2<sup>+</sup> release channel, RyR represents an important target in oxidative cell damage (Liu et al., 1994; Zable et al., 1997; Marengo et al., 1998). Activity of the RyR channel complex is regulated as a response to changes in transmembrane redox potential (Feng et al., 2000). When Ca2<sup>+</sup> channel activators lower the redox potential of the RyR, the thiol groups get oxidized and the channel opens. When Ca2<sup>+</sup> channel inhibitors increase the redox potential of the RyR, the disulfides are reduced and the channel closes (Feng et al., 2000). ROS, such as O•− <sup>2</sup> and H2O2, can activate the channel by direct oxidation of redox-sensing thiols (Rowe et al., 1983; Boraso and Williams, 1994; Zima et al., 2004). The endogenous ligand of RyR, FKBP12, stabilizes RyR in the absence of activation and prevents Ca2<sup>+</sup> leakage from the ER. The concentration of FKBP12 was decreased in ALS patients indicating the importance of equilibrium between FKBP12 and RyR in neurodegeneration (Kihira et al., 2005).

#### *IP***3***R are stimulated by oxidation*

A second receptor that induces the release of Ca2<sup>+</sup> from the ER is the IP3R. Ca<sup>2</sup><sup>+</sup> overload in the ER discharges IP3R spontaneously (Missiaen et al., 1991; Rooney et al., 1991). The most important ligands that modulate IP3R channel activity are InsP3 and Ca2+. At low concentrations Ca2<sup>+</sup> activates the channel, whereas at higher concentrations, Ca2<sup>+</sup> inhibits the channel (Foskett et al., 2007). IP3R can be directly activated by oxidative agents, such as O•− <sup>2</sup> (Suzuki and Ford, 1992) and H2O2 (Wesson and Elliott, 1995). Thimerosal, a sulfhydryl-oxidizing agent, stimulates IP3R channels isolated from rat cerebellum and incorporated into artificial lipid bilayer (Thrower et al., 1996) and HeLa cells (Bootman et al., 1992). Overexpression of the IP3R2 shortened the lifespan in SOD1G93A mice, which implicates the importance of ER Ca2<sup>+</sup> release by IP3Rs and that impaired function of this receptor can be destructive in ALS (Staats et al., 2012a). IP3-gated Ca<sup>2</sup><sup>+</sup> seems to be a key regulator of TDP-43 nucleoplasmic shuttling and proteostasis. Pathologic TDP-43 aggregation disturbs Ca2+-dependent TDP-43 shuttling, indicating pharmacological manipulation of IP3R as a target in TDP-43 induced neurodegeneration *in vivo* (Kim et al., 2012).

Phospholipase C delta 1 (PLCδ1) increases InsP3 formation which releases calcium from ER through IP3R. The expression of PLCδ1 is increased in ALS mouse spinal cord and neurons. Genetic ablation of PLCδ1 prevented shrinkage of motor neurons in ALS mice, suggesting that PLCδ1 is also a candidate for new targets in ALS research (Staats et al., 2013).

#### *SERCA and plasma membrane Ca***2+***-ATPase are inhibited by oxidation*

SERCA transfers Ca2<sup>+</sup> from the cell cytosol to the lumen of the SR (MacLennan et al., 1997). SERCA is very sensitive to redox state but contrary to RyR and IP3R, oxidation inhibits SERCA activity (Kaplan et al., 2003). SERCA is reversibly regulated through NOdependent S-glutathiolation of specific cysteine residues (Adachi

et al., 2004), where irreversible sulfonylation reduces SERCA (Ying et al., 2008). Thiol oxidizing agents inhibit and glutathione stimulate SERCA (Scherer and Deamer, 1986). Amino acid peroxides selectively oxidize cysteine residues of SERCA and inactivate the pump (Dremina et al., 2007). O•− <sup>2</sup> and H2O2/ •OH have been shown to inhibit Ca2<sup>+</sup> uptake into the sarcoplasmic reticulum (Rowe et al., 1983; Kukreja et al., 1988; Xu et al., 1997). H2O2/ •OH directly interfere with the ATP binding site. Since the Ca2<sup>+</sup> uptake into ER is coupled to the ATP hydrolysis, restriction of ATPase activity decreases Ca2<sup>+</sup> uptake (Xu et al., 1997). Ca2<sup>+</sup> transport and ATPase activity of plasmalemmal Ca2<sup>+</sup> ATPase can be inhibited by ROS due to oxidation of SH groups and peroxidation of membrane phospholipids (Kaneko et al., 1989). In SOD1G93A motor neurons, ER Ca2<sup>+</sup> uptake by SERCA was shown to be increased (Lautenschlager et al., 2013). Ca2<sup>+</sup> handling is reshaped during disease progression in the SOD1G93A mouse model. Increased plasma membrane extrusion upon mitochondrial failure likely indicates a compensatory mechanism in the disease. This study puts the focus on further investigations of mitochondrial and plasma membrane Ca2<sup>+</sup> transporters such as plasmalemmal Ca2<sup>+</sup> ATPase and plasma membrane Na+/Ca2<sup>+</sup> exchanger (Fuchs et al., 2013).

#### *Therapeutic strategies targeting RyR, IP***3***R, and SERCA*

Blocking RyR using dantrolene has provided protection of motor neurons exposed to a brief excitotoxic insult *in vitro*, but did not show a protective effect in SOD1G93A mice. This indicates that Ca2<sup>+</sup> release through RyRs have a modest role in SOD1 mice (Staats et al., 2012b). Inhibiting SERCA by cyclopiazonic acid showed protective effects against kainate induced excitotoxicity in SOD1G93A cultured motor neurons (Lautenschlager et al., 2013). Although there are not many studies targeting ERMCC Ca2<sup>+</sup> channels, these could be valuable targets for further investigation.

#### **CONCLUDING REMARKS**

Riluzole is currently the only approved drug for ALS, but at best it only slows disease progression for some months. It is crucial to understand disease pathophysiology and to recognize the major upstream events that lead to motor neuron death. Disturbances of Ca2<sup>+</sup> homeostasis and ER function are well-known features of motor neuron degeneration in ALS. Dysregulation in between the ERMCC is therefore characterized accumulation of misfolded proteins, oxidative stress and motor neuron death. Therapeutic drugs aiming to stabilize the ERMCC, reduce ER stress and support the UPR may be effective in a wide range of neuron diseases. However, genetic and the majority of pharmacologic strategies to protect ER and mitochondria against exitotoxicity have been unsuccessful. Nevertheless, these negative results, added to the many failed trials in the past, raise the question of the suitability of our experimental models, which are mostly murine. Perhaps we should focus on new tools such as induced pluripotent stem cells taken from ALS patients and derived into motor neurons. They could generate more suitible models.

Newly discovered genes that cause ALS may also offer new therapeutics for ALS. These strategies are currently underway. The SOD1G93A mice model has been extensively investigated so far, but there is an urgent requirement for additional models of ALS such as TDP-43, FUS and VCP. Moreover, the drugs that failed in clinical trials could still prove to play a valuble role as part of a combination strategy with other molecules in the future, such as drugs that operate in distinct or overlapping pathways. Developing of "smart drugs", such as Arimoclomol that enhance protein folding capacity only under conditions of cellular stress, may also be good direction in drug development.

Finally a significant point comprises establishing improved pharmacokinetic profiles. The safety properties and most efficient dose of the drug in humans have to be adequately established prior to phase III trials. Taken together, the key for success is in basic and clinical researchers continuing to work together.

#### **AUTHOR CONTRIBUTIONS**

All authors contributed in the conception and design of the present review, as well as in drafting and revising the manuscript.

#### **ACKNOWLEDGMENTS**

This research was supported by BMBF (the Bundesministerium für Bildung und Forschung) grants to Julian Grosskreutz in the framework of the E-RARE program (PYRAMID) and JPND program (SOPHIA) of the European Union. We thank N. Kroegel for language editing. The authors report no conflicts of interests.

#### **REFERENCES**


A53T mutant alpha-synuclein-induced toxicity. *Hum. Mol. Genet.* 14, 3801–3811. doi: 10.1093/hmg/ddi396


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 27 February 2014; accepted: 07 May 2014; published online: 30 May 2014. Citation: Tadic V, Prell T, Lautenschlaeger J and Grosskreutz J (2014) The ER mitochondria calcium cycle and ER stress response as therapeutic targets in amyotrophic lateral sclerosis. Front. Cell. Neurosci. 8:147. doi: 10.3389/fncel.2014.00147 This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2014 Tadic, Prell, Lautenschlaeger and Grosskreutz. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Riluzole But Not Melatonin Ameliorates Acute Motor Neuron Degeneration and Moderately Inhibits SOD1-Mediated Excitotoxicity Induced Disrupted Mitochondrial Ca2<sup>+</sup> Signaling in Amyotrophic Lateral Sclerosis

#### Manoj Kumar Jaiswal\* †

Center of Physiology, University of Göttingen, Göttingen, Germany

#### Edited by:

Chao Deng, University of Wollongong, Australia

#### Reviewed by:

Raman Chandrasekar, Kansas State University, USA Gundars Goldsteins, University of Eastern Finland, Finland

#### \*Correspondence:

Manoj Kumar Jaiswal mj2750@cumc.columbia.edu

#### †Present address:

Manoj Kumar Jaiswal, Divisions of Molecular Imaging and Neuropathology, New York State Psychiatry Institute, New York, NY, USA; Department of Psychiatry, Columbia University, New York, NY, USA

> Received: 11 July 2016 Accepted: 12 December 2016 Published: 06 January 2017

#### Citation:

Jaiswal MK (2017) Riluzole But Not Melatonin Ameliorates Acute Motor Neuron Degeneration and Moderately Inhibits SOD1-Mediated Excitotoxicity Induced Disrupted Mitochondrial Ca2<sup>+</sup> Signaling in Amyotrophic Lateral Sclerosis. Front. Cell. Neurosci. 10:295. doi: 10.3389/fncel.2016.00295 Selective motoneurons (MNs) degeneration in the brain stem, hypoglossal motoneurons (HMNs), and the spinal cord resulting in patients paralysis and eventual death are prominent features of amyotrophic lateral sclerosis (ALS). Previous studies have suggested that mitochondrial respiratory impairment, low Ca2<sup>+</sup> buffering and homeostasis and excitotoxicity are the pathological phenotypes found in mice, and cell culture models of familial ALS (fALS) linked with Cu/Zn-superoxide dismutase 1 (SOD1) mutation. In our study, we aimed to understand the impact of riluzole and melatonin on excitotoxicity, neuronal protection and Ca2<sup>+</sup> signaling in individual HMNs ex vivo in symptomatic adult ALS mouse brain stem slice preparations and in WT and SOD1-G93A transfected SH-SY5Y neuroblastoma cell line using fluorescence microscopy, calcium imaging with high speed charged coupled device camera, together with immunohistochemistry, cell survival assay and histology. In our experiments, riluzole but not melatonin ameliorates MNs degeneration and moderately inhibit excitotoxicity and cell death in SH-SY5YWT or SH-SY5YG93A cell lines induced by complex IV blocker sodium azide. In brain stem slice preparations, riluzole significantly inhibit HMNs cell death induced by inhibiting the mitochondrial electron transport chain by Naazide. In the HMNs of brainstem slice prepared from adult (14–15 weeks) WT, and corresponding symptomatic SOD1G93A mice, we measured the effect of riluzole and melatonin on [Ca2+]<sup>i</sup> using fura-2 AM ratiometric calcium imaging in individual MNs. Riluzole caused a significant decrease in [Ca2+]<sup>i</sup> transients and reversibly inhibited [Ca2+]<sup>i</sup> transients in Fura-2 AM loaded HMNs exposed to Na-azide in adult symptomatic SOD1G93A mice. On the contrary, melatonin failed to show similar effects in the HMNs of WT and SOD1G93A mice. Intrinsic nicotinamide adenine dinucleotide (NADH) fluorescence, an indicator of mitochondrial metabolism and health in MNs, showed enhanced intrinsic NADH fluorescence in HMNs in presence of riluzole when respiratory chain activity was inhibited by Na-azide. Riluzole's inhibition of excitability and Ca2<sup>+</sup>

signaling may be due to its multiple effects on cellular function of mitochondria. Therefore formulating a drug therapy to stabilize mitochondria-related signaling pathways using riluzole might be a valuable approach for cell death protection in ALS. Taken together, the pharmacological profiles of the riluzole and melatonin strengthen the case that riluzole indeed can be used as a therapeutic agent in ALS whereas claims of the efficacy of melatonin alone need further investigation as it fail to show significant neuroprotection efficacy.

Keywords: ALS, SOD1G93A, riluzole, melatonin, mitochondria, excitotoxicity, cell death, Ca2<sup>+</sup> signaling

#### INTRODUCTION

One of the hallmarks of ALS is extensive loss of MNs in the brain stem, hypoglossal nucleus, facial nucleus, and the spinal cord (Spalloni et al., 2004; Von Lewinski and Keller, 2005a,b; Spiller et al., 2016). Evidence is increasing that Ca2<sup>+</sup> dysregulation and mitochondrial dysfunction is involved in the SOD1 G93A (SOD1G93A) mouse model (Kong and Xu, 1998; Jaiswal and Keller, 2009; Jaiswal, 2013, 2014; Chang and Martin, 2016; Paine and Jaiswal, 2016). The precise mechanisms leading to the selective loss of MNs in ALS patients as well as a possible determinant of selective MNs death in transgenic (Tg) mouse models of this disorder remain elusive. Many findings indicated two important characteristic features of this disorder: (1) the presence of low level of calcium binding proteins in combination with high buffering capacity in affected MNs (Alexianu et al., 1994; Kaal et al., 1998; Laslo et al., 2000; Jaiswal et al., 2009; Jaiswal, 2014) and (2) involvement of AMPA receptors (AMPAR; Rothstein, 1995; Van Den Bosch et al., 2000; Udagawa et al., 2015; Chang and Martin, 2016).

Riluzole is the only FDA-approved neuroprotective compound used in clinics for the treatment of ALS/MN disease (Bensimon et al., 1994; Jehle et al., 2000; Obinu et al., 2002; Gutierrez et al., 2016). Riluzole's neuroprotective properties appear to arise from both presynaptic and postsynaptic mechanisms for attenuating glutamatergic neurotransmission, allowing riluzole to act as an antiglutamate drug (Bryson et al., 1996; Doble, 1996; Wokke, 1996; Geevasinga et al., 2016) and it block depolarization-evoked Ca2<sup>+</sup> transients in MNs (Hubert et al., 1998). Moreover, many groups reported that riluzole works by inhibiting Na<sup>+</sup> channels (Stefani et al., 1997; Yokoo et al., 1998; Zona et al., 1998; Geevasinga et al., 2016), inhibits Cl<sup>−</sup> channels (Bausch and Roy, 1996), increases glutamate uptake and has an impact on glutamate receptors (Azbill et al., 2000) and inhibits GABA reuptake (Kretschmer et al., 1998). Additionally, riluzole was shown to attenuate neurotoxicity in tissue preparations and an animal model as well (Lang-Lazdunski et al., 2000; Mu et al., 2000). Earlier it was shown that an increase in [Ca2+]<sup>i</sup> is a critical messenger for diverse pathophysiological events or in case of ALS and other neurodegenerative disorder and generally abnormal [Ca2+]<sup>i</sup> increases are cytotoxic (Bootman et al., 1993; Berridge, 1997; Budd, 1998). Riluzole is known to block persistent sodium currents and earlier it was shown that riluzole block the Trpm4 (pore-forming subunits of Sur1-regulated NCCa-ATP channels), a key mechanism for the beneficial effect of riluzole in spinal cord injury (Simard et al., 2012). Moreover, Trpm4 is a Ca2+-activated cation channel that were found to expressed at higher levels in Th2 cells and inhibition of Trpm4 expression increased Ca2<sup>+</sup> influx (Weber et al., 2010) via Ca2<sup>+</sup> release-activated Ca2<sup>+</sup> (CRAC) channel and Ca2+-activated K<sup>+</sup> current (Kca) channels (Dolmetsch and Lewis, 1994). More recently it was discovered that Ca2<sup>+</sup> influx and oscillations are also regulated by Trpm4 in Jurkat T cells (Launay et al., 2004). Work from all these previous studies indicated three compelling reasons for the beneficial effects of riluzole (i) block of persistent Na+ currents (molecular mechanism not known), (ii) inhibition of glutamatergic signaling pathways and (iii) an indirect neurotrophic effect. In spite of all these discoveries, it is not precisely established yet the precise mechanism through which riluzole acts on ion channels either directly or through second messenger signaling cascades, which attenuate ion influx-induced neuronal death.

Another effective cellular antioxidant known to be efficiently scavenging toxic free radicals, ROS and associated reactants is Melatonin (Jou et al., 2004; Reiter et al., 2016). Recent findings in which azides/CN-induced cell death via blocking complex IV of the mitochondrial ETC is reversed by melatonin and prevention of mitochondrial damage induced by ruthenium red (inhibits the mitochondrial Ca2<sup>+</sup> uniporter for Ca2<sup>+</sup> uptake), suggest that melatonin may act at the mitochondrial subcellular level and reduce ROS induced damage (Yamamoto and Tang, 1996; Martin et al., 2000). Furthermore, administrations of melatonin repair mitochondrial functioning in aging mice and shown to be neuroprotective in in vitro models of Alzheimer's disease (Pappolla et al., 1997; Okatani et al., 2003).

Collectively, these results demonstrate the protective effects of riluzole and melatonin in neurodegenerative diseases involving mitochondrial dysfunction. However, we do not know if the therapeutic effects of these drugs in various pre-clinical mitochondria related neurodegenerative diseases also effective in ALS. Moreover, since mitochondrial ROS-induced cell death

**Abbreviations:** aCFS, artificial cerebrospinal fluid; ALS, amyotrophic lateral sclerosis; AM, acetoxy methyl ester; CCD, charge coupled device; CRAC, Ca2<sup>+</sup> release-activated Ca2+; DMSO, dimethyl sulfoxide; ER, endoplasmic reticulum; ETC, electron transport chain; F, fluorescence; fALS, familial amyotrophic lateral sclerosis; HMNs, hypoglossal motoneurons; MNs, motor neurons; NADH, nicotinamide adenine dinucleotide; ROI, regions of interest; ROS, reactive oxygen species; SD, standard deviation; SEM, standard error of mean; SOD1, superoxide dismutase 1; WT, wild-type; [Ca2+]<sup>i</sup> , cytosolic calcium; [Ca2+]mito, mitochondrial calcium; 19m, mitochondrial membrane potential.

and excitotoxicity play a prominent role in ALS disease, specific neuroprotection mechanisms of riluzole and melatonin at the mitochondrial sub cellular level involving excitotxicity and Ca2<sup>+</sup> signaling need to be further investigated. In this context, we determine whether treatment with riluzole or melatonin could attenuate sodium-azide (Na-azide) induced ROS and eventually cell death. Additionally, we investigated the impact of riluzole and melatonin on vulnerable HMNs of adult symptomatic SOD1G93A mice and corresponding wild type (WT) littermates. Imaging experiments were performed on adult brain stem slices where mitochondrial function of HMNs was disturbed by a bath application of 3 mM Na-azide, which inhibits complex IV and thereby disturbs mitochondrial metabolism. Riluzole and melatonin protection studies were similarly designed where application of sodium azide was performed because of, (1) its quick and reversible action, and (2) In both sALS and fALS a decrease in mitochondrial complex IV observed (Nowicky and Duchen, 1998; Menzies et al., 2002). The cell death, cell survival and Ca2<sup>+</sup> signals were characterized and evaluated in the presence or absence of drugs.

We found that riluzole moderately ameliorates MN degeneration, inhibits excitotoxicity and mildly and reversibly inhibits Ca2<sup>+</sup> signaling. However, melatonin fails to show significant MN protection and significant impact on Ca2<sup>+</sup> signaling both in adult WT and corresponding SOD1G93A mice as well as in cell culture model of ALS. We anticipate that this experimental system can be used to screen the drugs targets either alone or in combination of drug cocktails that can be used for large-scale compound screening.

## MATERIALS AND METHODS

## SH-SY5YWT and SH-SY5YG93A Neuroblastoma Cell Lines

SH-SY5Y human neuroblastoma cell line transfected with either WT or G93A mutant linked with fALS were routinely cultured in growth medium and were used as a valid and robust in vitro system to investigate the cellular excitotoxicity and mitochondria mediated alterations associated with SOD1-G93A mutations (Carri et al., 1997; Goos et al., 2007; Jaiswal et al., 2009). Detail descriptions about the transfection strategy, cell culture maintenance, cell culture growth medium and procedures are described earlier (Jaiswal et al., 2009).

### Measurement of Cell Viability Using Mitochondrial Toxin Sodium Azide and Neuroprotection by Riluzole and Melatonin

Toxicity of Na-azide induced mitochondrial inhibition and thereby cell death assay in SH-SY5YWT and SH-SY5YG93A was done using bright field microscopy, hematoxylin and eosin (H&E) histochemistry and WST-1 assay. Neuroprotection study was done using similar method. SH-SY5YWT and SH-SY5YG93A transfected cells on glass coverslips were treated for 3 min (acute), 30 min (moderate) and 60 min (chronic) exposure with 3 mM Na-azide (60 min data not shown), cells were fixed with 4% paraformaldehyde (PFA), dehydrated through water/ethanol and histolene and then stained with Meyer's hemalum reagent (1:1 in H2O). While choosing the Na-azide concentrations we followed the previous literature and accordingly we selected three different Na-azide concentrations, 1, 3, and 10 mM for cytotoxicity induced cell death assays and neuroprotection study (Hunter et al., 1959; Harvey et al., 1999). Concentration of 1 mM Na-azide not always leads to swelling of mitochondria for acute (3 min) exposure. Also shown by other groups at 0.1 mM or less than 2 mM sodium cyanide (NaCN) was not always adequate to prevent completely spontaneous swelling or swelling induced by low concentrations of phosphate (Hunter et al., 1959; Harvey et al., 1999). Higher concentration, e.g., 10 mM Na-azide might react with components other than cytochrome oxidase, therefore based on our pilot toxicity experiments and previous literature, we have chosen 3 mM Na-azide concentration suitable for our acute toxicity (3 min exposure) and neuroprotection assays. To study the differences in impact of Na-azide–induced cell death and neuroprotection mediated by riluzole and melatonin, SY5YWT and SH-SY5YG93A cells at a density of 10<sup>5</sup> cells/cm<sup>2</sup> were seeded into 24-well plates. SH-SY5YWT and SH-SY5YG93A cells were exposed to culture medium with 3 mM Na-azide for 3 min (acute), 30 and 60 min (chronic). After 3, 30, and 60 min (60 min data not shown) of exposure, cell photomicrograph was acquired using brightfield microscope. Furthermore, for neuroprotection assay by riluzole and melatonin, both cell lines were exposed to DMEM growth medium with 3 mM Naazide+100 µM riluzole or 3 mM Na-azide+100 µM melatonin for 3 min (acute). After rinsing the cover slips twice with water, cover slips were dehydrated and Vectashield Medium (Vector Laboratories, Burlingame, CA, USA) was used for mounting the cover slips.

To further investigate the impact of Na-azide exposure on cell viability and neuroprotection by riluzole and melatonin, cells at a density of 10<sup>5</sup> cells/cm<sup>2</sup> in 24-well plates were treated with culture medium or medium containing 3 mM Na-azide final concentration. After 3, 30, and 60 min (60 min data not shown) of incubation with 3 mM Na-azide, WST-1 cell proliferation reagent (Roche Applied Science, Mannheim, Germany) was used to determine cell viability. For neuroprotection, similar WST-1 assay were done using 3 mM Na-azide+100 µM riluzole or 3 mM Na-azide+100 µM melatonin. The WST-1 assay is comprises a cleavage of the tetrazolium salt by healthy mitochondria to soluble orange formazan, by the enzyme succinate dehydrogenase present in the ETS of healthy mitochondria. The formazan dye formed after incubation of cells with WST-1 for 2 h was quantified by measuring the absorbance at 490 nm using the Genios multiplate reader (Tecan, Crailsheim, Germany) and correlates with the total number of metabolically viable cells.

### Animals and Genotyping

Superoxide dismutase 1G93A Tg 1Gur (Fast Line) mice strain is considered a well-established animal model for human ALS. This mice strain acquired from the Jackson Laboratory (Bar Harbor, ME, USA) and in-house breeding was carried out in our animal facility. The SOD1G93A Tg mouse develop paralysis in the

limb and eventually dies at 4–5 months of age due to the loss of MNs in the brain stem, HMNs, FMNs, and the spinal cord whereas WT littermates were unaffected. The Tg mouse carries a variation of the human mutant G93A-SOD1 gene in which at position 93 glycine is substituted by alanine (Gurney et al., 1994). Breeding and genotyping procedures were adapted from Hoyaux et al. (2000; for further details also see Jaiswal and Keller, 2009). Animal breeding and experimental procedures were approved and carried out in accordance with the guidelines of the ethics committee of the Medical Faculty of Georg-August University, Göttingen, Germany.

### Preparation of Acute Brain Stem Slices from Adult SOD1G93A/WT Mice Littermates

Mice were anesthetized until the paw withdrawal reflex disappeared in a chamber containing an ether vapor-enriched atmosphere and immediately sacrificed. Mice brains were quickly removed and then put in 4◦C ice-cold aCSF. Transverse 250 µm thick acute brain stem slices were prepared from the 14–15 weeks old WT and symptomatic SOD1G93A mice in late stage of motor dysfunction according to the procedures described earlier (Ladewig and Keller, 2000; Jaiswal and Keller, 2009) using a vibroslicer (Leika VT 1000S, Göttingen, Germany). Selectively vulnerable brain stem HMNs (Nuc. hypoglossus; XII) region were visually identified by their location nearby to the IV ventricle during the brain stem slice preparations transversely (**Figure 3**). To achieve maximum oxygen supply to slices, aCSF (in mM: NaCl 118, KCl 3, NaH2PO<sup>4</sup> 1, CaCl<sup>2</sup> 1.5, MgCl<sup>2</sup> 1, NaHCO<sup>3</sup> 25, glucose 20; pH 7.4; 320 mOsm) was constantly simmered with a mixture of 95% O<sup>2</sup> and 5% CO<sup>2</sup> (carbogen). An essential requirement for the microfluorometric measurements was the preservation of MNs in adult slices near to the slice surface in a physiologically viable and intact condition. This challenging requirement for Ca2<sup>+</sup> imaging in MNs was accomplished by reducing mechanical tissue disturbances during the thick adult brain slice preparations, slice cutting at 4◦C temperature and regular maximum oxygen delivery to keep metabolic conditions optimum for cells viability (For details, see Jaiswal and Keller, 2009). Slices were kept at ∼27◦C in continuous carbogenbubbled aCSF solution at pH 7·4 for 30 min and then let it cool down to room temperature (RT, ∼20–21◦C) prior to dye loading. Unless indicated otherwise, all experiments were performed at RT.

## Charge Coupled Device (CCD) Fast Optical Imaging

Intracellular calcium [Ca2+]i fluorescence measurement were achieved using an optical recording system including an upright Axioscope microscope (Zeiss, Göttingen, Germany) equipped with a computer-operated monochromator (TILL Photonics, Martinsried, Germany) built on a galvanometric scanner (Polychrome II, TILL Photonics, Martinsried, Germany). Briefly, a modified version of the CCD camera system (TILL Photonics, Planegg, Germany) and Achroplan W 63×, 0.9W objective were used to collect fluorescent signal changes in defined "regions of interest" (ROIs) in cell soma. Fluorescence signals changes were digitized using a 12-bit CCD camera (PCO, Kelheim, Germany), binning of which was put to 4 × 4, with a sampling rate between 3 and 13 Hz using TILL Vision Software (TILL Vision Software V3.3, TILL Photonics, Germany). Dichroic mirror with mid reflection at 425 nm was used for Fura-2 AM calcium dye (Invitrogen, Carlsbad, CA, USA) to direct the emitted light. To achieve the ratiometric calcium imaging, swapping between wavelengths (360 and 390 nm) was achieved in ∼3 ms, thereby allowing fast ratiometric Ca2<sup>+</sup> measurements.

### Microfluorometric Ca2<sup>+</sup> Measurements in SH-SY5Y Transfected Cells and HMNs from WT and SOD1G93A Brain Stem Slice Preparations

WT (SH-SY5YWT) and SOD1 (SH-SY5YG93A) transfected cell layers on cover slips were stained with membrane-permeable Fura-2 AM (K<sup>d</sup> ∼ 0.2 µM) dye by incubating with RPMI-1640 medium having 10% FCS (Invitrogen, Carlsbad, CA, USA) and 0.846 mM Ca2<sup>+</sup> (5 µM Fura-2 AM dissolved in DMSO containing 10% pluronic acid for 30 min at 37◦C and continuous bubbled with 95% O<sup>2</sup> and 5% CO2). After the dye incubation cells were washed with RPMI and incubated in culture medium for ∼20 min at 37◦C to permit de-esterification of dye. Baseline [Ca2+]i were measured in transfected cells 2–5 days in culture. Chemical induction of Ca2<sup>+</sup> measurements were done by incubating the cells for 3 min with 3 mM Na-azide and effects of 100 µM riluzole and 100 µM melatonin were measured by inhibition of Ca2<sup>+</sup> with similar measurements (data not shown).

In HMN<sup>S</sup> of brain stem slice, calcium signals were measured by defining appropriate ROIs of MN somas as previously described (Lips and Keller, 1998). [Ca2+]<sup>i</sup> signals were monitored by bath-loading of the brain stem slices with 5 µM final dye concentration of Fura-2 AM for 40 min at 27◦C with continuous carbogen bubble (5% CO<sup>2</sup> and 95% O2). After the dye incubation, prior to onset of imaging slices were washed with aCSF and incubated another 30 min in aCSF for de-esterification of Fura-2 AM. Excitation of Fura-2 was done alternately at 360 and 390 nm by UV light and emitted light was directed to a dichroic mirror having mid-reflection at 425 nm and filtered by a band pass filter of 505–530 nm (Zeiss, Goettingen, Germany) using a computercontrolled monochromator. Changes in [Ca2+]<sup>i</sup> are denoted as variations in Fura-2 ratio (F/F0) where F = fluorescence at different time points of experiment and F<sup>0</sup> = baseline fluorescence value before drug/chemical applications (for details see Jaiswal and Keller, 2009). Further calculations and analysis of [Ca2+]<sup>i</sup> signals were performed offline using IGOR Pro (Wavemetrics, Lake Oswego, OR, USA) and OriginPro 6.0 (OriginLab Corporation, Northampton, MA, USA) Software.

### Intrinsic nicotinamide adenine dinucleotide (NADH) Fluorescence

The metabolic state of HMNs from SOD1G93A mice in the presence and absence of 100 µm riluzole was monitored by measuring and monitoring intrinsic nicotinamide adenine dinucleotide (NADH) fluorescence of HMNs in the slice

preparations using a high-speed CCD camera. Briefly, NAD(P)H excitation at 360 nm was carried via a computer-controlled monochromator ((TILL Photonics, Martinsried, Germany) reflected onto the surface of the slice using dichroic mirror (400 nm, Zeiss, Goettingen, Germany). The hypoglossal nucleus was easily recognized by a high level of intrinsic fluorescence. Fluorescence emission was collected by using a CCD camera (TILL Photonics, Planegg, Germany). All experiments used a 410 nm LP filter between the dichroic mirror and camera to maximize light capture. Imaging was performed after focusing onto the surface of slices in hypoglossal area using 63× (0.9 NA) Achroplan water objective (Zeiss, Göttingen, Germany) and collected by a 12-bit CCD camera (PCO, Kelheim, Germany) described earlier. Calculations and analysis of NAD(P)H signals were performed offline after the experiment using IGOR Pro (Wavemetrics, Lake Oswego, OR, USA) and OriginPro 6.0 (OriginLab Corporation, Northampton, MA, USA) Software.

### Chemical Reagents

Fura-2 AM dye (50mg/vial) was purchased from Invitrogen (Carlsbad, CA, USA) and dissolved in 50 µL DMSO containing 20% Pluronic F-127 to a concentration of 1 mM. Na-azide, riluzole, melatonin and all other chemicals were purchased from Sigma–Aldrich (St. Louis, MO, USA). To prepare the stock solutions, riluzole and melatonin were dissolved in 100% ethanol to make 10 mM stock solutions and all the time kept at 4◦C to avoid ethanol evaporation. Na-azide was dissolved in distilled water. To keep cells viable drug solutions when mixed in the perfusate always simmered with carbogen (95% O2, 5% CO2) during the experiments.

### Statistical Analysis

SH-SY5YWT and SH-SY5YG93A transfected cells attached with cover slip was used for single experiments and replicate in five separate experiments (five field of view) for each condition in cell viability assay (cell survival/death) and seven separate experiments (seven field of view for each cover slides for SH-SY5YWT and SH-SY5YG93A) for neuroprotection assay. All slices were used for a sole experiment and 5–6 HMNs taken from each slice in Fura-2 calcium imaging data experiments. 2–3 slices were used per mouse and 5–6 HMNs selected from each slice for bar graph presentation. In slice experiments, bar diagram represents 12 imaging experiments recorded from 12 separate slices obtained from 5 WT mice and 10 imaging experiments recorded from 10 slices obtained from 5 SOD1G93A mice in acute Na-azide experiments. In case of riluzole experiments, bar diagram represent 14 imaging experiments recorded from 14 separate slices obtained from 6 WT mice and 13 imaging experiments recorded from 13 slices obtained from 6 SOD1G93A mice. In case of NADH fluorescence imaging, 5–6 cells from slice for five separate experiments for each condition from five different mice of same genotypes was used for bar diagram representation. Values in the text and error bars in experimental figures are represented as mean ± SD. Significance for all the in vitro assay, slice imaging and pharmacological intervention was calculated using the two-tailed unpaired student t-test and a p-value < 0.05 was considered statistically significant. Oneway analysis of variance (ANOVA) with Bonferroni's method was used as the post hoc test used to recognize statistically significant Na-azide effects, and neuroprotection effects of riluzole and melatonin exposed cells as compared with controls with in groups.

## RESULTS

### Cell Death Assessment by Inhibition of Mitochondria and Neuroprotective Actions of Riluzole and Melatonin against Sodium Azide-Induced Cell Death in In vitro SH-SY5Y Cell Culture Model of ALS

We utilized differentiated SH-SY5YWT and SH-SY5YG93A cells used previously to generate WT and SOD1G93A cell culture model of ALS (Jaiswal et al., 2009). Cell death, cell viability and antiaggregate formation were assessed using bright field microscopy, immunocytochemistry and WST-1 cell proliferation assay (**Figures 1** and **2**). SH-SY5YG93A cells showed slightly higher levels of cell death compared to SH-SY5YWT after 3 and 30 min (**Figures 1A–Da–c** and **2A**) exposure to mitochondrial toxin Naazide (**Figures 1A–Da–c** and **2A**). A similar pattern was observed while investigating the apoptosis and aggregation assessed by bright field microscopy demonstrating typical apoptotic morphology, e.g., condensation/fragmentation of nuclear material and elongation of cell size and cell membranes. Cells displaying fragmentation of nuclear materials (pyknosis)/cell membrane elongation were classified as apoptotic and showed aggregates formation with both cell lines after incubation for 30 and 60 min (**Figures 1Aa–c,Ca–c**; data not shown for 60 min exposure) but not with acute 3 min (acute) induction (**Figures 1Aa–c,Ca–c**). We have not quantified the cell numbers after 30 and 60 min exposure to Na-azide due to lump formation after 30 min exposure and inability to separate each cells with in aggregates and almost complete cell death 60 min after the inhibition of mitochondria by Na-azide (**Figures 1Ac,Cc**). Cell survival/death was assessed by H&E staining and WST-1 test. Cell count for H&E staining for both SH-SY5YWT and SH-SY5YG93A cells treated with Na-azide for 3 min (77.0 ± 4.7 for WT and 71.5 ± 6.1 for SOD1G93A) and 30 min (59.2 ± 5.8 for WT and 54.0 ± 5.5 for SOD1G93A) were significantly lower compared to untreated cells with Na-azide (114.5 ± 2.8 for WT and 107.0 ± 5.1 for SOD1G93A; **Figures 1A–Da–c** and **2A**; N = 5; ∗∗∗P < 0.001, ∗∗P < 0.005, Students two sample t-test, **Table 1**).

It was previously reported by several groups that riluzole and melatonin ameliorates ALS progression in SOD1G93A mouse model and extend morbidity of hALS patients. We next attempted to determine whether riluzole and melatonin confers neuroprotection and ameliorate cell death in ALS cell culture model and if yes at what extent. Previously it was shown that riluzole up to 10 µM have no effect on cell survival in DA neurons and SH-SY5Y cells. However, concentrations greater than 50 µM have a remarkable effect on cell survival (Storch et al.,

showed 3 min acute exposure of 3 mM Na-azide induced cell death is partially ameliorated in the presence of 3 mM riluzole (d), but not in the presence of 3 mM melatonin (e). (C) Bright field images of differentiated SH-SY5YG93A cells exposed to DMEM (a), DMEM+3 mM Na-azide for 3 min (b), DMEM+3 mM Na-azide for 30 min (c), DMEM+3 mM Na-azide+3 mM riluzole for 3 min (d), and DMEM+3 mM Na-azide+3 mM melatonin for 3 min (e). Note the decrease in cell density in presence of 3 mM (acute) and 30 min (chronic) Na-azide exposure. Note the increase in cell density of SH-SY5YWT cells compared to SH-SY5YG93A cells in presence of 3 mM riluzole suggesting greater susceptibility of SH-SY5YG93A to mitochondrial-induced damage. (D) H&E staining showed a similar substantially higher density of living SH-SY5YG93A cells like SH-SY5YWT cells exposed to DMEM (a) compared to cells exposed to 3 min (acute, b) and 30 min (chronic, c) exposure with 3 mM Na-azide. Cell death of H&E stained SH-SY5YG93A cells in presence of 3 mM Na-azide exposure is partially ameliorated in presence of 3 mM riluzole (d), but not in presence of 3 mM melatonin exposure (e). Scale bar for (A) and (C) = 20 µm and (B) and (D) = 100 µm.

2000); therefore, here we used 100 µM riluzole and melatonin exposure with superfusate and evaluate the protective effects of riluzole on cell survival in SH-SY5YWT and SH-SY5YG93A cells in vitro (**Figures 1A–Dd,e** and **2A**). Our findings show that the acute inhibition of mitochondrial metabolism and function and thereby cell death by Na-azide is ameliorated by riluzole but there is no significant effects of melatonin in both SH-SY5YWT and SH-SY5YG93A cells (**Figures 1A–Dd,e** and **2A**). The average number of H&E positive cells in a given field of view (seven field of view) after the treatment of 100 µM riluzole to SH-SY5YWT and SH-SY5YG93A cells is significantly higher (**Figure 2A**; 110.0 ± 7.8 for WT and 101.2 ± 6.7 for SOD1G93A) compared to non-treated cells stimulated with 3 mM Na-azide (**Figure 2A**; 77.0 ± 4.7 for WT and 71.5 ± 6.1 for SOD1G93A; for riluzole treated and non-treated cells; ∗∗P < 0.005, Students two sample t-test; **Table 1**). In addition to decreasing the cell death, riluzole had significant protective effects on the neuritis, fine cell processes and cell membrane integrity being preserve from mitochondrial inhibition and the effect was most pronounced in SH-SY5YWT compared to SH-SY5YG93A cells (**Figures 1Ad–Dd**). However, impact of 100 µM melatonin treatment on Na-azide induced inhibition of mitochondria is not significantly different then nontreated cells in both SH-SY5YWT and SH-SY5YG93A cells though there are minor increase in cell count of WT and SOD1G93A cells (**Figures 1A–De** and **2A**; 87.2 ± 7.0 for WT and 82.5 ± 5.3 for SOD1G93A, **Table 1**).

FIGURE 2 | Riluzole but not melatonin stimulation increased cells survival and neuroprotective against sodium azide induced mitochondrial ETC blockade both in SH-SY5YWT and SH-SY5YG93A cells. (A) Live cell count of SH-SY5YWT (gray bar) and SH-SY5YG93A (red bar) neuroblastoma cells after incubation with DMEM, DMEM+3 mM Na-azide for 3 min, DMEM+3 mM Na-azide for 30 min, DMEM+3 mM Na-azide+3 mM riluzole for 3 min and DMEM+3 mM Na-azide+3 mM melatonin for 3 min (average of cell counts for seven field of view in each conditions). Please note the significant difference between riluzole-treated and non-treated SH-SY5YWT and SH-SY5YG93A cells exposed to Na-azide (p < 0.005). (B) Toxicity of Na-azide for SH-SY5YWT and SH-SY5YG93A cells was measured using the WST-1 cell viability test in presence of DMEM, DMEM+3 mM Na-azide for 3 min, DMEM+3 mM Na-azide for 30 min, DMEM+3 mM Na-azide+3 mM riluzole for 3 min and DMEM+3 mM Na-azide+3 mM melatonin for 3 min. SH-SY5YWT and SH-SY5YG93A cells were more vulnerable to the action of Na-azide in absence of riluzole compared to 3 mM riluzole presence in DMEM growth medium (p < 0.01). Data are expressed as the mean ± SEM. ∗∗∗P < 0.005, ∗∗P < 0.01, <sup>∗</sup>P < 0.05 two-tailed unpaired student t-test.

TABLE 1 | Cell count of H&E staining for SH-SY5YWT and SH-SY5YG93A cells treated with sodium azide in presence and absence of riluzole and melatonin.


Cell viability assessed by WST-1 assay showed significant dose dependent ROS-induced mitochondrial toxicity and cell death. Cell viability of SH-SY5YWT and SH-SY5YG93A cells in presence of Na-azide for 3 and 30 min leads to a decrease in cell survival (75.0 ± 3.4 and 61.2 ± 6.1% for 3 and 30 min incubation, respectively, for WT; **Figure 2B**) and were significantly not different with SOD1G93A cells (73.5 ± 6.2 and 54.0 ± 8.4% for 3 and 30 min incubation, respectively; **Figure 2B**; **Table 2**). However, cell viability for both WT and SOD1G93A treated cells were much lower and significantly different compared to untreated cells without Na-azide toxicity (100 ± 3.4% for WT and 91 ± 8.6% for SOD1G93A; **Figures 1A–Da–c** and **2A**; N = 5; ∗∗∗P < 0.001, Students two sample t-test; **Table 2**). Similarly % of SH-SY5YWT and SH-SY5YG93A viable cells after the treatment of 100 µM riluzole is significantly higher (**Figure 2B**; 91.0 ± 4.5 for WT and 88.2 ± 5.2 for SOD1G93A; **Table 2**) compared to non-treated cells stimulated with 3mM Na-azide for 3 min

(**Figure 2A**; 75 ± 6.3% for WT and 73.5 ± 6.3% for SOD1G93A; ∗∗P < 0.01, Students two sample t-test; **Table 2**). There is no significant increase in cell survival after melatonin treatment (**Figure 2B**; 80.2 ± 5.3 for WT and 78.5 ± 4.2 for SOD1G93A; **Table 2**).

### Impact of Riluzole on Sodium Azide-Induced Mitochondrial [Ca2+]<sup>i</sup> Release and Excitotoxicity

To test further the efficacy of riluzole on Na+-azide induced mitochondrial Ca2<sup>+</sup> signaling, Fura-2 AM stained HMNs of brain stem slices were exposed to Na-azide in presence and absence of 100 µM riluzole. Sketch diagram of hypoglossal MNs in brainstem slice preparations (**Figure 3A**) and representative photomicrograph of fura-2 AM stained HMNs in WT and (left) and symptomatic SOD1G93A mice (right) shown in



**Figure 3B**. We evaluated the protective impact of riluzole on calcium signaling of HMNs of adult WT (**Figures 3C,E**) and symptomatic SOD1G93A mice brain stem slices in vitro (**Figures 3D,F**). We found that riluzole has a moderate effect on inhibition of Na-azide (3 mM) induced [Ca2+]<sup>i</sup> signaling of MN's mediated by Na+-azide induced excitotoxicity. At higher concentration (>50 µM) riluzole moderately inhibited [Ca2+]<sup>i</sup> fluorescence in both WT and selectively impaired HMNs of symptomatic SOD1G93A mice whereas at lower concentrations (<50 µM) there is no significant effect (data not shown). Sodium azide-induced mitochondrial [Ca2+]<sup>i</sup> release response is significantly different between WT and symptomatic SOD1G93A mice. The impact of riluzole inhibition is not specific to only SOD1G93A mice and it also moderately inhibits [Ca2+]<sup>i</sup> signaling in WT mice as well (**Figure 3**; **Table 3**).

As illustrated in **Figures 3C,D,G**; 3 mM Na-azide caused a rise in the [Ca2+]<sup>i</sup> transient by mitochondrial complex IV inhibition, with average peak amplitudes (F/F0) reaching 0.15 ± 0.02 and 0.11 ± 0.01 (**Figures 3C,D,G**; **Table 3**) respectively, in HMNs of adult WT (5 mice/N = 12 slice; n = 58 HMNs) and symptomatic SOD1G93A mice (5 mice/N = 10 slice; n = 46 HMNs) and subsequently comes to baseline over 2–3 min wash out with aCSF. In the presence of 100 µM riluzole, application of 3 mM Na-azide caused a rise in the [Ca2+]<sup>i</sup> transient, with average of the peak amplitudes (F/F0) reaching 0.089 ± 0.01 and 0.065 ± 0.01 (**Figures 3E–G**; **Table 3**) respectively, in HMNs of adult WT (6 mice/N = 14 slice; n = 63 HMNs) and symptomatic SOD1G93A mice (6 mice/N = 13 slice; n = 58 HMNs). The inhibition of the peak amplitude of [Ca2+]<sup>i</sup> in SOD1G93A mice (F/F<sup>0</sup> = 0.065 ± 0.01) is prominent compared to WT mice (F/F<sup>0</sup> = 0.089 ± 0.01; P < 0.05, Students two sample t-test; quantitative fluorescence value compared in **Table 3**). The inhibition of [Ca2+]<sup>i</sup> peak amplitude in SOD1G93A mice by riluzole might be attributed to a decrease in the entry of Ca2<sup>+</sup> through a voltage dependent calcium channel (VDCC) and need further investigations. We conclude that riluzole inhibits Na-azide-induced mitochondrial Ca2<sup>+</sup> signaling and it modestly ameliorate MN degeneration by inhibition of excitotoxicity trigger by [Ca2+]<sup>i</sup> efflux.

### Impact of Riluzole on Metabolic State of HMNs in Presence of Sodium Azide-Induced Mitochondrial Respiratory Chain Inhibition and Excitotoxicity

Mitochondrial ETC associated proton efflux causes accumulation of negative charges in the matrix and electrochemical gradient across the mitochondrial membrane. Intrinsic NADH fluorescence in MNs has been shown to serve as a valuable tool to characterize the metabolic signature of intrinsic energy profiles. Accordingly, evaluation of intrinsic NADH fluorescence after inhibition of mitochondrial complex IV with Na-azide indicated that the mitochondrial respiratory chain was significantly disturbed in SOD1 G93A animals which is rescues in acute condition by riluzole by yet unknown mechanism (**Figure 4**). To confirm that impact of riluzole on mitochondrial NADH fluorescence, experiments were performed, in which respiratory chain activity of HMNs was inhibited by 3 mM Na-azide in absence (aCSF; **Figure 4A**) and presence of riluzole (100 µM riluzole; **Figure 4B**). Kinetic profile of HMNs (n = 5) NADH fluorescence in single slice in aCSF (black) and in presence of riluzole (red) shown in **Figure 4C** for three consecutive application.

To analyze the comparative efficiency of mitochondria of SOD1G93A mice, HMNs were evoked by a 30 mM K<sup>+</sup> depolarizing stimulus. When SOD1G93A mice HMNs were exposed to 30 mM K<sup>+</sup> for 20 s, there is immediate and slow plateau shape decrease in the NADH fluorescence and subsequently returned to the basal level. To determine whether the mitochondrial depolarization by 30 mM K<sup>+</sup> has any apparent influence on the lifetime of the NADH fluorescence and riluzole still rescues HMNs metabolism, 100 µM riluzole was added to the aCSF superfusate after K+-induced depolarization (**Figures 4D,E**). Application of 3 mM Na-azide after 20 s depolarization induced by K<sup>+</sup> resulted in an increase of NADH fluorescence of 0.024 ± 0.003 and 0.049 ± 0.001 in the presence of aCSF (N = 5 slice; n = 38 HMNs) and aCSF+100 µM riluzole (N = 5 slice; n = 41 HMNs), respectively, in SOD1G93A mice (**Figure 4F**). Following previous trends there was a slight increase in the NADH fluorescence in presence of 100 µM riluzole but there was no significant difference in NADH fluorescence mean value after 3 mM Na-azide-evoked responses (normalized, F/F0) post K+-induced depolarization with or without 100 µM riluzole.

### DISCUSSION

We have shown earlier that [Ca2+]<sup>i</sup> in HMNs increases in response to stimulation with mitochondrial electron transport complex IV inhibitors Na-azide/cyanide and differentially regulated in WT and SOD1G93A mice and cell culture model of SOD1 (Bergmann and Keller, 2004; Jaiswal and Keller, 2009; Jaiswal et al., 2009). Given that riluzole and melatonin has been shown to avert excitotoxicity induced cell death by various mechanism, including oxidative stress and calcium

Goöttingen, Univ., Diss., 2008. Copyright© 2009 Jaiswal, M. K. ediss.uni-goettingen.de. Used with permission]. (B) A CCD camera images (4 × 4 binning) showing 15 weeks adult WT (left) and symptomatic SOD1 (right) mice HMNs stained with ratiometric calcium dye fura-2 AM (excitation at 360 nm). (C) 3 mM Na-azide was added in aCSF superfusate and the calcium fluorescence signal was recorded in 14–15 weeks old WT mice HMNs. (D) 3 mM Na-azide was added in aCSF superfusate and the calcium fluorescence signal was recorded in 14–15 weeks old symptomatic SOD1G93A mice HMNs. (E) 3 mM Na-azide was added in aCSF superfusate together with 100 µM riluzole and the calcium fluorescence signal was recorded in 14–15 weeks old WT mice HMNs. (F) 3 mM Na-azide was added in aCSF superfusate together with 100 µM riluzole and the calcium fluorescence signal was recorded in 14–15 weeks old symptomatic SOD1G93A mice HMNs. Average peak amplitude (gray color) shown as means from 5 to 6 cells in an imaging field (C–F). (G) Bar diagram of effect of 100 µM riluzole on Na-azide induced blockade of [Ca2+]<sup>i</sup> measured in adult WT (Light gray) and symptomatic SOD1G93A (Gray sparse) mice HMNs. Data are expressed as means 12 imaging experiments from five WT mice and corresponding 10 imaging experiments from five symptomatic SOD1G93A mice in acute 3 mM Na-azide experiments. In case of riluzole experiments, bar diagram represent 14 imaging experiments obtained from six WT mice and corresponding 13 imaging experiments obtained from six symptomatic SOD1G93A mice. The changes in peak amplitude of [Ca2+]<sup>I</sup> were calculated as F/F<sup>0</sup> (360/390). <sup>∗</sup>P < 0.05, ∗∗P < 0.01 compared with adult WT and symptomatic SOD1G93A mice after riluzole application. Data are expressed as the mean ± SD.

#### TABLE 3 | Sodium azide-induced mitochondrial [Ca2+]<sup>i</sup> release and excitotoxicity in absence and presence of riluzole.


signaling (Wang, 2009; Mazzone and Nistri, 2011; Jaiswal, 2012; Geevasinga et al., 2016; Gutierrez et al., 2016; Reiter et al., 2016), inhibition of persistent sodium currents and Trpm4 (Simard et al., 2012), Trpm4 mediated increase of Ca2<sup>+</sup> influx (Weber et al., 2010) via Ca2<sup>+</sup> CRAC and Kca channels (Dolmetsch and Lewis, 1994), we examined whether riluzole and melatonin affects mtSOD1-mediated cell death and excitotoxicity and improve cell survival. Furthermore, since Na+-azide blocks the mitochondrial ETC in HMNs of ALS, we further analyzed whether riluzole and melatonin protect the HMNs from Na-azide induced MN death induced by excitotoxicity and mediated by inhibition of [Ca2+]mito signaling cascades. The effect of drug riluzole and melatonin on the excitotoxicity-induced cell death and [Ca2+]<sup>i</sup> concentration induced by mitochondrial blocker was measured, and the impact of riluzole and melatonin-modulated Ca2<sup>+</sup> influx and Ca2<sup>+</sup> release were evaluated. We found that riluzole but not melatonin, moderately inhibits excitotoxicity-induced [Ca2+]mito signaling and thereby modestly protects the cell death in WT and SOD1G93A cell culture model of ALS and SOD1G93A mice model (**Figures 1–3**). Previously, it was shown that melatonin enhance mitochondrial ETC I and IV and thereby improving mitochondrial respiration, ATP synthesis and energy metabolism under stress circumstances (Leon et al., 2005). However, in our experiments melatonin fail to show neuroprotective action on either cell death (**Figures 1** and **2**) or Na+-azide induced [Ca2+]<sup>i</sup> signaling (data not shown) in adult WT and corresponding symptomatic SOD1G93A mice. Our results indicate that the melatonin mitochondrial neuroprotective action might be governed by respiratory complex I inhibition or by yet unknown unidentified mechanism(s). We therefore believe that the riluzole-targeted inhibition of mitochondrial signaling could be beneficial in ALS. In this paper, we provide evidence using three independent methods about inhibition of depolarizationevoked calcium transients and excitotoxcity in healthy and diseased MNs and cell death/survival assay in cell culture model of MNs by neuropharmacological substrates riluzole

to protects cells against excitotoxic damage and hence its therapeutic effects in ALS. These observations may be relevant to understand the vulnerability of MNs to the excitotoxic insult that may contribute to the aetiopathology of ALS.

We found a substantial increase in cell viability and inhibition of excitotoxicity induced cell death in MNs by riluzole but not by melatonin (**Figures 1** and **2**), thus providing evidence for the efficacy of riluzole in in vitro transfected cell culture model in the treatment of ALS. Riluzole has been shown to alleviate neurological symptoms in the SOD1G93A Tg mouse model of ALS by suppressing oxidative stress (Seo et al., 2015). Indeed, we observed that riluzole reduced Na-azide induced and SOD1G93A-mediated cell death (**Figure 1**) and increase cell viability (**Figure 2**) in SH-SY5YWT and SH-SY5YG93A cells, suggesting that mitochondrial excitotoxicity and ROS is associated with the mechanism of SOD1G93A-induced cytotoxicity. However, compared to riluzole, melatonin fail to prevent Na-azide induced and SOD1G93A-mediated cell death (**Figure 1**) and there is no significant increase in cell viability (**Figure 2**) in SH-SY5YWT and SH-SY5YG93A cells, suggesting that mechanism of cell death protection by melatonin might adopt different cellular mechanism as it did not affect cell viability induced by Na-azide induced ROS generation through mitochondrial inhibition.

Furthermore, we demonstrated that the inhibition of peak amplitudes of [Ca2+]<sup>i</sup> in symptomatic SOD1G93A and adult WT mice is significant when measured with application of Na+-azide together with riluzole compared to of Na+-azide alone (**Figure 3**, p < 0.05). The overall inhibition of the rise in peak [Ca2+]<sup>i</sup> transients in the presence of riluzole after a Na+-azide application in adult WT and symptomatic SOD1G93A might be due to the potential inhibition of complex IV and decrease in the entry of Ca2<sup>+</sup> through VDCC. Additionally, in symptomatic SOD1G93A mice, the presence of riluzole slightly diminished the Na+-azide induced [Ca2+]mito increase and delayed the baseline recovery by up to 30–60 s. Our results therefore indicate that riluzole's moderate inhibition of [Ca2+]mito signaling may be because of a partial blockade of intracellular Ca2<sup>+</sup> release from the mitochondria. Our finding further supports the assumption that the riluzole ameliorates excitability and Ca2<sup>+</sup> signaling and this pathway may have role in its manifold effects on mitochondria mediated cellular metabolism in ALS.

NADH fluorescence is a good experimental measure to monitor the metabolic profile of the MNs. In the intact HMNs besetting the mitochondria by Na-azide adversely affects

the production of NADH in the mitochondrial metabolism. Our data shows slight increase in the NADH fluorescence in presence of 100 µM riluzole compared to aCSF but there was no significant difference in NADH fluorescence mean value (**Figure 4**). This implies that there is a significant mitochondrial metabolism dysfunction and mitochondria in SOD1G93A mice might not able to sequester enough calcium after K+-induced depolarization (dysfunctional Ca2<sup>+</sup> accumulation activity) or Na-azide-evoked release response is hampered in HMNs of SOD1G93A mice.

Overall, our results indicate that the inhibition and stabilization of Ca2<sup>+</sup> transient by riluzole under physiological situation, e.g., diffusion-restricted and in homeostatically controlled mitochondrial domains might indeed have numerous functional advantages (**Figure 5**). Accordingly, drugs targeting protection of mitochondrial function and homeostasis could be useful in various forms of ALS by utilizing multidrug therapy where combined pharmacological intervention target different domains of mitochondria and Mito-cycle which ameliorates excitotoxicity both at the MNs and at the neighboring cells and most likely prolong survival of ALS patients. Our experiments promise a limited but effective and methodological way to test the multiple drugs and their targets most likely tested with different drug cocktails to check the efficacy of multi-drug therapy. However, more detailed study allowing identification of mechanism and molecular targets would be necessary.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

MKJ established the experimental protocol, designed the experiments, carried out all of the experiments, analyzed the data, and wrote the manuscript. Bernhard U. Keller provides laboratory support and financial support for research work.

### FUNDING

This study was partially supported by the Bundesministerium für Bildung und Forschung (BioRegioN/ERANET) [Grant 0313610A] and the Göttingen Center for Molecular Physiology of the Brain [Grant 1356834].

### ACKNOWLEDGMENTS

I would like to thank Prof. Dr. Bernhard U. Keller for comments on the manuscript, valuable discussions and mentorship and financial help. I would like to thank Dr. Diane M. Jaworski from University of Vermont and Drs. Namboodri and Arun from USUHS/Walter reed army institute for SH-SY5Y cell line, Cornelia Hühne for excellent technical assistance and Dr. Ananta Paine for comments on the manuscript.

cytosolic Ca2+ concentration in transfected neuroblastoma SH-SY5Y cells. FEBS Lett. 414, 365–368.



vulnerability of motor neurons. J. Neurol. Sci. 180, 29–34. doi: 10.1016/S0022- 510X(00)00414-7


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Jaiswal. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Mechanisms of mutant SOD1 induced mitochondrial toxicity in amyotrophic lateral sclerosis

#### **Piia Vehviläinen, Jari Koistinaho and Gundars Goldsteins \***

Department of Neurobiology, A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland

#### **Edited by:**

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### **Reviewed by:**

Tibor Kristian, University of Maryland School of Medicine, USA Pavle R. Andjus, University of Belgrade, Serbia Ping Liu, University of Connecticut Health Center, USA

#### **\*Correspondence:**

Gundars Goldsteins, Department of Neurobiology, A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, P.O. Box 1627, Neulaniementie 2, 70211 Kuopio, Finland

email: gundars.goldsteins@uef.fi

In amyotrophic lateral sclerosis (ALS), mitochondrial dysfunction is recognized as one of the key elements contributing to the pathology. Mitochondria are the major source of intracellular reactive oxygen species (ROS). Increased production of ROS as well as oxidative damage of proteins and lipids have been demonstrated in many models of ALS. Moreover, these changes were also observed in tissues of ALS patients indicative of important role for oxidative stress in the disease pathology. However, the origin of oxidative stress in ALS has remained unclear. ALS linked mutant Cu/Zn-superoxide dismutase 1 (SOD1) has been shown to significantly associate with mitochondria, especially in the spinal cord. In animal models, increased recruitment of mutant SOD1 (mutSOD1) to mitochondria appears already before the disease onset, suggestive of causative role for the manifestation of pathology. Recently, substantial in vitro and in vivo evidence has accumulated demonstrating that localization of mutSOD1 to the mitochondrial intermembrane space (IMS) inevitably leads to impairment of mitochondrial functions. However, the exact mechanisms of the selectivity and toxicity have remained obscure. Here we discuss the current knowledge on the role of mutSOD1 in mitochondrial dysfunction in ALS from the novel perspective emphasizing the misregulation of dismutase activity in IMS as a major mechanism for the toxicity.

**Keywords: superoxide dismutase activity, oxidative stress, mitochondria, intermembrane space, misfolding**

#### **INTRODUCTION**

Amyotrophic lateral sclerosis (ALS) is a fatal paralyzing disease characterized by the selective degeneration of motor neurons in the spinal cord, brain stem and motor cortex. The majority (∼90%) of ALS cases are sporadic whereas some 10% of patients have dominantly inherited familial ALS (FALS). However, the forms share clinical features. The suggested causative factors for the motor neuron degeneration include oxidative stress, neuroimmune reactions, protein aggregation (Boillée et al., 2006) and mitochondrial abnormalities such as disturbed calcium homeostasis (Jaiswal and Keller, 2009; De Vos et al., 2012) and dysfunctional axonal transport (De Vos et al., 2007). Nevertheless, the mechanism is yet to be identified. Mutations in several genes, including C9orf72, superoxide dismutase 1 (SOD1), TAR DNA binding protein (TARDBP), Fused in sarcoma (FUS), angiotensin, alsin, senataxin, and vesicle-associated membrane protein (VAPB) genes, cause FALS (Finsterer and Burgunder, 2014; Kiernan, 2014). Mutations in the C9orf72 gene are estimated to account for 35% and SOD1 mutations for 20% of FALS cases. The first gene linked to FALS was SOD1 (Rosen et al., 1993) and by today more than 150 FALS associated mutations in SOD1 have been identified. There are two types of mutant SOD1 (mutSOD1) proteins inducing clinically similar disease: those that show wild type like enzymatic activity and those with markedly reduced activity (Valentine et al., 2005). In addition, lack of SOD1 does not lead to development of ALS in mice (Reaume et al., 1996). Thus, the deleterious effect of mutSOD1 is thought to involve acquired toxic property possibly due to misfolding, aberrant enzymatic activity (Bruijn et al., 2004) or disturbance of cellular homeostasis by incorporating into the lipid bilayer membrane (Allen et al., 2012).

#### **THE ROLE OF SUPEROXIDE DISMUTATION IN MITOCHONDRIAL HYDROGEN PEROXIDE GENERATION AND OXIDATIVE STRESS**

Reactive oxygen species such (ROS) as superoxide and hydrogen peroxide (H2O2) are products of normal oxygen metabolism in cells (Dröge, 2002). They serve as signaling molecules but when present in excess can harm the structure and functions of the cell. In the cytoplasm, SOD1 dismutates superoxide anion radical to H2O<sup>2</sup> that is further reduced to H2O by catalase, glutathione peroxidases or peroxiredoxins (Rhee et al., 2005). In intermembrane space (IMS), SOD1 has been suggested to play the similar protective role in handling of superoxide as in the cytosol (O'Brien et al., 2004; Aquilano et al., 2006; Klöppel et al., 2010; Fischer et al., 2011; **Figure 1**, reaction V). However, in this location, the scavenging systems might not be efficient enough to eliminate the H2O<sup>2</sup> produced by dismutation. Upon cellular stress and pathological conditions such as ALS, the elevated H2O<sup>2</sup> levels could contribute to mitochondrial damage.

Alternatively, superoxide in IMS could be detoxified by cytochrome C (cytC) which oxidizes it to O<sup>2</sup> without producing secondary ROS (**Figure 1**, reaction Vb), thus providing a safe pathway to eliminate superoxide. However, cytC can also catalyze peroxidation in the presence of H2O<sup>2</sup> yielding oxoferryl-cytC (CytCFe4+) which is a highly reactive oxidant (**Figure 1**, reaction VI). Indeed, we have demonstrated earlier a novel mechanism by which increased dismutase activity in IMS leads to cytC dependent peroxidation, followed by mitochondrial dysfunction and apoptosis (Goldsteins et al., 2008). Cells lacking SOD1 produced less H2O<sup>2</sup> and were less sensitive to apoptosis induced by inhibition of mitochondrial respiratory complex III. Moreover, mitochondria isolated from spinal cords of G93A mice showed increased dismutase activity and elevated ROS production compared to wild type (wt) mitochondria. Further *in vivo* studies revealed that mutSOD1 in spinal cords of G93A SOD1 rodents is destabilized and bound to the inner mitochondrial membrane (Ahtoniemi et al., 2008). These results raise two essential issues for understanding the mechanisms of mutSOD1 induced mitochondrial toxicity: translocation of SOD1 to IMS and activity control of IMS localized SOD1.

### **FACTORS REGULATING THE IMPORT OF SOD1 INTO MITOCHONDRIA**

A substantial proportion of SOD1 is found in the mitochondria (Weisiger and Fridovich, 1973; Jaarsma et al., 2001; Mattiazzi et al., 2002), concentrated in the IMS (Okado-Matsumoto and Fridovich, 2001; Sturtz et al., 2001). In mammalian cells, the mitochondrial localization of SOD1 is regulated by its folding as it can cross the mitochondrial membrane only in unfolded disulfide-reduced and non-metallated apo-form (Kawamata and Manfredi, 2008; **Figure 1**, reaction I). In addition, it was demonstrated that the subcellular localization of SOD1 is dictated by distribution of the copper chaperone for SOD1 (CCS; Kawamata and Manfredi, 2008). The maturation of SOD1 requires zinc binding followed by copper insertion and oxygen-dependent intramolecular disulfide bond formation (Furukawa et al., 2004), the last two actions being executed by CCS (Culotta et al., 1997). For proteins such as SOD1 and CCS that lack the mitochondrial localization signal, there is a disulfide relay system in IMS which drives their import (Mesecke et al., 2005). This Mia40/Erv1 system is involved in the translocation of yeast and mammalian SOD1 and CCS (Kawamata and Manfredi, 2008; Reddehase et al., 2009; Groß et al., 2011; Klöppel et al., 2011). When the reduced apo-proteins enter the IMS, the Mia40/Erv1 machinery introduces intramolecular disulfide bonds into them (Herrmann and Riemer, 2010). This folding step traps the proteins in the IMS as only unfolded precursors can use the translocator on the mitochondrial outer membrane (TOM) for exit.

ApoCCS is a direct substrate for Mia40 (**Figure 1**, reaction IIa) and the critical cysteines for their interaction in domain I of CCS were identified by Klöppel et al. (2011). The IMS-located CCS promotes folding of apoSOD1 and its subsequent trapping in IMS (Kawamata and Manfredi, 2008; **Figure 1**, reaction IIIa). High levels of CCS have been observed in the IMS of yeast and mammalian cells (Field et al., 2003; Kawamata and Manfredi, 2008) and CCS was found essential for mitochondrial localization of SOD1 in yeast (Field et al., 2003). Accordingly, the amount of mitochondrial SOD1 increases upon overexpression of CCS in mice (Son et al., 2007). In contrast, under the normal cell culture conditions overexpression of CCS resulted in decreased translocation of both CCS and SOD1 into the mitochondria (Kawamata and Manfredi, 2008). This appeared to happen because of the high oxygen concentration which promoted the oxidative folding of CCS and subsequent maturation of SOD1 in the cytosol. Lowering the O<sup>2</sup> concentration from 20% to 6% that more resembles the conditions in tissues directed CCS and SOD1 to IMS (Kawamata and Manfredi, 2008). These data indicate tight O2-dependent physiological regulation of CCS and SOD1 localization.

Redox status of the cysteine residues in human SOD1 is critical for its retention in mitochondria due to their involvement in intramolecular disulfide bonds and in the interaction with CCS. There are four Cys residues in SOD1 (C6, C57, C111 and C146) which all have a role in mitochondrial import and accumulation (Kawamata and Manfredi, 2008). An intramolecular disulfide bridge is formed between C57 and C146 upon maturation (Arnesano et al., 2004) thus making these residues essential for the retention of SOD1 in IMS. C6 and C111 are not involved in intramolecular bonds but rather in intermolecular interactions (Niwa et al., 2007; Cozzolino et al., 2008). These interactions appear to be important as Cys-111 was identified as a key mediator of mitochondrial association *in vitro* (Ferri et al., 2006) and mutation in either of the two residues leads to poor accumulation of SOD1 in IMS (Kawamata and Manfredi, 2008).

A mitochondrial import mechanism for SOD1 independent of CCS has been described in yeast (Varabyova et al., 2013). This utilizes the mitochondrial inner membrane organization system (MINOS) and involves also Mia40. Interestingly, reduced SOD1 was retained in the mitochondria of CCS null cells in yeast (Varabyova et al., 2013). In mammalian cells, CCS independent mitochondrial import and retention exists for mutSOD1 (Kawamata and Manfredi, 2008).

#### **CONTROL OF SOD1 ACTIVITY BY ITS REDOX STATE, METAL BINDING AND QUATERNARY STRUCTURE**

For maturing into an active enzyme, SOD1 requires three posttranslational modifications: the insertions of zinc and copper, intramolecular disulfide bond formation, and dimerization. Zinc is deposited close to the active site stabilizing the structure (Potter et al., 2007). Copper is required for the enzymatic activity and is delivered to the catalytic center by assistance of CCS (Culotta et al., 1997). CCS also participates in the oxidation of the intramolecular disulfide bond between C57 and C146 in SOD1 (Furukawa et al., 2004) which results from disulfide transfer involving C244 and C246 in CCS (Banci et al., 2012). Finally, two metallated and disulfide-oxidized SOD1 monomers form a dimeric active enzyme via hydrophobic and hydrophilic interactions. The mature SOD1 is extremely stable due to the intramolecular disulfide bond which brings stability to the quaternary structure (Arnesano et al., 2004; Hörnberg et al., 2007).

Full activation of human SOD1 requires CCS (Wong et al., 2000) but an activation pathway independent of CCS also exists (Carroll et al., 2004). This alternative way of activation requires reduced glutathione but is independent of oxygen (Carroll et al., 2004; Leitch et al., 2009). Thus, in contrast to oxygen-dependent CCS-mediated activation this pathway allows SOD1 activation in low oxygen conditions. Notably, in tissues of mice null for CCS as much as 10–20% of SOD1 activity has been observed compared to wt mice (Wong et al., 2000; Subramaniam et al., 2002).

### **ABERRANT MITOCHONDRIAL IMPORT AND ACTIVITY CONTROL OF SOD1 IN ALS**

#### **ACCUMULATION OF MUTSOD1 IN MITOCHONDRIA**

In ALS, mutSOD1 is recruited to mitochondria especially in the spinal cord (Higgins et al., 2002; Liu et al., 2004). Oxygen concentration critically influences cellular distribution of CCS and subsequently that of wtSOD1 whereas mutSOD1 appears to be able to escape this physiological regulation leading to mitochondrial accumulation of mutants also in the absence of CCS (Kawamata and Manfredi, 2008). Recently, it was demonstrated that reduced mutSOD1 enter the mitochondria via CCS-independent route involving MINOS and Mia40 (Varabyova et al., 2013; **Figure 1**, reaction IIIb). In accordance with the implications of the redox status of Cys residues in human SOD1 for its mitochondrial association, a variety of SOD1 mutants have been observed to possess susceptibility to the reduction of disulfide bonds *in vitro* (Tiwari and Hayward, 2003). These findings are in agreement with studies implicating the presence of reduced SOD1 in spinal cords of ALS mice (Jonsson et al., 2006; Karch et al., 2009; Zetterström et al., 2013). Even though the *in vivo* studies did not assess the localization of reduced mutSOD1, one could expect that it would be accessible to mitochondria where a part of that pool could be oxidized. In cell culture, oxidation of Cys111 residues led to accumulation of several SOD1 mutants in mitochondria (Ferri et al., 2010). However, another study demonstrated that, in contrast to wtSOD1, mitochondrial association of mutSOD1 does not depend on Cys111 (Kawamata and Manfredi, 2008).

Furukawa and O'Halloran (2005) suggested that the reduction of conserved disulfide bond predisposes SOD1 to incorrect crosslinking and aggregation (**Figure 1**, reaction IVb). Certainly, formation of intracellular aggregates has caught much attention as it is a hallmark of ALS. Indeed, several mouse models show correlation between insoluble aggregated forms of mutSOD1 and manifestation of symptoms (Johnston et al., 2000; Wang et al., 2002, 2003) in concert with findings from human tissues (Kato, 2008). Moreover, mutants possessing a higher aggregation propensity correlated with shorter disease duration in patients when 33 different SOD1 mutants were studied (Prudencio et al., 2009). In contrast to that correlation, D101N, I113T and L144F mutations with low or modest propensity to aggregate provoked rapidly progressing disease (Ayers et al., 2014). Despite of intensive investigations the mechanism how aggregates would result in toxicity has remained unclear. It has been suggested that soluble forms of mutSOD1 initiate the disease (Zetterström et al., 2007; Karch et al., 2009) and are, in fact, more toxic than insoluble aggregates (Zetterström et al., 2007; Brotherton et al., 2012; Weichert et al., 2014). Thus, the key appears to involve misfolding and the localization in the mitochondria rather than aggregate formation. Soluble misfolded subfractions of mutSOD1 were found enriched in spinal cords of ALS mice (Zetterström et al., 2007), and the conformation of such species could facilitate mitochondrial translocation. Investigators utilizing antibodies selectively recognizing misfolded/non-native SOD1 have demonstrated association of misfolded SOD1 with spinal cord mitochondria in mutSOD1 rodents (Velde et al., 2008; Brotherton et al., 2012). In cultured motor neuronal cells, obligate expression of mutSOD1 in IMS leads to mitochondrial toxicity and cell death (Cozzolino et al., 2009; Magrané et al., 2009). The toxicity was abolished by overexpression of glutaredoxin 2 possibly via modulation of disulfide bonds or by affecting mitochondrial redox environment (Ferri et al., 2010).

Most forms of mutSOD1 show instability *in vitro* (Rodriguez et al., 2002, 2005; Lindberg et al., 2005) and mitochondrial mut-SOD1 has been reported to exist in partially unfolded oligomerized state (Deng et al., 2006; Ferri et al., 2006). Interestingly, Synofzik et al. (2012) reported on patients carrying a novel SOD1 mutation, L117V, which was indistinguishable from wtSOD1 in terms of stability and dismutase activity resulting in slowly progressing form of ALS. These findings are consistent with earlier data demonstrating that even small amounts of misfolded mutSOD1 are enough to induce ALS (Jonsson et al., 2006). In addition, overexpressed wtSOD1 is also able to acquire an unstable, misfolded conformation, leading to motor neuron disease in mice (Graffmo et al., 2013).

#### **THE ROLE OF DISMUTASE ACTIVITY IN MUTSOD1 INDUCED MITOCHONDRIAL TOXICITY**

The IMS-located SOD1 has been suggested to be inactive in the intact mitochondria and activated through oxidative modification of its critical thiol groups by yet poorly known mechanisms (Iñarrea et al., 2005). Strikingly, overexpression of CCS in G93A-SOD1 mice results in enhanced mitochondrial recruitment of SOD1 and drastically accelerates disease progression (Son et al., 2007). Further, the worsened outcome was independent of protein aggregation, leaving the possibility of the involvement of aberrant dismutase activity. Indeed, increased dismutase activity was found in mitochondria isolated from spinal cords of G93A mice compared to wt mitochondria and it was accompanied with elevated ROS production (Goldsteins et al., 2008). Accordingly, mitochondrial state 4 respiration was increased in spinal cord and brain mitochondria of presymptomatic G93A rats and it was coupled with increased production of hydrogen peroxide (Panov et al., 2011). This is in agreement with observed hypermetabolism in ALS patients (Bouteloup et al., 2009). In addition, recent data on pulmonary fibrosis shows that translocation of SOD1 to IMS leads to increased production of H2O<sup>2</sup> (He et al., 2011).

Insertion of copper is required for the activity of SOD1 and altered copper homeostasis is a common feature in mutSOD1 mice (Tokuda et al., 2013). Cu(atsm) treatment prolongs the survival of ALS mice (Soon et al., 2011). Importantly, in the spinal cords of those mice the cytoplasmic dismutase activity was increased upon copper supplementation. Even though the mitochondrial activity was not assessed one could hypothesize that it could be decreased: if more SOD1 is activated in the cytoplasm then the pool of apo-form capable of entering mitochondria should decrease. When mutSOD1 mice were treated with copper chelator the onset of disease was delayed and the mice survived longer (Hottinger et al., 1997; Tokuda et al., 2008). This improvement was accompanied with decreased dismutase activity in spinal cord homogenates (Tokuda et al., 2008). In cell culture, copper depletion markedly increased the viability of G93A motoneurons (Azzouz et al., 2000). In another study, copper depletion has been shown to increase mitochondrial association of SOD1 (Arciello et al., 2011). In the latter, the cell viability was not assessed and the mitochondrial morphology was not altered. It is possible that in the above mentioned studies on copper depletion, the mitochondrial import of SOD1 was increased but due to the lack of copper the enzyme remained inactive and thus unable to cause mitochondrial damage. However, the lack of CCS, which provides SOD1 with copper, does not rescue mice from mutSOD1 induced ALS (Subramaniam et al., 2002). Moreover, when all four histidine residues that coordinate copper binding are knocked out from mutSOD1 mice they still develop motor neuron disease (Wang et al., 2003). These latter two findings could be explained by alternative ways of SOD1 activation and by heterodimer formation (Witan et al., 2008, 2009).

In addition to proper folding and metal acquisition, dimerization is critical for dismutase activity. Mutant and wtSOD1 are able to form either homo- or heterodimers both in animal and cell culture models (Furukawa et al., 2006; Witan et al., 2008; Wang et al., 2009). Interestingly, wtSOD1 monomer is able to provide enzymatic activity to the heterodimer where the other monomer is an inactive mutant (Witan et al., 2008). Furthermore, it was found that the toxicity of mutSOD1 does not correlate with its aggregation potential but with the ability to form active dimeric molecules (Witan et al., 2008). There is some discrepancy in current data published on the role of wtSOD1 in FALS. However, most of the rodent models of ALS where wtSOD1 is co-expressed with mutSOD1 show earlier onset of the disease (Jaarsma et al., 2000; Fukada et al., 2001; Deng et al., 2006; Wang et al., 2009). Notably, in the work by Wang et al. (2009) the phenotype was associated with markedly increased dismutase activity in spinal cord extracts. Both Furukawa and Wang also confirmed the existence of heterodimers *in vivo*, suggesting that formation of heterodimers may be physiologically relevant in FALS patients as well. It has been proposed that the heterodimers comprise a pool of more soluble and more toxic SOD1 whereas homodimers form less toxic aggregates (Weichert et al., 2014). Furthermore, it was demonstrated that wtSOD1 actually reduces aggregate formation of mutSOD1 in cell culture (Witan et al., 2009). How different dimers affect mitochondrial dismutase activity and H2O<sup>2</sup> production is not known and warrants further elucidation. In the model proposed (**Figure 1**, reaction IV) enzymatically active homoand heterodimers all contribute to dismutation of superoxide. Accumulation of these dimers leads to increased activity which is likely augmented by loss of physiological controlling mechanism. Subsequent increase in H2O<sup>2</sup> leads to cytC catalyzed peroxidation and mitochondrial damage.

#### **CONCLUSIONS**

Under physiological conditions SOD activity in IMS is suppressed by redox state control. However, upon mitochondrial stress SOD1 may undergo oxidative activation and compete with cytochrome C for superoxide released in the IMS, leading to increased ROS production. In ALS, mutSOD1 accumulates in IMS, thus contributing to the misregulation of dismutase activity. The presence of endogenous wt enzyme may provide dismutase activity to inactive mutant by heterodimerization. Here we propose that aberrant regulation of dismutase activity in IMS as a major mechanism for the toxicity should be further evaluated in future studies. Furthermore, the possible involvement of mitochondrial IMS dismutase activity misregulation in sporadic ALS warrants careful assessment.

#### **REFERENCES**


pathologically affected tissues in familial ALS. *Proc. Natl. Acad. Sci. U S A* 109, 5505–5510. doi: 10.1073/pnas.1115009109


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; accepted: 22 April 2014; published online: 09 May 2014*. *Citation: Vehviläinen P, Koistinaho J and Goldsteins G (2014) Mechanisms of mutant SOD1 induced mitochondrial toxicity in amyotrophic lateral sclerosis. Front. Cell. Neurosci. 8:126. doi: 10.3389/fncel.2014.00126*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2014 Vehviläinen, Koistinaho and Goldsteins. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

# Calcium dysregulation links ALS defective proteins and motor neuron selective vulnerability

Sónia S. Leal 1,2 and Cláudio M. Gomes 1,2 \*

<sup>1</sup> Faculdade de Ciências, Biosystems and Integrative Sciences Institute and Department of Chemistry and Biochemistry, Universidade de Lisboa, Campo Grande, Lisboa, Portugal, <sup>2</sup> Instituto Tecnologia Química e Biológica, Universidade Nova de Lisboa, Oeiras, Portugal

More than 20 distinct gene loci have so far been implicated in amyotrophic lateral sclerosis (ALS), a fatal neurodegenerative disorder characterized by progressive neurodegeneration of motor neurons (MN) and death. Most of this distinct set of ALSrelated proteins undergoes toxic deposition specifically in MN for reasons which remain unclear. Here we overview a recent body of evidence indicative that mutations in ALSrelated proteins can disrupt fundamental Ca<sup>2</sup><sup>+</sup> signalling pathways in MN, and that Ca<sup>2</sup><sup>+</sup> itself impacts both directly or indirectly in many ALS critical proteins and cellular processes that result in MN neurodegeneration. We argue that the inherent vulnerability of MN to dysregulation of intracellular Ca<sup>2</sup><sup>+</sup> is deeply associated with discriminating pathogenicity and aberrant crosstalk of most of the critical proteins involved in ALS. Overall, Ca<sup>2</sup><sup>+</sup> deregulation in MN is at the cornerstone of different ALS processes and is likely one of the factors contributing to the selective susceptibility of these cells to this particular neurodegenerative disease.

#### Edited by:

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Adán Dagnino-Acosta, Baylor College of Medicine, USA Marco Martina, Northwestern University, USA Karim Mekhail, University of Toronto, Canada

#### \*Correspondence:

Cláudio M. Gomes, Faculdade de Ciências, Biosystems and Integrative Sciences Institute and Department of Chemistry and Biochemistry, Universidade de Lisboa, Ed C8, Campo Grande, Lisboa 1749-016, Portugal cmgomes@fc.ul.pt; folding.fc.ul.pt

> Received: 21 April 2015 Accepted: 28 May 2015 Published: 16 June 2015

#### Citation:

Leal SS and Gomes CM (2015) Calcium dysregulation links ALS defective proteins and motor neuron selective vulnerability. Front. Cell. Neurosci. 9:225. doi: 10.3389/fncel.2015.00225 Keywords: neurodegenerative diseases, calcium homeostasis, ALS, proteinopathies, SOD1

### Sporadic and Familial ALS Aggregates Share Identical Proteins

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the selective degeneration of motor neurons (MN) in the spinal cord, brainstem and cerebral cortex (Rowland and Shneider, 2001). Most cases of ALS are sporadic (sALS) with no known genetic linkage, while approximately 10% are associated with familial forms (fALS), presenting mutations in over 20 genes encoding for distinct proteins with varied functions (**Table 1**). Despite the heterogeneous genetics of fALS and unknown etiology of sALS and its clear multifactorial character, most ALS patients present similar phenotypes with formation of cytoplasmic proteinaceous aggregates in the affected MN (Al-Chalabi et al., 2012). Notably, many proteins involved in fALS forms are also found in sALS toxic aggregates (Maekawa et al., 2009; Deng et al., 2010; Forsberg et al., 2010; Blokhuis et al., 2013) and found to cross-talk and impact on each other in ALS pathology (Kanekura et al., 2004; Volkening et al., 2009; Tudor et al., 2010; Nihei et al., 2012; Pokrishevsky et al., 2012; Stoica et al., 2014; Osaka et al., 2015). This suggests that apart from mutations, additional chemical and/or biological factors influence the selective involvement of these proteins also in sALS neurodegeneration. In agreement, most of the proteins which are involved in ALS are ordinarily expressed in many distinct cell tissues other than the nervous system (e.g., SOD1, FUS, TDP-43, VAPB, matrin-3, ataxin-2, alsin) but are only found to generate toxicity among MN. The selective vulnerability of such cells suggests that environmental triggers within those neurons are mandatory for the onset of ALS.

#### TABLE 1 | Heterogeneity of fALS causative genes.


Proteins are indicated in parenthesis for some cases: SOD1, Superoxide Dismutase 1; TDP-43, TAR DNA-binding protein 43; FUS, Fused in Sarcoma protein.

### Calcium Dysregulation in ALS-Affected Motor Neurons—The Factual Case of ALS1

A particular feature that distinguishes ALS affected MN from other cells relates to their inherent vulnerability to Ca2<sup>+</sup> overload; indeed, these neurons highly express Ca2<sup>+</sup> permeable AMPA receptors (Williams et al., 1997; Shaw and Eggett, 2000; Van Den Bosch et al., 2000; Vandenberghe et al., 2000; Guatteo et al., 2007) concurrently with a low Ca2<sup>+</sup> buffering capacity due to endogenous low expression of Ca2<sup>+</sup> buffering proteins (CaBPs) such as parvalbumin and calbindin (Alexianu et al., 1994; Palecek et al., 1999; Jaiswal, 2013) albeit the presence of EF-hand Ca2<sup>+</sup> binding proteins in MN (Migheli et al., 1999; Zhang et al., 2014). This combination of inherent physiological features of MN to manage Ca2<sup>+</sup> levels, though essential for normal functioning (von Lewinski and Keller, 2005), are likely a predisposition risk for the systematic intracellular Ca2<sup>+</sup> overload that is detected in ALS1 affected MN (Siklós et al., 1996, 1998; Kruman et al., 1999; Grosskreutz et al., 2010; Kawamata and Manfredi, 2010). Interestingly, the levels of calretinin and parvalbumin in MN axons is found further decreased in ALS patients (Hayashi et al., 2013), thus establishing an increased deficit in MN Ca2<sup>+</sup> buffering capability under pathological conditions.

In fact, a direct outcome of the low expression of CaBPs in MN is that mitochondria are likely to assume a major role in buffering calcium in these cells. In agreement, it might not be a coincidence that in ALS1, mutated SOD1 was shown to abnormally accumulate in the mitochondrial intermembrane space (Jaarsma et al., 2001; Liu et al., 2004), affecting mitochondrial function leading to disturbance of Ca2<sup>+</sup> homeostasis and ERMCC cycle (Jaiswal and Keller, 2009; Lautenschläger et al., 2013). Moreover, studies on cellular and animal ALS1 models have shown that Ca2<sup>+</sup> overload is linked with SOD1 aggregation (Tateno et al., 2004; Tradewell et al., 2011). Indeed, we have recently shown that Ca2<sup>+</sup> can bind to SOD1 immature states promoting its aggregation (Leal et al., 2013; Estácio et al., 2015) thus establishing an additional ALS1 pathological pathway for the impact of Ca2<sup>+</sup> overload on SOD1 toxic deposition. In agreement, a decrease of intracellular Ca2<sup>+</sup> overload in ALS1 models through AMPA channel antagonists or overexpression of calcium-buffering proteins has been shown to reduce SOD1 toxic aggregation and neurodegeneration (Beers et al., 2001; Van Damme et al., 2003; Tateno et al., 2004; Tortarolo et al., 2006; Yin et al., 2007; Parone et al., 2013).

### Crossways of ALS Critical Proteins with Calcium

In spite of the scarcity of available data regarding more recently discovered fALS models linked with systematic Ca2<sup>+</sup> deregulation, compelling evidence argues that many critical proteins involved in ALS (other than SOD1), are directly or indirectly involved with Ca2+. It is noteworthy that in such instances, ALS-associated mutations in many of these ALS critical proteins happen to potentiate Ca2<sup>+</sup> deregulation and/or result in an increased vulnerability to the effects of Ca2+. For example, VAP-B which is involved in ALS8, is a ERmembrane MAM protein directly engaged in Ca2<sup>+</sup> exchange between the ER and mitochondria (De Vos et al., 2012). The ALS VAPBP56S variant was shown to disrupt Ca2<sup>+</sup> homeostasis leading to a perturbation of the anterograde mitochondrial axonal transport and affecting the Miro1/kinesin-1 interaction with tubulin (Mórotz et al., 2012). Alsin, a protein implicated in juvenile ALS2, is involved in endossome/membrane trafficking that undergoes Ca2<sup>+</sup> dependent binding to the NCS regulating neurocalcin alpha protein (Masutani et al., 2008). This suggests that alsin membrane binding might be a Ca2<sup>+</sup> dependent process, and therefore passible to become affected by dysregulation of Ca2<sup>+</sup> levels. Moreover, alsin is also found to play a role in AMPAR trafficking, where ALS2 mutations lead to distinct subcellular GRIP1 localization and reduction of the calciumimpermeable GluR2 containing AMPA receptors, thus likely rendering neurons susceptible to deviant Ca2<sup>+</sup> influxes (Lai et al., 2006). Matrin 3, a multifunctional nuclear matrix protein involved in ALS21, is suggested to be regulated through a Ca2<sup>+</sup> dependent interaction with CaM (Valencia et al., 2007) and is therefore likely to be affected by deregulated Ca2<sup>+</sup> levels. As only very recently matrin 3 has been implicated in ALS (Johnson et al., 2014), future studies will be needed to clarify this possibility. Ataxin-2, which is involved in ALS13 is an ubiquitous cytoplasmic protein proposed to induce defects in the ER–Golgi pathway and disrupt Ca2<sup>+</sup> signalling (van den Heuvel et al., 2014). The possibility that ataxin-2 influences ER–Golgi Leal and Gomes Calcium, ALS and motor neuron vulnerability

function is inferred from a recent study where it was shown that intermediate-length polyQ expansions in ataxin-2 mutants enhance FUS-induced ER stress and Golgi fragmentation (Farg et al., 2013). The suggestion that the ALS13 related intermediatelength polyQ expansions in ataxin-2 can lead to Ca2<sup>+</sup> signalling disruption derives from the causal association that polyQexpanded forms of ataxin-2, or huntingtin and ataxin-3 in other diseases, associates with the C-terminal domain of the intracellular calcium release channel receptor—InsP3R1 and enhance InsP3R1-mediated calcium release in neurons (Tang et al., 2003; Chen et al., 2008; Liu et al., 2009). Thus, it is tempting to speculate that intermediate-length polyQ expansions in ataxin-2 may also cause it to associate with InsP3R1 and thereby modulate calcium signaling, though this was not yet shown. In ALS6, the FUS protein leads to CAMK2N2 upregulation (Convertini et al., 2013), an inhibitor of CAMKII that regulates neuronal synaptic plasticity through phosphorylation of AMPA receptors. Given that it has been shown that abnormal CAMKII inhibition by small molecules and peptides results in dysregulation of Ca2+/glutamate signalling (Ashpole et al., 2012), it may be hypothesized that CAMKII inhibitor—CAMK2N2 upregulation by mutant FUS will also result in Ca2<sup>+</sup> dysregulation in ALS6. Moreover, deregulated Ca2<sup>+</sup> levels can activate the Ca2+-dependent calpain protease that cleaves TDP-43 at the C-terminal, generating aggregation prone N-terminal segments that are found misallocated in the majority of ALS patients (even in those that do not carry TDP-43 mutations associated with ALS10) driving TDP-43 toxicity across ALS pathology (Aggad et al., 2014; Yamashita and Kwak, 2014). In addition, TDP-43 is simultaneously found to interact with other critical proteins in ALS namely matrin-3 (Johnson et al., 2014), ataxin-2 (Nihei et al., 2012), VAPB (Stoica et al., 2014), and SOD1 (Volkening et al., 2009) and is therefore rather tempting to conjecture about a tight interrelation between Ca2<sup>+</sup> dyshomeostasis and the involvement of critical proteins in ALS.

### ALS Toxic Processes and the Role of Calcium

In fact, major pathological processes in ALS involving excitotoxicity and the ER-mitochondria Ca2<sup>+</sup> cycle are deeply connected and potentially trigger or/and are enhanced by intracellular Ca2<sup>+</sup> deregulation: (a) Glutamatergic excitotoxicity is tough to be mediated by an excessive influx of extracellular ions, including Ca2+, resulting in elevated intracellular Ca2<sup>+</sup> levels that can activate cytoplasmatic Ca2+-dependent apoptotic proteins (e.g., calcineurin, calpain) which promote cell death (Wang et al., 1999; Kim et al., 2002); (b) elevated intracellular levels of Ca2<sup>+</sup> also lead to mitochondrial Ca2<sup>+</sup> overload, that is

#### References

Aggad, D., Vérièpe, J., Tauffenberger, A., and Parker, J. A. (2014). TDP-43 toxicity proceeds via calcium dysregulation and necrosis in aging Caenorhabditis elegans motor neurons. J. Neurosci. 34, 12093–12103. doi: 10.1523/jneurosci. 2495-13.2014

deeply interconnected with mitochondrial dysfunction resulting in ROS production, oxidative stress and eventually to apoptosis or necrosis (Kawamata and Manfredi, 2010; Cozzolino and Carrì, 2012); and (c) depletion of Ca2<sup>+</sup> levels in the ER which is suggested to occur via a persistent shift of Ca2<sup>+</sup> from the ER to the mitochondria due to deregulated ER MCC leads to protein folding dysfunction and proteasome impairment, resulting in ER stress and apoptosis (Prell et al., 2013; Tadic et al., 2014). Mutations in critical proteins associated with ALS actually seem to increase the susceptibility for these toxic processes to occur. For example, misfolded and aggregated SOD1 mutants localized within the mitochondrial membrane of spinal cord MN cause dysfunction in oxidative phosphorylation and bind aberrantly to Bcl-2, generating toxicity (Jung et al., 2002; Mattiazzi et al., 2002; Liu et al., 2004; Vande Velde et al., 2008; Pedrini et al., 2010); also, the ALS linked P56S mutation in VAPB, or th A4V, G85R and G93A SOD1 mutations leads to toxic protein aggregation and ER stress (Prosser et al., 2008; Kim et al., 2010; Atkin et al., 2014).

#### Conclusions

Overall, we here argue that Ca2<sup>+</sup> deregulation seems to establish a converging point for major ALS dysfunctional pathways and critical associated proteins, and can therefore be a key environmental factor to better understand ALS etiology and its pathomechanisms. However, we do not intend to ground that all proteins implicated in ALS will necessarily lead to Ca2<sup>+</sup> deregulation; rather, we seek to discuss that processes involving Ca2<sup>+</sup> could directly or indirectly (e.g., via Ca2<sup>+</sup> effects on processes dependent of other divalent cations) account for their mutual involvement in ALS. Interestingly, the so-called ''calcium hypothesis'' establishing a close link between Ca2<sup>+</sup> deregulation and neurodegeneration, has also been suggested to play a central role in other neurodegenerative disorders such as Alzheimer's, Ataxia, Parkinson's and Huntington Diseases (Bezprozvanny, 2010; Kasumu and Bezprozvanny, 2012), where Ca2<sup>+</sup> channels and proteins involved in neuronal Ca2<sup>+</sup> signalling systems are likely potential targets for therapeutic strategies (Zundorf and Reiser, 2011).

### Acknowledgments

This work was supported by Fundação para a Ciência e a Tecnologia through grant PTDC/QUI-BIQ/117789/2010 (to CMG), post-doctoral fellowship SFRH/BPD/47477/2008 (to SSL), strategic grant PEst-OE/EQB/LA0004/2011 (to ITQB-Laboratório Associado) and grant UID/MULTI/04046/2013 from FCT/MCTES/PIDDAC (to BioISI).


vulnerability in amyotrophic lateral sclerosis. Ann. Neurol. 36, 846–858. doi: 10. 1002/ana.410360608


amyotrophic lateral sclerosis. Neurobiol. Dis. 28, 90–100. doi: 10.1016/j.nbd. 2007.07.002


vulnerability of motor neurons. J. Neurol. Sci. 180, 29–34. doi: 10.1016/s0022- 510x(00)00414-7


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Leal and Gomes. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Compartment-dependent mitochondrial alterations in experimental ALS, the effects of mitophagy and mitochondriogenesis

Gianfranco Natale<sup>1</sup> , Paola Lenzi <sup>1</sup> \*, Gloria Lazzeri <sup>1</sup> , Alessandra Falleni <sup>2</sup> , Francesca Biagioni <sup>3</sup> , Larisa Ryskalin<sup>1</sup> and Francesco Fornai 1, 3 \*

*<sup>1</sup> Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Italy, <sup>2</sup> Department of Clinical and Experimental Medicine, University of Pisa, Italy, <sup>3</sup> I.R.C.C.S., Neuromed, Pozzilli, Italy*

#### Edited by:

*Manoj Kumar Jaiswal, Columbia University Medical Center, USA*

#### Reviewed by:

*Nina Kaludercic, National Research Council of Italy, Italy Raman Chandrasekar, Kansas State University, USA Nektarios Tavernarakis, Foundation for Research and Technology - Hellas, Greece*

#### \*Correspondence:

*Paola Lenzi paola.lenzi@med.unipi.it; Francesco Fornai francesco.fornai@med.unipi.it*

Received: *30 June 2015* Accepted: *15 October 2015* Published: *06 November 2015*

#### Citation:

*Natale G, Lenzi P, Lazzeri G, Falleni A, Biagioni F, Ryskalin L and Fornai F (2015) Compartment-dependent mitochondrial alterations in experimental ALS, the effects of mitophagy and mitochondriogenesis. Front. Cell. Neurosci. 9:434. doi: 10.3389/fncel.2015.00434*

Amyotrophic lateral sclerosis (ALS) is characterized by massive loss of motor neurons. Data from ALS patients and experimental models indicate that mitochondria are severely damaged within dying or spared motor neurons. Nonetheless, recent data indicate that mitochondrial preservation, although preventing motor neuron loss, fails to prolong lifespan. On the other hand, the damage to motor axons plays a pivotal role in determining both lethality and disease course. Thus, in the present article each motor neuron compartment (cell body, central, and peripheral axons) of G93A SOD-1 mice was studied concerning mitochondrial alterations as well as other intracellular structures. We could confirm the occurrence of ALS-related mitochondrial damage encompassing total swelling, matrix dilution and cristae derangement along with non-pathological variations of mitochondrial size and number. However, these alterations occur to a different extent depending on motor neuron compartment. Lithium, a well-known autophagy inducer, prevents most pathological changes. However, the efficacy of lithium varies depending on which motor neuron compartment is considered. Remarkably, some effects of lithium are also evident in wild type mice. Lithium is effective also *in vitro*, both in cell lines and primary cell cultures from the ventral spinal cord. In these latter cells autophagy inhibition within motor neurons *in vitro* reproduced ALS pathology which was reversed by lithium. Muscle and glial cells were analyzed as well. Cell pathology was mostly severe within peripheral axons and muscles of ALS mice. Remarkably, when analyzing motor axons of ALS mice a subtotal clogging of axoplasm was described for the first time, which was modified under the effects of lithium. The effects induced by lithium depend on several mechanisms such as direct mitochondrial protection, induction of mitophagy and mitochondriogenesis. In this study, mitochondriogenesis induced by lithium was confirmed *in situ* by a novel approach using [2-3H]-adenosine.

Keywords: mitochondria, amyotrophic lateral sclerosis, autophagy, lithium, G93A transgenic mouse, motor neuron, electron microscopy, biogenesis of mitochondria

## INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is characterized by progressive neurodegeneration of motor neurons within spinal cord, brainstem, and motor cortex (Charcot, 1874; Boillée et al., 2006). This may occur either sporadically (sALS) or depending on a number of gene mutations with familial transmission (fALS), all these mutations leading to autophagy impairment (Pasquali et al., 2014; Cirulli et al., 2015). Twenty percent of fALS occurs due to mutations in the gene coding for the enzyme copper-zinc superoxide dismutase (SOD-1, Rosen et al., 1994), which produces mitochondrial toxicity (Vehviläinen et al., 2014) while being an autophagy substrate (Kabuta et al., 2006). Both fALS and sALS are characterized by severe mitochondrial alterations. In fact, altered mitochondria within motor neurons of ALS patients were originally demonstrated by using electron microscopy (Hart et al., 1977; Hirano et al., 1984a,b; Sasaki, 2011; Ruffoli et al., 2015). Mitochondrial alterations represent a milestone within degenerating motor neurons in ALS. This may occur either primarily, through direct neurotoxicity of mutated proteins toward specific mitochondrial targets (Martin et al., 2007; Vehviläinen et al., 2014), or it may be produced by a defective clearance of altered mitochondria (impaired autophagy and mitophagy).

The occurrence of defective autophagy in ALS was demonstrated for the first time in the last decade (Fornai et al., 2008a,b); and it was progressively validated through a growing number of studies (Pasquali et al., 2009; Ferrucci et al., 2010; Otomo et al., 2012; Shimada et al., 2012; Wang et al., 2012; Castillo et al., 2013; Ikenaka et al., 2013; Barmada et al., 2014; Cheng et al., 2015; Lee et al., 2015; Philips and Rothstein, 2015; Xiao et al., 2015; Yang et al., 2015) up to the very recent confirmative work by Xie et al. (2015a,b). This evidence was reviewed in this issue by Ruffoli et al. (2015). At the same time, within ALS mitochondria an altered calcium buffering activity is constantly described (Higgins et al., 2002; von Lewinski and Keller, 2005; Jaiswal, 2013). This was intensely investigated in the past two decades to elucidate the process of primary mitochondrial damage (Barrett et al., 2014; Vehviläinen et al., 2014), which still requires a prompt mitochondrial removal (mitophagy; Fornai et al., 2008a; Laird et al., 2008; Pasquali et al., 2009; Ruffoli et al., 2015) as well as the synthesis of novel mitochondria (Fornai et al., 2008a,b; Melser et al., 2015; Roy et al., 2015; Ruffoli et al., 2015). A few years ago we demonstrated that, when removal of aged mitochondria within motor neurons is promoted through the autophagy machinery (mitophagy), a concomitant stimulation of the biogenesis of novel mitochondria takes place (Fornai et al., 2008a,b). This is now explained mechanistically by the recent work of Palikaras et al. (2015a,b) who unraveled a specific signaling pathway which binds mitophagy to mitochondriogenesis. Thus, an increase of mitochondrial removal eventually leads to the biogenesis of novel mitochondria (Palikaras et al., 2015a,b). This provides the molecular mechanism explaining why lithium, which is a powerful autophagy (and mitophagy) inducer (Sarkar et al., 2005; Sarkar and Rubinsztein, 2006; Pasquali et al., 2010; Klionsky et al., 2012; Motoi et al., 2014) concomitantly stimulates mitochondriogenesis (Fornai et al., 2008a,b), while protecting against primary mitochondrial damage (Bachmann et al., 2009) and excitotoxicity in the spinal cord (Young, 2009; Calderó et al., 2010; Fulceri et al., 2011). Despite mitochondria are key in the course of motor neuron degeneration in ALS, recent data demonstrate that, when mitochondria are protected in ALS mice, the time course of motor palsy and lethality is not modified (Parone et al., 2013). These data were obtained by suppressing the synthesis of cyclophilin D (CycD-KO), which contributes to the opening of the mitochondrial permeability transition pore. These CycD-KO mice have been generated in different strains of SOD-1 transgenic ALS mice. In all these ALS mice knocking out CycD preserves mitochondria and prevents the loss of motor neurons in the spinal cord. However, no improvement of motor symptoms and survival is obtained (Parone et al., 2013).This is due to the ongoing loss of peripheral motor innervation which persists despite the preservation of motor neuron perikaria.

Unexpectedly, in CycD-KO mice axonal mitochondria were protected as it occurred upstream in the cell body despite muscle denervation persisted. However, mitochondrial preservation within motor axons of CycD-KO mice was measured by Parone et al. (2013) only within ventral roots (proximal axons), while motor axon degeneration was measured at peripheral level within denervated muscles (distal axons). This leaves open the chance that mitochondrial damage within peripheral axons still occurs compared with proximal axons and cell bodies. This calls for additional studies to document the ultrastructure of distal (peripheral) compared with proximal axons before ruling out the protective role of mitochondrial preservation. In fact, it may occur that mitochondrial integrity measured in the ventral root is no longer present in peripheral axons. On the other hand, if mitochondrial alterations in distal axons are prevented despite an ongoing muscle denervation, alternative hypothesis of pathological targets need to be taken into account. In this context, altered axonal structure beyond mitochondria needs to be investigated. In recent years, the axonal transport was shown to be impaired in ALS. This may lead to abnormal accumulation of organelles and protein aggregates. In this way, a combined defect in axonal transport including mitochondria may converge in determining axonal damage. For instance, a specific impairment of axonal transport for mitochondria in G93A mice was clearly documented (Magrané et al., 2012, 2014). Since mitochondria are mostly synthesized at the level of the cell body, this may lead to increasing deficit of mitochondria along axonal length. This is expected to produce the most severe mitochondrial reduction at the level of the most remote segments of distal axons. If this is the case, apart from mitochondrial protection, the presence of mitochondria, comparing proximal with distal axons, should be evaluated.

In any case, motor axon degeneration represents a dominant effect compared with the loss of cell bodies in the course of ALS. Similarly, the ultimate anatomical target to promote neuroprotection in ALS shifts from the spinal cord to the muscle. This leads to focus on diverse motor neuron compartments (primarily distal axons compared with proximal axons, cell body, and dendrites).

Therefore, in the present study ultrastructural morphometry of compartment-dependent mitochondrial alterations occurring in G93A mice was carried out. In detail, we used quantitative ultrastructural morphometry to count mitochondrial alterations occurring at the level of dendrites, neuronal cell bodies, and motor axons. This latter compartment was further evaluated to measure mitochondrial alterations within distal compared with proximal axons. We examined at transmission electron microscopy (TEM) changes of mitochondrial size, amount, distribution, and mitochondrial alterations involving selective organelle architecture and swelling. Moreover, to understand the phenomena of axonal degeneration beyond mitochondrial level we investigated the fine ultrastructure of additional sub-cellular components within ALS proximal and distal motor axons. This was measured both in baseline conditions, in G93A SOD-1 transgenic mice compared with wild types, and following lithium to produce autophagy stimulation (Sarkar et al., 2005; Sarkar and Rubinsztein, 2006; Fornai et al., 2008a; Pasquali et al., 2010; Klionsky et al., 2012; Motoi et al., 2014). In fact, lithium is known for a long time as a mood stabilizer and it was recently shown to possess neuroprotective effects in various degenerative disorders, including those affecting motor neurons (Madeo et al., 2009; Luo, 2010; Shimada et al., 2012; Chiu et al., 2013; Agam and Israelson, 2014; Lieu et al., 2014; Mohammadianinejad et al., 2014; Motoi et al., 2014; Yáñez et al., 2014; Dwivedi and Zhang, 2015). This is in line with recent data showing that all autophagy inducers being tested so far protect in ALS experimental models. This is the case of rapamycin (Wang et al., 2012; Cheng et al., 2015), trehalose (Schaeffer et al., 2012; Castillo et al., 2013; He et al., 2015), resveratrol (Han et al., 2012), spermidine (Wang et al., 2012), tamoxifene (Wang et al., 2012), carbamazepine (Wang et al., 2012), valproate (Boll et al., 2014; Wang et al., 2015), histone deacetylase 6 (Chen et al., 2015), food starvation (Wiesner et al., 2015). This was also achieved by increasing expression of autophagy activating genes (Hetz et al., 2009; Oliván et al., 2015; Yang et al., 2015). Lithium administration may not induce autophagy if lithium plasma levels are way below those required to stimulate the autophagy pathway and produce any therapeutic effect (Pizzasegola et al., 2009; Chiu et al., 2013). Since, lithium-induced ultrastructural changes produce marked effects on mitochondria, involving size, shape, number and architecture, we further detailed these effects in vitro in cell lines to reduce experimental variables. In these experimental conditions, we could replicate the effects observed ex vivo in the spinal cord and we set up the ultrastructural measurement of mitochondriogenesis. In particular, we wish to provide for the first time in situ morphological evidence for mitochondriogenesis by high resolution (TEM) autoradiography of [2-3H]adenosine.

### MATERIALS AND METHODS

### Ex Vivo Study

#### Animals

Male B6SJL-TgN (SOD1-G93A)1 Gur mice, expressing human G93A SOD-1 mutation (N = 10) and wild type (WT) littermates (N = 10) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) via Charles River (Calco, LC, Italy). Mice received food and water ad libitum and were housed under controlled conditions: 12 h light/dark cycle, 21◦C room temperature.

This study was carried out in agreement with the European Council directive (86/609/ EEC). The experiments were approved by the Ethical Committee at the University of Pisa (protocol n. 14540, November, 21st, 2011).

#### Experimental Groups and Treatments

Transgenic G93A SOD-1 mice and their WT littermates were divided into four experimental groups: lithium-treated G93A mice (N = 5); vehicle (saline, sodium chloride 0.9%)-treated G93A mice (N = 5); lithium-treated WT mice (N = 5) and vehicle-treated WT mice (N = 5). All treatments started at 67 days of age, which corresponds to a pre-symptomatic stage, and they were carried out every other day up to a tetraplegic stage (end point), when mice were sacrificed with deep anesthesia (chloral hydrate). The tetraplegic stage was defined when mice were no longer able to get up from a lying position within a 30 s time interval (Parone et al., 2013; Fornai et al., 2014). This end point was chosen in order to avoid discomfort due to impaired feeding, drinking and breathing (Tankersley et al., 2007; Fornai et al., 2014). Lithium chloride (Sigma, St. Louis, MO, USA) was administered (both to G93A and WT mice) at the dose of 1 mEq/Kg, i.p. dissolved in 200µl of saline. All treatments were carried out between 9.00 and 12.00 am.

#### Transmission Electron Microscopy of Spinal Cord and Muscle

Following deep anesthesia mice were trans-cardially rinsed with saline solution (0.9%) and they were fixed by perfusion with 2% paraformaldehyde/0.1% glutaraldehyde in 0.1 M phosphate buffered saline, pH = 7.4. The spinal cord and gastrocnemius muscle were dissected. Spinal cords were kept overnight at 4◦C in situ using the same fixing solution.

#### Specimens from Spinal Cord

The lumbar tract of the spinal cord was surgically dissected to avoid any abnormal pressure and post-fixed in a 1% OsO<sup>4</sup> buffered solution for 1 h and 30 min, and dehydrated in ethanol, and embedded in Epon-araldite.

For each spinal cord sample two tissue blocks (volume of 5 mm<sup>3</sup> ) were cut to obtain an average of 20 grids. Each grid included at least 5 cells, which were analyzed along nonserial sections. Motor neurons were selected based on classic morphological features (multipolar cells with dispersed nuclear chromatin and prominent nucleoli). In order to improve the selection of motor neurons we also applied a size exclusion criterion. This consists of excluding from counts those lamina IX neurons measuring less than 30µm of maximum diameter. This limits the analysis to phasic alpha motor neurons ruling out gamma motor neurons and most tonic alpha motor neurons, but it guarantees to rule out type I Golgi projecting neurons. This size-exclusion criterion is validated by several previous studies adjusted to various mouse strains (Morrison et al., 1998; Martin et al., 2007; Fornai et al., 2008a, 2014; Ferrucci et al., 2010; Fulceri et al., 2012).

Morphometry of motor neurons was carried out by using TEM at 8000x magnification. Specimens form spinal cord were used to analyze proximal motor axons and glial cells surrounding motor neurons in the ventral horn.

#### Specimens from Muscle

Each gastrocnemius muscle from each animal was dissected and gently stretched for 10 s before being immersed for 1 h and 30 min in the fixing solution used for perfusion. Samples were then post-fixed in buffered 1% OsO<sup>4</sup> for 1 h and 30 min, dehydrated in ethanol and processed as described above. To guarantee a homogeneous analysis of the same muscle area in each mouse, we cut little blocks in the central part of the belly at the level of the wider muscle size each measuring a volume of 5 mm<sup>3</sup> . Since various muscle fibers orientation may lead to different structural perspectives and different measurements, each block was cut following the same longitudinal orientation. This procedure allows to keep constant the reference points and to follow the course of peripheral nerve fibers within muscle length allowing to reduce as much as possible experimental bias. Analysis at TEM was oriented by a previous light microscopy observation on 1–2µm thick serial semi-thin sections, which were cut using a Porter Blum MT-1 or an ultramicrotome Reichert-Jung. These slices were stained with 1% toluidine blue and 1% methylene blue in 1% sodium tetraborate and they were analyzed to verify homogeneity of muscle segments and nerve fiber tracts.

Analysis of gastrocnemius muscle and peripheral nerve fibers was carried out by using TEM at 2500–6000x magnification. Muscle architecture was examined as follows: (i) intersarcomeric area, defined by the space intermingled between two sarcomers; (ii) density of mitochondria in the muscle, defined by the number of mitochondria per surface unit; (iii) percentage of altered mitochondria; (iv) mitochondrial diameter; (v) diameter of sarcoplasmic reticulum cisternae.

In peripheral nerve fibers the following parameters were analyzed: (i) density of mitochondria; (ii) percentage of altered mitochondria (defined below); (iii) mitochondrial diameter; (iv) axons owing enlarged and/or fused myelin layers.

#### Analysis of Mitochondrial Alterations

Mitochondria were defined as altered according to criteria being validated by previous morphological studies (Ghadially, 1988; Fornai et al., 2008a; Lenzi et al., 2012) as follows: (i) significantly decreased electron density of the matrix (dilution, vacuolization, cavitation); (ii) fragmented and ballooned cristae (intracristal swelling); (iii) partial or complete separation of the outer and inner membranes; (iv) mitochondrial swelling. This latter criterion allows to distinguish mitochondrial enlargement which might develop in physiological conditions from pathological swelling which appears as an increase in mitochondrial size due to a swelling of the mitochondrial structure.

Accordingly, the following data were calculated: (i) density of mitochondria in dendrites; (ii) density of mitochondria in the cell body; (iii) density of mitochondria in proximal and distal motor axons; (iv) percentage of altered mitochondria in dendrites, cell bodies, and motor axons; (v) mitochondrial swelling in dendrites, cell bodies and proximal and distal motor axons.

#### Additional Morphometric Analysis

In order to document ultrastructural alterations other than mitochondrial changes we analyzed the following: (i) dendrites diameter; (ii) ultrastructure of glial cells; (iii) clogging of motor axons. This was measured both as the percentage of clogged axonal area and percentage of motor axon with total clogging. Clogged motor axons were defined as axons filled with an electron-dense compact material in which membranes, mitochondria, and neurofilaments were no longer distinguishable but jamming in the axoplasm. Total clogging was defined by a condition in which axoplasm was no longer visible in the ultra-thin section due to the total filling with such electron-dense compact material. In some cases, the concomitant disarrangement of myelin sheet led inner myelin layers to project within axoplasm, thus contributing to axonal clogging.

In order to document the autophagy status, we analyzed double and multiple membranes vacuoles. This was done to confirm the occurrence of big stagnant autophagy vacuoles in ALS mice and their removal following lithium. These vacuoles were further identified by immunogold particles staining for LC3 and beclin-I according to the gold standard procedure to identify autophagy vacuoles.

For immuno-electron microscopy we did not apply the routine method, consisting of acrylic resins for embedding, since this procedure preserves antigens but impairs preservation of the fine morphological architecture (Lenzi et al., 2012). Thus, samples were embedded in Epon-araldite, as previously described, and incubated for 24 h at 4◦C with primary antibodies anti-beclin-I (Santa Cruz Biotechnology Inc., Dallas, TX, USA, diluted 1:10) and LC3 (Santa Cruz Biotechnology Inc., diluted 1:10) in a buffer solution (PBS, 1% goat-serum, and 0.2% saponin). After washing in phosphate buffer, sections were incubated with gold-conjugated secondary antibodies (with gold particles averaging 10 nm for beclin-I and 20 nm for LC3, respectively, Sigma) diluted 1:20 in PBS for 1 h, at room temperature. This allowed co-staining for both antigens. Finally, sections were fixed with 1% glutaraldehyde and stained with uranyl acetate and lead citrate and they were examined under TEM. Control sections were not exposed to the primary antibody but they were directly incubated with the secondary antibody. This immunocytochemistry was aimed at specific recognition of autophagy vacuoles, therefore no quantitative measurement of immunogold particles was carried out.

On the other hand, immunocytochemistry for mitochondrial SOD-1 (Assay Designs, Ann Arbor, MI, USA, dilute 1:10) was carried out for quantitative purposes. SOD-1 was revealed stoichiometrically by immunogold particles (10 nm, Sigma). The count of particles localized within mitochondria was obtained from two plastic blocks randomly chosen from each experimental group. Immunogold particles were counted at TEM with a magnification of 15,000x.

#### Statistics for Ex Vivo Studies

Values were expressed either using the absolute value or as percentage of normal numerical distributions. Data are reported as the mean or the mean percentagee ± S.E.M. This was carried out for altered mitochondria, SOD-1 immunogold particles or clogged motor axons. Inferential statistics to compare groups was carried out by using One-way analysis of variance, ANOVA, with Sheffè's post-hoc analysis (H<sup>0</sup> probability was rejected when less than 5%, P ≤ 0.05).

We wish to emphasize here the need of calculating mitochondrial density (the number of mitochondria per surface unit expressed in µm<sup>2</sup> ). In fact, the rough number of mitochondria within compartments of motor neuron may introduce experimental bias, since both disease and treatment modify motor neuron diameter (Martin et al., 2007; Fornai et al., 2008a), which in turn could bias the mitochondrial count (which could be either diluted or concentrated). Again, despite size exclusion criteria, different sizes between various compartments would not be comparable. This also applies to measurement carried out within muscle fibers. Changes in mitochondrial volume were considered separately from mitochondrial swelling which was a pure mitochondrial damage.

### In Vitro Studies

#### Primary Spinal Cord Cultures

Mixed primary cell cultures from spinal cords were obtained from 14-days-old mice embryos as previous described (Carriedo et al., 1996). Pregnant (14 days gestation) Swiss Webster female albino mice were purchased from Charles River Laboratories (Calco, Lecco, Italy). The spinal cords were dissected and both meninges and dorsal root ganglia were removed. The cords were then incubated for 10 min in 0.025% trypsin and cells were dissociated. The cell cultures were plated at a density of 3 "spinal cords" per well plate (35 mm dishes) on poly-D-lysine-coated glass coverslips and maintained in D-MEM supplemented with 5% FBS and 5% HS. Twenty-four hours after plating, the medium was replaced with Neurobasal supplemented with B-27 and 0.5 mM glutamine. Three days after plating cytosine arabinoside (10µM) was added and medium was changed every 3 days. For analysis at light microscopy, cells grown on glass coverslips, within 35 mm diameter dishes, and they were transferred into cell culture plates 24 well, to perform immunohistochemistry for SMI32.

Spinal cord cultures at 8–9 days were used for toxicity experiments. The cells were administered the autophagy blocker 3-methyladenine (3-MA) 5, 10, and 20 mM for 4 h with or without lithium chloride at the dose of 0.5 mEq/L. After this time, the cultures were rapidly washed and maintained in the free medium with lithium or vehicle for 24 h. To identify motor neurons, immunocytochemistry for SMI32 was performed on primary cultures. At first, cells were fixed in a solution containing 4% paraformaldehyde in 0.1 M PBS, pH 7.3 (10 min); after washing in PBS cell cultures were incubated with Triton-X 0.1% in PB (15 min), followed by 3% hydrogen peroxide (10 min). Before incubation with the primary antibody, a blocking solution (10% normal goat serum in PBS) was added for 1 h at room temperature. The SMI32 primary antibody (mouse, 1:1000; Covance, Emeryville, CA, USA) was dissolved in PBS containing 2% normal goat serum and incubated overnight at 4◦C. After rinsing in PBS the anti-mouse secondary biotinylated antibody (Vector Laboratories, Burlingame, CA USA) was used at a dilution of 1:200 for 1 h at room temperature, followed by incubation with ABC kit (1 h at room temperature, Vector Lab) and diaminobenzidine (Vector Lab). Immunostaining was analyzed at light microscopy (Nikon Eclipse 80i, Japan).

The effects of each treatment on the survival of motor neurons were measured by counting the number of SMI32 positive motor neurons in a total of 3 slides per group. All measurements were carried out at 20x magnification. Values are given as the percentage mean ± S.E.M. Comparisons between groups were made by using ANOVA with Sheffè's post-hoc analysis.

#### Cell Lines

PC12 cells were obtained from the American Type Culture Collection and they were grown in RPMI 1640 medium supplemented with 10% heat-inactivated horse serum, 5% fetal bovine serum, penicillin (50 IU/mL), and streptomycin (50 mg/mL). Cells were seeded in 6-well plates at 1 × 10<sup>6</sup> cells in a final volume of 1 ml/well and incubated at 37◦C in 5% CO<sup>2</sup> for 24 h.

PC12 cells were treated either with lithium chloride (1 mEq/L for 48 h) which was dissolved in culture medium (3 ml per dish) or they were added 3 ml of plain culture medium (controls). When studied for autoradiography both groups of cells received 1µCi of [2-3H]adenine for 2 h, starting at 46 h after lithium/medium administration. Both lithium concentration, and the amount and time of exposure to [2- <sup>3</sup>H]adenine were selected based on pilot studies. [2-3H]Adenine (15-25 Ci mmol−<sup>1</sup> ) was purchased from the Radiochemical Centre (Amersham, Bucks, UK). L4 photographic emulsion for electron microscopy autoradiography was obtained from Ilford (UK). In addition, plain ultrastructural morphometry (no autoradiography) was carried out from PC12 dishes which were treated with lithium or plain culture medium but they were not exposed to [2-3H]adenine. This was necessary to avoid that electron density of [2-3H]adenine impairs the fine morphometry of mitochondria and other organelles.

Cells were centrifuged at 1000 g for 5 min. After removal of the supernatant, each pellet was thoroughly rinsed in PBS. Fixation was carried out with a solution containing 2.0% paraformaldehyde/0.1% glutaraldehyde in 0.1M PBS (pH 7.4) for 90 min at 4◦C. Specimens were post-fixed in 1% OsO<sup>4</sup> for 1 h at 4◦C, dehydrated in ethanol and embedded in Epoxy-resin. For routine TEM, ultrathin sections were contrasted with uranyl acetate and lead citrate, and examined using a Jeol JEM SX 100.

For ultrastructural morphometry, sections were examined directly at TEM at a magnification of 6000x. Each grid containing enough non-serial sections to count at least 50 cells. Several grids were observed in order to obtain a total number of at least 200 cells for each experimental group. For routine microscopy, the total number of mitochondria per cell, the number of altered mitochondria and their diameter was also measured. The altered mitochondria were described as mentioned above for ex vivo samples according to Ghadially (1988).

For electron microscopy autoradiography, ultra-thin sections were collected on Formvar-coated nickel grids and mounted on cork caps made adhesive with double-sided sticky tape. A monolayer of jellified Ilford L4 emulsion was then placed onto the nickel grids firmly held by the cork caps using a 2 cm wide platinum wire loop. They were then placed in black light tight boxes and exposed for up to 30 days at 4◦C in a dry environment. By the end of this period, the emulsion over the grids was developed in D19 developer, fixed and thoroughly washed in distilled water.

The number and size of silver grain clusters were carefully calculated in each cell. In particular, their distribution in nuclear and cytosolic compartments was distinguished. Autoradiography grains as following: (i) number of grain clusters placed within or close to mitochondria up to 250 nm distance between the cluster contour and the closest mitochondrial outer membrane; (ii) diameter of grain clusters placed within or close (up to 250 nm distance) mitochondria (defined as previously); (iii) ratio between grain clusters placed within or close (up to 250 nm distance) to mitochondria and total grain clusters in the cytoplasm.

We counted a total of 20 cells per group. Comparisons between groups were made by using ANOVA with Sheffè's posthoc analysis to compare various distances from mitochondria. Data were analyzed using Student's t-test for unpaired data when two groups at a single distance were compared.

As for ex vivo studies, we calculated both the total number and percentage of altered mitochondria. Due to a lack of dendritic/somatic/axonal compartments in PC12 cells and homogeneous diameter of the cell line, we did not find any significant difference in providing mitochondrial data as expressed in rough number or in density. In contrast, when analyzing autoradiography data it was essential to compartmentalize morphometrical calculation of autoradiography grains based on their distance from mitochondria. In detail, the distance of 250 nm was selected as the critical length for which the highest difference was observed between lithium and vehicle-treated cells. At this distance, in lithium treated cells the number of grain clusters was the highest and it does not differ significantly from the number of grain clusters counted on the mitochondrial surface (**Figure 1**). We counted cytoplasmic grain clusters placed both on mitochondria and at different lengths from mitochondrial contour. This was done at progressive (100 nm) distances. Grain clusters were counted within homogeneous areas which corresponded to squares measuring 100 nm side. Under lithium administration these measurements were homogeneous within mitochondria and within an area which was distant 250 nm from the outer mitochondrial membrane. After the critical length of 250 nm, clusters which were counted within squares at progressive 100 nm distance possessed a number of grain clusters which decreased geometrically. In this way we sought to determine

FIGURE 1 | Compartmentalization of autoradiography grain clusters. Cytoplasmic grain clusters are counted in the presence of lithium (squares-labeled curve) or vehicle only (circles-labeled curve). Grain clusters express the presence of [2-3H]adenine within mitochondria (*X* = 0) and at various distances from the mitochondrial outer membrane which was expressed in nm (*X* = Nnm). The number of grain clusters for each *X*-value is reported on the Y axis. Grain clusters were counted within homogeneous areas which correspond to squares measuring 100 nm side. Under lithium administration these measurement were homogeneous within mitochondria and within an area which was distant 250 nm from the outer mitochondrial membrane. After the critical length of 250 nm, squares counted at progressive 100 nm distance possessed a number of grain clusters which decreased geometrically. Each point reports the mean ± S.E.M. of clusters. This generated a highly polarized distribution. In contrast, when cells were administered vehicle, the number of grain clusters was distributed quite homogeneously with no significant difference. This indicates that lithium produces a source of autoradiography which is centered on mitochondria according to the model published by Salpeter et al. (1978). For each point values represent the mean ± S.E.M. Comparisons between points were made by using ANOVA with Sheffè's *post-hoc* analysis. \**P* ≤ 0.05 compared with control.

whether a theoretical source of autoradiography could be identified at mitochondria. In fact, assuming that [2-3H]adenine binds to novel synthesized mitochondrial DNA, mitochondria become the source of autoradiography diffusion according to the model provided by Salpeter et al. (1978). This assumption is based on the occurrence of authentic mitochondriogenesis under the effects of lithium. As reported in **Figure 1**, for lithiumtreated cells, the number of autoradiography clusters starts to decrease in areas placed at a distance of more than 250 nm from mitochondrial contour. In contrast, in vehicle-treated cells there is no critical length. This demonstrates the occurrence of a polarization in the distribution of autoradiography clusters, which occurs only when lithium is present. Thus, squared area of 100 nm side ranging within mitochondria and around up to 250 nm length from mitochondria, correspond to the critical spot in which the difference in the number of autoradiography clusters reaches the maximal polarization and the maximal difference between lithium and vehicle treated cells. This starting point pertains to methodology while providing already experimental data speaking out for in situ evidence of mitochondriogenesis induced by lithium. The occurrence of nuclear staining was clearly independent from mitochondria (ranging way in excess of 250 nm and it was bound to nuclear DNA). Of course this staining does occur but it does not interfere with the count in the cytosol where the staining was really scarce at distance from mitochondria. Thus, two sources of radioactive staining occur: (i) nuclear chromatin; (ii) cytosolic mitochondria. Nuclear and cytosolic compartments are generally separated one from each other by large areas of faint randomly placed staining, which is similar to the loss of radioactivity occurring in the cytosol for areas placed between mitochondria. When a mitochondrion was placed at a distance not exceeding 250 nm from the nucleus (less than 10% of mitochondria) this was not included in the count. This was done to avoid bias of contaminating counts with the nuclear source of clusters. The cluster density we counted within mitochondria corresponds to that measured within a 250 nm range from mitochondrial contour. This explains graph in **Figure 1** where the Y-values of radioactive clusters was constant in the interval between X = 0 (corresponding to the mitochondrial surface) and X = 250 nm. When counts were carried out in space intervals extending above 250 nm, at 100 nm progressive intervals we measured a geometric fall in the number of clusters (**Figure 1**). This curve of cluster distribution produced by lithium was abolished in vehicle-treated cells.

## RESULTS

### Morphometry of Mitochondria within Cell Bodies of Motor Neurons, Lithium–and Disease-modifying Effects

**Figure 2A** shows representative semi-thin sections from the anterior horn of WT and G93A mice treated either with vehicle or lithium. It is evident the general shape of motor neurons which appear as multipolar cells, in which nucleus is not condensed and the nucleolus is well evident. The size of motor neurons corresponds to the size exclusion criterion. Nonetheless, in these representative pictures we reported the disease-dependent increase in motor neuron size (compare G93A vehicle with WT vehicle) and the effects of lithium, which brings back motor neuron diameter to control values (compare G93A lithium with WT vehicle). Interestingly, as we already published (Fornai et al., 2008a), the size of motor neurons in WT is further decreased by the effects of lithium. This size change does not exceed 10% of diameter and it does not affect significantly the size exclusion criterion. In addition, within motor neurons of G93A mice a prominent vacuolization is evident which is markedly reduced by lithium administration (**Figure 2A**). These semi-thin sections addressed further studies at electron microscopy reported in representative **Figure 2B** showing well-shaped and regularly-sized mitochondria in WT mice compared with swollen mitochondria of G93A mice, which possess matrix dilution and breaking of the cristae. This is further documented in **Figure 3** showing swollen mitochondria with fragmented and swollen cristae with a rupture of both inner and outer membranes. Remarkably, mitochondria from ALS mice are abnormally swollen up to an average diameter which is way in excess to what it was measured in WT mice (0.69 ± 0.05 and 0.38 ± 0.025 respectively, **Figure 4**). These altered mitochondria often appear within autophagy vacuoles which are identified by the gold standard procedure according to Klionsky et al. (2012) as TEM-identified double membrane organelles (**Figure 3A**), staining both for LC3 and beclin I (**Figure 3B**). In these mice, the amount of giant/altered mitochondria, which was found within big stagnant autophagy vacuoles, was very high (percentage of 75.25 ± 3.1 %, **Figure 4**). As shown in **Figure 2B**, the mitochondrial alterations were reduced by lithium, thus providing a gold standard TEM validation for lithium-induced autophagy. In detail, as counted in graphs of **Figure 4** lithium increases the density of mitochondria (**Figure 4A**), it reduces the amount of altered mitochondria (**Figure 4B**), and it reduces mitochondrial diameter (both in normal and swollen mitochondria, **Figure 4C**). Thus, in ALS mice administered lithium for about 170/180 days we found a marked suppression of giant swollen mitochondria with slighter matrix dilution and rare cristae fragmentation. These mitochondria were barely detectable within autophagy vacuoles, which in turn did not appear as big stagnant vacuoles but they were similar to controls.

### Morphometry of Mitochondria within Motor Neuron Dendrites, Lithium–and Disease-modifying Effects

Dendrites were partly reminiscent of what we already described within the neuron cell body (**Figure 5**). However, a striking effect concerns dendritic volume. G93A mice possess giant dendrites compared with WT mice as evident in representative pictures and in graph of **Figure 5A**.This effect is way more pronounced of what already described for the cell body of motor neurons. This explains why, despite an increase in the total number of mitochondria the density of these organelles was attenuated in the dendrites of G93A mice (**Figure 5B**). In detail, mitochondrial morphology was deranged

in G93A mice compared with WT. These effects were prevented by lithium, which produces a mitochondrial morphology similar to WT both concerning the architecture and diameter of dendritic mitochondria (**Figures 5C,D**). Lithium further augmented mitochondrial number but this effect was mitigated by the measurement of mitochondrial density (which is in line with the giant dendritic volume of G93A mice). When lithium was administered to WT mice an improvement of mitochondrial architecture and increase in mitochondrial density were observed (**Figure 5**). The percentage of altered mitochondria, which was significantly higher in G93A mice administered vehicle, was brought back to WT values following lithium administration (**Figure 5C**).

### Morphometry of Mitochondria within Proximal Motor Axons, Lithium–and Disease-modifying Effects

When observing representative pictures of motor axons, in G93A mice (**Figure 6**) dramatic mitochondrial alterations were evident which surpassed those described for mitochondria in the cell body and dendrites of motor neurons. Severe architectural alterations of mitochondria were fairly prevented by lithium (**Figure 6**).

Unexpectedly, while lithium increases the density of mitochondria in the cell body (**Figure 4A**) this does not occur within proximal axons where the density of mitochondria in G93A and WT mice treated with lithium was reduced similarly to G93A mice treated with vehicle (**Figure 6A**). Thus, within proximal axons lithium and the disease condition independently produce a similar mitochondria reduction (**Figure 6A**). However, altered mitochondria were markedly suppressed by lithium administration of G93A mice (**Figure 6B**). The percentage of altered mitochondria in the proximal axon was similar to that counted in the cell body. This is corroborated by representative pictures, which demonstrate mitochondrial impairment in proximal motor axons (see insert of **Figure 6**). However, in the proximal axon there is a significant decrease in the total number of mitochondria which does not occur in the cell body. This suggests that disease progression leads to a similar mitochondrial damage in the cell body and proximal axons but the number of mitochondria is frankly diminished only in the proximal axon. Interestingly, this effect was observed also in G93A mice administered lithium and, most remarkably it was measured in lithium-administered WT mice. One might argue that this count is explained by the lack of mitochondrial replacement (likely sustained by mitochondriogenesis) within proximal axons both in disease conditions as well as under accelerated mitophagy promoted by lithium. Noteworthy, the few mitochondria counted both in G93A and WT mice treated with lithium appear to be structurally preserved (representative pictures of **Figure 6**). This is confirmed by

counts of altered mitochondria and mean mitochondrial diameter within motor axons (**Figures 6B,C**). Moreover, one should consider that, when counting mitochondria at this proximal level within the cord or in the ventral root as carried in this paragraph and in the work by Parone et al. (2013), this does not necessarily reflect the mitochondrial status at distal axons where the critical damage occurs. This is the first time, which a separate count for axonal mitochondria, was carried out to distinguish what occurs in the proximal axons compared with peripheral axon which is reported in the next paragraph.

### Morphometry of Mitochondria within Distal Axons, Lithium–and Disease-modifying Effects

The morphometry of distal axons reveals remarkable differences compared with proximal axons. As shown in representative pictures of **Figure 7** the mitochondrial damage is further worsened in vehicle-administered G93A mice compared with every other compartment including proximal axons.

The axoplasm of distal axons is entirely jammed by abnormal structures where normal constituents are no longer distinguishable (see next paragraph). These abnormal structures include the myelin sheet, which seems to intrude the axoplasm itself and is altered (see insert). This is reverted by lithium. In fact in the peripheral axons the loss of mitochondria is prevented by lithium which increases mitochondrial number similarly to WT mice (**Figure 7A**) Lithium also provides significant protection against mitochondrial damage which is mostly evident in G93A mice administered vehicle (**Figure 7B**). The amount of mitochondrial damage measured in the distal axons involves almost the total number (more than 90%). This implies that preserved mitochondria in the distal axons are more than two-fold the preserved mitochondria within proximal axons, where only a subtotal (less than 80%) mitochondrial damage takes place (compare **Figure 7B** with **Figure 6B**).

These data fully confirm the occurrence of the worst mitochondrial damage in peripheral axons as described also in ALS patients (Shi et al., 2010).

These data indicate the need to extend the ultrastructural morphology of ALS axons toward the peripheral segments to unravel the significance of axonal compared with perikaryal alterations. This is reciprocated by the occurrence of protective effects of lithium both on mitochondrial density (**Figure 7A**), and mitochondrial alterations (**Figure 7B**). At this level, lithium reduced mitochondrial diameter due to a specific effect on preventing mitochondrial swelling (**Figure 7C**).

### Extra-mitochondrial Morphometry of Axons, Lithium–and Disease-modifying Effects

When observing the extra-mitochondrial alterations within motor axons (**Figure 8**) low magnification provides a clear perspective of axonal clogging which was further detailed in previous **Figure 7** by abnormal structures, which were jamming the axoplasm. In G93A mice administered vehicle this effect leads to electron-dense material which fills almost totally the axoplasm, which is cleared back significantly by lithium as shown by representative pictures and graphs of **Figure 8**. These jams are clearly visible in the insert at high magnification of **Figure 8** in vehicle-administered G93A mice as mixed structures clogging the axoplasm. These axonal inclusions are frankly pathological as witnessed by their amorphous structure and their total absence in axons from WT mice (**Figure 8A**). This contrasts with the occurrence of a critical clogging (more than 60%) of axoplasm which was present in several motor axons from G93A mice (**Figures 8A,B**), while it was limited

and Methods Section. Mitochondrial density within cell body is dramatically augmented by lithium administration both in WT and G93A mice (A). In line with the preservation of mitochondrial architecture observed in Figure 2B, lithium significantly reduces the percentage of altered mitochondria (B). As noticed from representative pictures a re-shaping effect of lithium is also counted The size of mitochondria was increased in G93A mice administered vehicle but it is brought back to WT values when G93A mice received lithium. Remarkably, such re-sizing effects of lithium also occur in WT mice, where lithium further decreases mitochondrial diameter in WT mice (C). Values are given as the mean ± S.E.M. Comparisons between groups was made by using One-way ANOVA. \**P* ≤ 0.05 compared with vehicle-treated mice in graph.

to 26% of the luminal area of a few motor axons in G93A mice administered lithium (**Figures 8A,B**). This material we described here selectively in the motor axon compartment is likely to produce significant deleterious effects concerning axonal physiology at least affecting axonal transport which indeed is functionally impaired (Magrané et al., 2012, 2014). In fact, the amount of clogged area almost reaches 70% of the axonal area measured in 2D electron microscopy. The administration of lithium produces a remarkable effect since it tears down the clogged area of more than two-fold from 63.58 to 26.56%, which means a net gain of axonal lumen of more than 50% (**Figure 8B**). Incidentally, the clearance induced by lithium of more than two-fold the clogging of motor axons is similar to the increase in mitochondrial density and the amount of spared mitochondria (more than two-fold), which was measured in the distal axons. This indicates that by increasing the available axonal lumen is associated to the increase of novel mitochondria transported downstream and the increased number of healthy mitochondria counted at distal level.

### Structure of Muscles, Lithium–and Disease-modifying Effects

In semi-thin sections from the gastrocnemius muscle (**Figure 9**) we could dissect the homogeneous segment to be analyzed under electron microscopy. The severe disarrangement of muscle structure in vehicle-administered G93A mice is clearly evident (**Figure 9**). Similarly, the re-alignment and the lack of abnormal architecture of muscle fibers can be appreciated in **Figure 9** showing G93A mice administered lithium.

When examined at TEM, representative pictures of **Figure 10** show that, within muscle fibers from saline-administered G93A mice, there was a severe loss of the normal sarcomeric structure as well as the regular alignment of sarcomers. These muscle fibers from vehicle-administered G93A mice could no longer be recognized as composed of sarcomeric units. The severe disarrangement which appears in G93A mice was measured by counting the distance between vestigial sarcomers (**Figure 10A**) which was increased more than ten-fold compared with WT. In these muscle fibers the effects of lithium were most remarkable since the intersarcomeric area of lithium-administered G93A mice was re-instated to values similar to WT. This is well evident also from representative pictures cited from **Figure 10**, showing that lithium substantially preserves the sarcomeric architecture in the muscle fibers. The fine alignment of sarcomers was even more geometrically correct when WT mice were administered lithium, but this pertains only to the appreciation of the pictures and it was not specifically measured. It is likely that, the density of mitochondria under the effects of lithium might contribute to this effect. In fact, when counting the number of mitochondria in G93A mice a massive decrease was measured, which was significantly prevented by lithium (**Figure 10B**). Interestingly, in WT mice there was a non-significant trend toward a decrease in the number of mitochondria. In any case dramatic mitochondrial alterations in the muscle of G93A mice were totally abolished by lithium administration as can be appreciated in representative pictures of **Figure 11** and as reported by the counts of graph in **Figure 11A**. This effect was also evident by measuring the mitochondrial swelling in each group which was graphed in **Figure 11B** as an increase in mitochondrial diameter. Remarkably, the swelling of sub-cellular structures in the muscle of G93A mice extends to the sarcoplasmic reticulum which was increased three-fold compared with wild types (representative pictures of **Figure 11** and graph of **Figure 11C**). Again, the administration of lithium abolished such an alteration. Thus, the preservation of various components of the gastrocnemius muscle structure was remarkable under the effects of lithium. This is key, since the loss of muscle structure and architectural disruption which occurs otherwise in vehicle-administered G93A

FIGURE 5 | Ultrastructural morphometry of mitochondria within dendrites of motor neurons. Mitochondria within dendrites of motor neurons of the anterior horn of mice spinal cords are representatively reported. Mitochondria from dendrites of vehicle-treated G93A mice possessed swollen cristae (arrows) which do not appear neither in WT nor in lithium-treated G93A mice. Low magnification representative picture from vehicle-treated G93A mice shows giant dendrites (arrowhead) compared with WT mice, this is fully reversed by lithium as counted in graph (A). Graph (B) reports a slight non-significant reduction of mitochondrial density in vehicle-treated G93A mice compared with WT, which was not affected by lithium. In contrast, altered mitochondria were significantly increased in vehicle-treated G93A mice compared with WT mice, although lithium treatments restore the percentages of altered mitochondria in G93A to values similar to WT mice (C). The mitochondrial diameter in dendrites from G93A mice was slightly but significantly increased following vehicle, but it was normal following lithium (D). Values are the mean ± S.E.M. Comparisons between groups were made by using One-way ANOVA. \**P* ≤ 0.05 compared with other groups in graph. Scale bars: WT vehicle and WT lithium = 0.3 µm, G93A vehicle and G93A lithium = 0.4 µm, G93A vehicle (low magnification) = 1.3 µm.

mice along with classic mitochondrial alterations were really impressive.

### In Vitro Study on Cell Lines and Primary Cultures from Spinal Cord

In order to better document the effects observed in vivo on mitochondrial morphology and number we measured the effects of lithium in a PC12 cell line which owns much less experimental variables. We further extended the mechanistic analysis on the role of lithium and autophagy in a primary cell cultures from spinal cord.

As shown ex vivo in the spinal cord, in PC12 line lithium increases two- fold the number of mitochondria compared with vehicle (**Figure 12A**). In this cell line, which owns aberrancies inherent with their shift from a rat pheocromocytoma (Fornai et al., 2007) the number of mitochondria, which eventually undergo spontaneous alterations is noticeable and reaches 15% of total mitochondria (**Figure 12B**). These spontaneous mitochondrial degeneration was suppressed by lithium administration (which reduced three-fold the percentage of altered mitochondria, **Figure 12B**). Thus, considering that lithium increases two-fold the total number of mitochondria, while it decreasing three-fold the percentage of altered mitochondria, the total number of mitochondria which are structurally preserved by lithium corresponds roughly to six-fold the number of preserved mitochondria measured in control. At the same time, lithium reduces significantly the mitochondrial diameter (**Figure 12C**). This size reduction was counted independently from mitochondrial swelling but it was counted considering only healthy mitochondria. This effect suggests

the occurrence of mitochondriogenesis which is known to produce small densely packed mitochondria with a well-marked architecture both concerning arrangement of the cristae and matrix electrondensity. In fact, when observed in representative TEM pictures (**Figure 12**), lithium-treated cells consistently show a dramatic increase in mitochondrial density, which clearly appears at first glance (see lower magnification in the left row of **Figure 12**). At the same time, at higher magnification (in the right row of **Figure 12**), the mitochondrial architecture was well-delineated in lithium-treated cells. These results confirm the hypothesis that lithium possesses mitochondriogenetic effects.

Based on cell imaging with mitotrack red and green and the quantitative analysis of mitochondrial genes by rtPCR we already indicated that lithium induces the biogenesis of mitochondria (Fornai et al., 2008a). In line with this, a few weeks ago evidence was provided that autophagy induction is dually bound to mitophagy and mitochondriogenesis (Palikaras et al., 2015a,b). However, no single in situ morphological evidence was ever produced to document the occurrence of mitochondriogenesis under the effects of lithium. Thus, in the present study we preadministered lithium chloride (1 mEq/L) or vehicle which was followed at 48 h by administration of [2-3H]adenine (1µCi). In these experimental conditions, which follow up experimental data reported in **Figure 1** and already commented in the Materials and Methods section we studied the distribution of autoradiography grains. As reported in representative **Figure 13**, these grains appeared as electrondense clusters of particles which were distributed in the cells very differently depending on exposure to vehicle or lithium. In detail, as shown in pictures of **Figure 13** in a vehicle-administered cell the clusters distribute quite homogeneously in the cytosol, while a marked hetereogenity, polarized toward mitochondria was observed in the cytoplasm following lithium. As reported in **Figure 1** this generates a hot area of [2-3H]adenine clusters which covers all the mitochondrial surface and the surrounding cytoplasm up to a 250 nm length from mitochondria. In this hot area, measurement

point to a particular zone of disrupted myelin sheaths with loss of stratification (see also insert). Mitochondrial vestigia (#) are also observed in the altered axoplasm of G93A vehicle mice. Graph (A) shows a significant reduction of mitochondrial density in vehicle-treated G93A mice compared with other groups. In addition, the percentage of altered mitochondria (B) and the mitochondrial diameter (C) are significantly increased in G93A vehicle mice. Both effects were prevented by lithium administration. Values are the mean ± S.E.M. Comparisons between groups were made by using One-way ANOVA. \**P* ≤ 0.05 compared with other groups. Scale bars: WT vehicle, WT lithium, G93A vehicle = 0.7 µm, G93A lithium = 0.6 µm, insert = 0.2 µm.

of cluster density in homogeneous areas of squares measuring a side of 100 nm were similar. At this point, clusters density falls according to a geometrical distribution which confirms the model drawn by Salpeter (Salpeter et al., 1978). This represents the first in situ evidence for lithium induced mitochondriogenesis. The effects of lithium were evaluated on various patterns of radioactive grains. Lithium increases the number of grain clusters (**Figure 13A**); lithium increases the mean diameter of clusters (**Figure 13B**); lithium increases the polarization of cytoplamic vs. nuclear clusters (**Figure 13C**); lithium increases the number of clusters within and immediately around (250 nm) mitochondria (**Figure 13D**); lithium increases the ratio between mitochondrial and scattered cytosolic clusters (**Figure 13E**). This is in line with the curve we measured in **Figure 1** for [2-3H]adenine distribution in the cytosol showing that, the distance of 250 nm corresponds to the critical length for which the highest difference is observed between lithium and vehicle-treated cells. At this distance, in lithium treated cells the number of grain clusters is still the highest and corresponds to the number of grain clusters counted on the mitochondrial surface. This is clearly visible in **Figure 1** where the Y-values for radioactive clusters counted at X = 0 (corresponding to the mitochondrial surface) was constant in a range of X = 0–250 nm. This curve of cluster distribution produced by lithium was abolished in vehicle-treated cells This indicates significantly the occurrence of mitochondriogenesis according to the model of Salpeter et al. (1978). As described in the method the counts ruled out the staining produced by nuclear DNA.

In order to further confirm the role of autophagy in motor neuron survival and the autophagy-dependency of most part of lithium-induced neuroprotection we analyzed motor neuron survival in a very controlled experiment where primary motor neurons from the spinal cord were selectively exposed to an autophagy inhibitor. As shown in **Figure 14**, we administered the autophagy inhibitor 3-MA. When the highest dose of 3- MA was administered, SMI32 positive motor neurons were lost dramatically. Thus, in this in vitro model of ALS we could document autophagy-dependent motor neuron loss which

was rescued by lithium administration (**Figure 14**). In these experimental conditions the subcellular changes replicated those described in vivo as well as the ultrastructural effects produced by lithium. Interestingly, as shown in representative pictures and graph of **Figure 15**, when we evaluated the immune-electronmicroscopy of the protein SOD-1 we documented the occurrence of a lithium-dependent clearance of the SOD-1 protein. This effect confirms the ability of lithium to exert a direct protective effect at mitochondrial level by removing toxic mutated G93A SOD-1. Remarkably, while lithium-induced removal of G93A SOD-1 was significant, no noticeable effect was documented for WT SOD-1 (**Figure 15**). Additional effects were produced by lithium on glial cells as reported in the Supplementary Material (Figure S1) where lithium which is known to inhibit glial cells activation still re-shapes mitochondrial size. This effect does not relate to mitochondrial swelling which we could not document in glia at significant level.

### DISCUSSION

In the present study we compared different motor neuron compartments in G93A SOD-1 mice. In each compartment we analyzed the mitochondrial pathology as well as extra-mitochondrial alterations. These studies were carried out both in baseline conditions and after autophagy induction in ALS mice compared with WT. Autophagy was induced by lithium administration which promoted effective progression of the autophagy pathway as documented in a variety of experimental studies as well as in the present work and in the Autophagy Guidelines by Klionsky et al. (2012). The analysis of motor neuron compartments was detailed at the level of dendrites, cell bodies and axons. The axonal compartment was further divided in proximal vs. distal axons. In this way we wish to evaluate whether the status of motor axons upstream in the spinal cord really reflects the ultrastructure of peripheral axons distributed in the muscle which are known to be early and severely damaged in ALS. In fact, when we measured mitochondrial alterations as well as other ultrastructural targets of ALS-related pathology we documented remarkable differences in distal compared with proximal axons. These differences are compatible with the disruption of axonal transport of mitochondria which was recently described by Magrané et al. (2012, 2014). In keeping with this, we documented here for the first time, the presence of axonal jams which clog significantly the axonal lumen suggesting a morphological correlate for altered mitochondrial transport along axons (Magrané et al., 2012, 2014). These axonal jams were

hardly deciphered in their structure but they appear to contain mitochondrial remnants as well as proteinaceous aggregates. Significantly, when G93A mice were administered lithium, this compound cleared the axonal lumen and produced a significant improvement of mitochondrial number and morphology which was much more evident in the distal compared with proximal segment of the axons. These data confirm the seminal role of an impairment of autophagy in the progression of ALS. The occurrence of defective autophagy in ALS was demonstrated for the first time in the last decade (Fornai et al., 2008a,b). This was progressively validated through a growing number of studies (Pasquali et al., 2009; Ferrucci et al., 2010; Otomo et al., 2012; Shimada et al., 2012; Wang et al., 2012; Castillo et al., 2013; Ikenaka et al., 2013; Barmada et al., 2014; Cheng et al., 2015; Lee et al., 2015; Philips and Rothstein, 2015; Xiao et al., 2015; Yang et al., 2015) up to the very recent confirmative work by Xie et al. (2015a,b). This evidence was also reviewed in this issue by Ruffoli et al. (2015). In order to evaluate directly the significance of impaired autophagy for motor neuron survival, in the present study we administered the autophagy blocker 3-MA, which produced a dose dependent motor neuron loss, which in turn, was rescued by lithium. In the present study we extended the structural and morphometrical analysis to the gastrocnemius muscle which possesses dramatic light and electron microscopy alterations in G93A mice which were reverted by lithium. While the effects of lithium on motor neuron cell body were already reported, the occurrence of a remarkable effect in the distal vs. the proximal axons as well as in muscles was reported here for the first time in ALS mice. Interestingly, lithium is known to protect against a variety of peripheral neurophathy as shown in recent and classic literature. In fact, lithium prevents peripheral neuropathy induced by paclitaxel (Mo et al., 2012; Pourmohammadi et al., 2012) as well as the neuropathy induced by vincristine (Alimoradi et al., 2012) and there is evidence showing that lithium may even reverse the neurological damage induced by vinca alkaloids (Petrini et al., 1999). Interestingly, these compounds are known to inhibit vesicle trafficking, thus impairing autophagy progression and axonal transport. Again, in keeping with the dramatic morphological effects of lithium on peripheral nerves described here, previous studies indicate that lithium improves peripheral nerve regeneration after injury (Nouri et al., 2009) and it promotes axonal re-growth after ventral root avulsion (Fu et al., 2014). Similarly to peripheral nerves, muscles are known to be significantly affected by lithium which protects from a variety of degenerative changes such those occurring after muscle accumulation of amyloid beta and tau protein and muscle atrophy due to peripheral denervation (Askanas and Engel, 2008; Du et al., 2008; Kitazawa et al., 2008; Terracciano et al., 2010; Askanas et al., 2012; Fu et al., 2014; Su et al., 2014). Additionally lithium promotes myogenic differentiation and muscle regeneration (Polesskaya et al., 2003; van der Velden et al., 2007). The amount of muscle protection

observed here along with a remarkable preservation of the sarcomeric architecture is similar to what demonstrated in these previous studies carried out administering lithium to muscles in pathological and baseline conditions. This study provides the first evidence describing a remarkable protection of muscle degeneration induced by lithium in the course of motor neuron disease. The protective effects induced by lithium were analyzed here by light end electron microscopy and they were measured by using morphometry of multiple targets within muscle cells. In fact, following lithium administration the dramatic loss of alignment of muscle fibers was prevented. Again, the abnormal inter-sarcomeric area was reduced back to WT values and the architecture of each sarcomer, which was dramatically altered, was largely reinstated. Similarly, the abnormal swelling of the sarcoplasmatic reticulum within muscle fibers from G93A mice was no longer present when mice were administered lithium. These effects, which are deeply grounded on the autophagy stimulating effects of lithium may also depend on a frankly primary protection by lithium of mitochondrial alterations such as altered calcium homeostasis. In fact lithium was shown to mitigate cytosolic calcium elevation (Bosche et al., 2013). Similarly, as shown by Scheibye-Knudsen et al. (2012), in the Cockayne syndrome type B mitochondrial protection is produced when gold standard autophagy activators such as lithium and rapamycin are administered. In the present study we did not assess directly potential neuroprotective effects exerted by lithium on mitochondrial activity; nonetheless we evaluated whether lithium was able to remove mutated SOD-1 from mitochondria, which in turn, is considered one potential mechanism of mitochondrial toxicity in ALS due to a G93A SOD-1 mutation. Other protective mechanisms exerted by lithium on mitochondria as a result of lithium-induced modulation of mitochondrial bioenergetics cannot be ruled out as convergent mechanisms to produce the overall beneficial results described here. Since the G93A mouse model is reported to suffer from an excess of mitochondrial accumulation of the mutant enzyme G93A SOD-1 within mitochondria, we evaluated the removal of this enzyme in lithium administered G93A mice. In fact, we documented in situ by representative pictures and morphometrical counts the removal of mutant but not WT SOD-1 in lithium administered G93A mice. This confirms at mitochondrial level the clearance of SOD-1 produced by lithium we already reported using western blotting (Fornai et al., 2008a). The present study provides significant evidence of a compartment-dependent damage to mitochondria and other sub-cellular structures in ALS, while demonstrating the compartment-specific neuroprotective effects of lithium. The removal of mutant SOD-1 from the

mitochondria of lithium administered G93A mice is expected to lead indirectly to a preservation of mitochondrial architecture. This effect which was shown in the cell body of motor neurons (**Figure 15**), is still effective in the proximal axons, where it may contribute to explain why, in the absence of an effective mitochondrial turnover, the few mitochondria preserve their structure.

When analyzing at TEM the spinal cord of G93A mice, mitochondrial swelling and mitochondria-containing large stagnant vacuoles were measured within motor neuron perikarya and dendrites of lamina IX. These vacuoles were identified as authentic autophagy vacuoles by the gold standard (electron microscopy for multi-membrane vesicles staining both for LC3 and beclin I). Giant swollen mitochondria were characterized by severe dilution of the matrix and cristae breaking. The increase in mitochondrial size was dramatic and it was accompanied by an increase in the amount of altered mitochondria. This was associated with a dramatic filling with mitochondria-containing stagnant autophagy vacuoles within cell bodies of motor neurons. These effects confirm previous findings indicating a defective autophagy flux in the spinal cord both in ALS patients (Sasaki, 2011) and experimental ALS models (Fornai et al., 2008a; Li et al., 2008; Xie et al., 2015a,b). In these experimental conditions we analyzed the effects of a gold-standard autophagy inducer, which is lithium (Scheibye-Knudsen et al., 2012). Lithium is known for a long time as a mood stabilizer and neuroprotective agent (Madeo et al., 2009; Luo, 2010; Shimada et al., 2012; Chiu et al., 2013; Agam and Israelson, 2014; Lieu et al., 2014; Mohammadianinejad et al., 2014; Motoi et al., 2014; Yáñez et al., 2014; Dwivedi and Zhang, 2015) and it represents a classic autophagy inducer which was validated along several studies (Sarkar et al., 2005; Sarkar and Rubinsztein, 2006; Fornai et al., 2008a; Pasquali et al., 2010; Motoi et al., 2014) including the guidelines on autophagy (Klionsky et al., 2012), which represent the reference point for studies on the autophagy pathway. We found that lithium administration reverses/prevents almost completely alterations of mitochondrial architecture, the occurrence of swollen mitochondria and the presence of big stagnant autophagy vacuoles. The effects of lithium on mitochondrial size, shape, number and architecture are already evident in representative semi-thin sections and they were quantified by ultrastructural morphometry. Interestingly, the mitochondrial re-shaping effects produced by lithium were not limited to diseased motor neurons of G93A mice since we could document a significant mitochondrial plasticity induced by lithium even in motor neurons from WT mice.

Interestingly, while in vehicle-administered G93A mice we never documented the occurrence of small electron-dense mitochondria, within motor neurons of G93A mice administered lithium the size of mitochondria was markedly reduced and these organelles possessed an electron-dense ultrastructure, which generally corresponds to newly synthesized organelles. In G93A mice treated with lithium giant mitochondria were suppressed and the fine mitochondrial ultrastructure was preserved with slighter matrix dilution and rare cristae fragmentation being similar to controls. In fact, when counting mitochondria in both G93A and WT mice treated with lithium we documented the reduction in mitochondrial diameter, the increase in mitochondrial electron-density, and the increase in the total number of mitochondria. These small electrondense mitochondria were barely detectable within autophagy vacuoles. Thus, following lithium mitochondria within cell bodies of motor neurons increased their number, but they also preserved a fine architecture and appear as electron dense wellstructured small organelles. These effects are reminiscent of what happens following mitochondriogenesis. In fact, the stimulation of the biogenesis of mitochondria as well as mitochondrial fission increases the number of these organelles. However, when mitochondriogenesis is increased, the new mitochondria are very small electron-dense and well-conformed organelles, whereas the sole fission produces an increased number of mitochondria which otherwise appear of regular size and conformation. These data lend substance to what previously documented (Fornai et al., 2008a,b) suggesting that lithium produces mitochondriogenesis. In previous studies we provided such an evidence by measuring quantitative rtPCR for mitochondrial genes or MitoTracker Red

and Green (Fornai et al., 2008a). This is in line with very recent data published by Palikaras et al. (2015a,b) showing the molecular mechanisms which eventually lead to a concomitant activation of mitophagy and mitochondriogenesis. In detail, when mitophagy is induced by autophagy activators, a molecular cascade is triggered which leads to a dual effect inducing concomitantly the biogenesis of novel mitochondria (Palikaras et al., 2015a,b).

In order to detail the occurrence of mitochondriogenesis in situ and to investigate these effects in other cell types, we measured the effects of lithium chloride in PC12 cells with or without pre-administration of the nucleotide precursor [2-3H]adenine in vitro. In this way, despite confirming a strong mitochondrial plasticity induced by lithium in baseline conditions, we could document by in situ morphometry that lithium produces a strong biogenesis of mitochondria as measured by incorporation of [2-3H]adenine. In detail, the administration of lithium to otherwise normal PC12 cells significantly modifies mitochondria, which is reminiscent of what observed ex vivo within motor neurons of WT mice treated with lithium. In fact, lithium administration further improved the architecture of mitochondria in PC12 cells. In particular, in vehicle administered PC12 cells, regularly shaped mitochondria with a slight degree of matrix dilution were observed. When PC12 cells were exposed to lithium mitochondrial morphology was well-defined with increased electron-density of the matrix and well-defined and densely packed cristae. Moreover, as described for ex vivo motor neurons' cell bodies, the number of mitochondria was dramatically increased following lithium exposure. This effect also extended to mitochondrial size, which was decreased by lithium administration. As previously mentioned, the occurrence at a high rate of small, well-structured, mitochondria suggests an ongoing mitochondriogenesis. Therefore, we measured directly in situ, by using TEM combined with [2-3H]adenine exposure, the occurrence of mitochondrial DNA replication. In detail, control and lithium-exposed PC12 cells were examined by high resolution autoradiography, following administration of [2- <sup>3</sup>H]adenine. Under these experimental conditions, radioactive labeling appears associated both with nucleus and cytoplasm.

However, following lithium administration a significant increase in the number of grains was detected in the cytoplasm. [2- <sup>3</sup>H]Adenine was clustered specifically in very small areas surrounding mitochondria (which is compatible with grains diffusion constant established here: within mitochondria up to 250 nm distance from mitochondrial contour). In keeping with this, when lithium was administered, the distance between [2-3H]adenine labeling silver grain clusters and mitochondria was polarized within a range of 250 nm compared with stochastic diffusion which was measured in vehicle-treated cells (see **Figure 1** again). Moreover, the mean size of electron microscopy clusters surrounding mitochondria was way higher in lithium compared with vehicle-treated cells. These data provide for the first time an in situ mechanistic evidence for mitochondrial biogenesis, thus lending substance to previous results we obtained by using quantitative rtPCR or MitoTracker Red or Green (Fornai et al., 2008a). This confirms that autophagy inducers which work as mitophagy activators produce a concomitant stimulation of mitochondriogenesis (Palikaras et al., 2015a,b).

In these latter works the mitophagy flux was coupled with the biogenesis of mitochondria. In fact, although autophagy induction was already suggested to be tightened with mitochondriogenesis (Fornai et al., 2008a; Ruffoli et al., 2015), very recently, elegant data indicated how mitophagy is coupled with biogenesis of novel mitochondria (Palikaras et al., 2015a,b). When damaged mitochondria occur even in physiological conditions, this triggers a dual pathway based on SKN-1 activation, which mediates mitophagy (the removal of damaged mitochondria), and increases mitochondrial biogenesis (Palikaras et al., 2015a,b). Interestingly, these dual effects were recently demonstrated for another autophagy inducer such as resveratrol (Meira-Martins et al., 2015), which is known to protect in ALS models (Han et al., 2012).

The significance of the present findings on motor neuron perikarya needs to take into account what recently published by Parone et al. (2013), who found that protecting mitochondria within motor neuron cell body and axons in the ventral root (through knocking out CycD thereby inducing the stabilization of the mitochondrial transition pore), does not extend the

survival of G93A mice, despite preserving the number of motor neurons counted within spinal cord. The occurrence of palsy and lethality in these experimental conditions is considered to be the consequence of the axonal degeneration which still persists producing muscle denervation despite mitochondrial protection (Parone et al., 2013).

If one assumes that even axonal damage is produced by mitochondrial dysfunction, then it is likely that the threshold for mitochondrial damage in the cell body is higher than axons. This would explain why treatments leading to a noticeable protection of mitochondria in the motor neuron cell body do not guarantee for the survival of the peripheral axons. This is in line with mice undergoing palsy and death despite a fair sparing of motor neurons counted in the anterior horn of the spinal cord (Parone et al., 2013). Within this scenario we extended our analysis considering the effects of lithium within motor neuron axons. Here we noticed a substantial difference of mitochondrial damage in the proximal compared with distal axons and with what observed in the cell body. In the study by Parone et al. (2013), the occurrence of lethality and neurological symptoms was not modified by preserving mitochondria in the cell body and proximal axons ruling out the hegemonic role for mitochondrial alterations in the genesis of ALS. However, in this study axonal mitochondria were analyzed only at the level of the spinal cord/ventral root which means at the level of proximal axons. In the present study a clear difference was observed in mitochondrial degeneration and protection when considering proximal compared with distal axons. In particular, the amount of damage at mitochondrial level and within extra-mitochondrial structures in distal axons exceeds the damage which can be measured in proximal axons and cell bodies and dendrites. This indicates as mandatory to analyze the ultrastructure of peripheral axons in order to judge about the occurrence of axonal loss and muscle denervation, while proximal axons do not correlate with disease severity. In keeping with this, recent studies documented remarkable beneficial effects of lithium to enhance motor regeneration at axonal level (Fu et al., 2014; Su et al., 2014). This is in line with data reported here showing that lithium produces beneficial effects in peripheral axons and muscles. These effects are much slighter in proximal axons. The diversity of these axonal compartments may relate to the occurrence of axonal jams which clogs the axonal lumen which is cleared by lithium administration. The occurrence of axonal clogging is likely to make the distal axonal compartment no longer reachable by the cell body, unless a clearance of the axonal lumen is produced. In fact, the clearance induced by lithium of axonal lumen matches the increase in mitochondrial density and healthy mitochondria which was produced by lithium in distal axons. In keeping with the diversity of effects within proximal vs. distal axons some issues appear as contradictory. It may be difficult to explain why the amount of mitochondria counted in the proximal axon of lithium-treated WT is similar to lithium- and vehicleadministered G93A mice. In fact assuming a detrimental role of axonal jamming to impede mitochondrial axonal transport, one should expect that the number of mitochondria in lithiumtreated WT was similar to vehicle-administered WT. Again G93A mice administered lithium should posses a number of mitochondria similar to that counted in G93A mice administered vehicle. However this point, despite being unexpected at first glance, confirms the profound difference between proximal and distal axon as distinct compartments in motor neuron pathology. In fact, when looking at distal axons, we measured the highest amount of mitochondrial damage and the presence of a significant protection by lithium. This is in line with the fact that proximal axons in the spinal cord do not represent the culprit of ALS degeneration which is instead localized within distant peripheral axons within muscles.

This is substantiated by the placement of axonal jamming which was detected upstream to distal axons and it was removed by lithium administration. Still, one might wonder why a lack of mitochondria in the proximal axon still occurs in lithium-treated mice. This is likely not to be a critical point for axonal physiology since, as in the cell bodies, proximal axons are not the site of neurotransmitter release and maximal energy consumption. These findings strengthen the introductory hypothesis of the manuscript that proximal compared with distal axonal damage

#### REFERENCES


is key to understand ALS pathology. Similarly functional data indicating impairment of axonal transport of mitochondria (Magrané et al., 2012, 2014) are now grounded on a solid morphological evidence provided here by axonal jamming. These data do not contradict the data obtained by Parone et al. (2013).

In fact in their study these Authors documented the preservation of mitochondria in the cell body and axons at the level of the ventral root (which means proximal axons), while the peripheral axonal degeneration and muscle denervation was still ongoing. In this study there was no ultrastructural analysis of mitochondria within peripheral axons where degeneration and denervation were taking place. The present findings shift the focus toward the very distal axonal compartment compared with proximal axons as key to understand the key role of mitochondria in producing axonal degeneration, muscle denervation and the course of ALS. It is likely that other neuronal alterations beyond mitochondria play a fundamental role in ALS; nonetheless the experiments reported here demonstrate that mitochondrial protection within distal axons is beneficial. Of course additional targets need to be considered. In fact, axonal clogging which occurs within axonal lumen is likely to play a critical role. Again, a frank beneficial effect on muscle ultrastructure shown here may improve the disease course also independently from motor innervation which remains fundamental. The protective effects of lithium described here on distal axonal mitochondria, axonal clogging and muscle derangement may not be powerful enough to arrest the disease course although they offer a solid proof of principle on how to target the critical points in the motor system which determine the onset and progression of motor symptoms as well as survival in ALS.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2015.00434


organotypic slice cultures of spinal cord. Neuroscience 165, 1353–1369. doi: 10.1016/j.neuroscience.2009.11.034


and valproate. Neuroscience 301, 276–288. doi: 10.1016/j.neuroscience.2015. 06.002


amyotrophic lateral sclerosis, is antagonized by riluzole. Neurodegener. Dis. 13, 171–179. doi: 10.1159/000357281


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Natale, Lenzi, Lazzeri, Falleni, Biagioni, Ryskalin and Fornai. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Gly482Ser PGC-1α Gene Polymorphism and Exercise-Related Oxidative Stress in Amyotrophic Lateral Sclerosis Patients

Angelique Pasquinelli <sup>1</sup> , Lucia Chico<sup>1</sup> , Livia Pasquali <sup>1</sup> , Costanza Bisordi <sup>1</sup> , Annalisa Lo Gerfo<sup>1</sup> , Monica Fabbrini <sup>1</sup> , Lucia Petrozzi <sup>1</sup> , Letizia Marconi <sup>2</sup> , Elena Caldarazzo Ienco<sup>1</sup> , Michelangelo Mancuso<sup>1</sup> and Gabriele Siciliano<sup>1</sup> \*

<sup>1</sup> Departments of Clinical and Experimental Medicine, Neurological Clinic, University of Pisa, Pisa, Italy, <sup>2</sup> Departments of Surgical, Medical and Molecular Pathology, and Critical Area, University of Pisa, Pisa, Italy

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Francesco Panza, University of Bari Aldo Moro, Italy Aaron Paul Russell, Deakin University, Australia Andrew John Whittle, Stanford University, USA Adam Keith Walker, Macquarie University, Australia

> \*Correspondence: Gabriele Siciliano

gabriele.siciliano@med.unipi.it

Received: 25 January 2016 Accepted: 05 April 2016 Published: 22 April 2016

#### Citation:

Pasquinelli A, Chico L, Pasquali L, Bisordi C, Lo Gerfo A, Fabbrini M, Petrozzi L, Marconi L, Caldarazzo Ienco E, Mancuso M and Siciliano G (2016) Gly482Ser PGC-1α Gene Polymorphism and Exercise-Related Oxidative Stress in Amyotrophic Lateral Sclerosis Patients. Front. Cell. Neurosci. 10:102. doi: 10.3389/fncel.2016.00102 The role of exercise in Amyotrophic lateral sclerosis (ALS) pathogenesis is controversial and unclear. Exercise induces a pleiotropic adaptive response in skeletal muscle, largely through the peroxisome proliferator-activated receptor γ coactivator 1α (PGC-1α), a transcriptional coactivator that regulates mitochondrial biogenesis and antioxidant defense mechanisms. It has been suggested that a Gly482Ser substitution in PGC-1α has functional relevance in human disorders and in athletic performance. To test this hypothesis, we examined the genotype distribution of PGC-1α Gly482Ser (1444 G > A) in ALS patients to evaluate whether or not the minor serine-encoding allele 482Ser is involved in oxidative stress responses during physical exercise. We genotyped 197 sporadic ALS patients and 197 healthy controls in order to detect differences in allelic frequencies and genotype distribution between the two groups. A total of 74 ALS patients and 65 controls were then comparatively assessed for plasmatic levels of the oxidative stress biomarkers, advanced oxidation protein products, ferric reducing ability and thiol groups. In addition a subgroup of 35 ALS patients were also assessed for total SOD and catalase plasmatic activity. Finally in 28 ALS patients we evaluated the plasmatic curve of the oxidative stress biomarkers and lactate during an incremental exercise test. No significant differences were observed in the genotype distribution and allelic frequency in ALS patients compared to the controls. We found significant increased advanced oxidation protein products (p < 0.001) and significant decreased ferric reducing ability (p < 0.001) and thiol groups (p < 0.001) in ALS patients compared to controls. When comparing different genotypes of PGC-1α, no relation between Gly482Ser polymorphism and oxidative stress biomarker levels was detected in resting conditions. On the other hand, when considering exercise performance, lactate levels were significantly higher

**Abbreviations:** ALS-FRS-R, Amyotrophic lateral sclerosis functional rating scale; AOPP, advanced oxidation protein products; ALS, amyotrophic lateral sclerosis; DTNB, dithiobis nitrobenzoic acid; fALS, fALS amyotrophic lateral sclerosis; FRAP, ferric reducing antioxidant power; MVC, maximum voluntary contraction; MEF2, myocyte enhancer factor 2; PGC-1α, peroxisome proliferator activated receptor gamma coactivator 1 alpha; PBS, phosphate buffered saline; ROS, reactive oxygen species; RFLP, restriction fragment length polymorphism; SNPs, single nucleotide polymorphisms; SOD, superoxide dismutase; SOD1, superoxide dismutase-1.

(between p < 0.01 and p < 0.001) and greater protein oxidative products were found in AA (Ser482Ser) compared to GG (Gly482Gly) and GA (Gly482Ser) ALS patients. Our findings highlight the importance and confirm the involvement of oxidative stress in ALS pathogenesis. Although not associated with 1444 G > A SNP, ALS patients with Gly482Ser allelic variant show increased exercise-related oxidative stress. This thus highlights the possible role of this antioxidant defense transcriptional coactivator in ALS.

Keywords: amyotrophic lateral sclerosis, exercise, oxidative stress, PGC-1α, polymorphisms

#### INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is the most common neurodegenerative disorder of motor neurons. Loss of pyramidal, spinal, and bulbar motor neurons affecting multiple regions of the body leads to progressive motor dysfunction, disability and death. It is clinically characterized by progressive paralysis and eventual death from respiratory failure within 2–5 years (Wijesekera and Leigh, 2009; D'Amico et al., 2013).

ALS is mainly a sporadic disease (sALS) however approximately 10% of ALS cases are familial (fALS; Ticozzi et al., 2011), due to mutations in several genes, which can play an important role in understanding the pathogenesis of fALS and sporadic ALS (Ingre et al., 2015). The distribution of fALS and sporadic cases also depends on the ancestral origin of the population, since the frequency of ALS-associated genes, as well as environmental factors vary significantly across different populations (Marin et al., 2015).

ALS is a complex and multifactorial disease characterized by the involvement of several pathogenic conditions, including glutamatergic excitotoxicity (Matsumoto et al., 2007), mitochondrial dysfunction (Shi et al., 2010), defects in RNA maturation (Matsumoto et al., 2007) and oxidative stress (Robberecht, 2000). Oxidative stress biomarkers have been investigated in several studies and show high protein carbonyl levels, increased lipid peroxidation and DNA/RNA oxidative modifications, in the nervous and peripheral tissues in both sporadic and fALS ALS patients (Chang et al., 2008; Barber and Shaw, 2009; Kabuta et al., 2015).

ALS occurs in all socio-professional categories with a higher prevalence in athletes and in individuals who practice intense and competitive activities. In fact the association of strenuous physical activity with genetic predisposition and external factors, such as chemicals used in the sanitizing sports grounds, or head injuries suffered by players during games, can increase the risk of developing ALS in some individuals (Scarmeas et al., 2002; Chiò et al., 2005). Conversely moderate activity has been shown to be protective for ALS in mice (Carreras et al., 2010).

Acute exercise is known to induce a transient increase in reactive oxygen species (ROS), as shown by several reports of increased oxidative damage following acute bouts of erobic and anerobic exercise (Fisher-Wellman and Bloomer, 2009). Although the results are somewhat mixed and appear disease dependent, individuals with chronic disease experience an exacerbation in oxidative stress following acute exercise compared to healthy individuals (Fisher-Wellman and Bloomer, 2009). However the relationship between physical activity and oxidative stress is far from linear (Pittaluga et al., 2006), since physical activity is associated with favorable changes in blood pressure, and lipid and lipoprotein profiles, which reduce the risks of several oxidative stress associated diseases. In addition, recent research shows that physical activity contributes significantly to a longer life in good health (Fraile-Bermúdez et al., 2015).

It is still unclear which of these mechanisms are the primary causes or consequences of ALS pathogenesis. Several proteins have been identified that may be involved in the mechanisms of ALS pathogenesis, including the Peroxisome proliferator activated receptor gamma coactivator 1 Alpha (PGC-1α; Qi et al., 2015). PGC-1α is a tissue-specific transcriptional coactivator, encoded by the gene PPARGC1A/PGC-1α, located on chromosome 4 (4p15.1), consisting of 13 exons (Arany, 2008).

PGC-1α has an N-terminal domain, containing the sites for the activator proteins, a central region containing the sites for the transcription factors which interact with PGC-1α, and a C-terminal domain containing a binding site for RNA (Monsalve et al., 2000). PGC-1α activates different metabolic pathways in response to nutritional and environmental stimuli such as fasting, cold and physical activity (Handschin et al., 2007a). In physiological conditions, PGC-1α is located in the cytoplasm in an inactive state. During muscle contraction or exercise, it is activated and moves into the nucleus where it promotes the transcription of genes involved in mitochondrial biogenesis, oxidative phosphorylation, skeletal muscle fiber composition remodeling and antioxidant defense (Puigserver and Spiegelman, 2003; Valle et al., 2005; Liang and Ward, 2006).

Studies in mice models of ALS (SOD1 G93A and SOD1 G37R mouse) and Duchenne muscular dystrophy (mdx mouse; Sheng et al., 2012) have shown that increased expression and activity of PGC-1α can improve the clinical manifestations (Handschin et al., 2007b; Liang et al., 2011) by increasing mitochondrial biogenesis and ROS detoxification (Austin and St-Pierre, 2012). Da Cruz et al. (2012) have shown that increased PGC-1α activity in G37R improves muscular endurance, reduces atrophy and improves locomotor activity also in the advanced stages of disease. Liang et al. (2011) showed that PGC-1α expression declines with ALS development in SOD1 G93A mice and that PGC-1α overexpression slows the disease progression also moderately extending the lifespan, without any effect on disease onset. The motor activity was also improved, with a significant decrease in motor neuron cell death and neuromuscular junction damage, thus suggesting that increased PGC-1α expression may be important for ameliorating ALS progression (Liang et al., 2011).

The gene PGC-1α exists in several allelic variants that differ from the common allele in terms of a single nucleotide (Steinbacher et al., 2015). Among single nucleotide polymorphisms (SNPs), the G to A substitution at position 1444 in exon 8 of the PCG-1α gene, results in the substitution of glycine with serine in codon 482 and has been associated with a lower gene expression and reduced PCG-1α protein activity (Prior et al., 2012).

Several studies have suggested that with increased levels of ROS, PCG-1α can bind several transcription factors that induce the transcription of genes coding for antioxidant enzymes, which in turn limit ROS production and oxidative stress (St-Pierre et al., 2003, 2006; Valle et al., 2005; Liang and Ward, 2006). Other studies have investigated whether SNP 1444 G > A could affect the protein activity as a transcriptional coactivator, since the A carriers have low levels of PCG-1α proteins (Liang and Ward, 2006). In fact, this could inhibit the transcription of antioxidant genes and increase oxidative stress. However, there are conflicting results regarding oxidative stress biomarkers in A carriers and A non-carriers. Weng et al. (2010) found different levels of oxidative stress biomarkers between A carriers and A noncarriers, whereas Lai et al. (2008) did not find any such differences.

During exercise, muscle contraction causes PCG-1α protein activation, increases mitochondrial biogenesis, and oxidative capacity and leads to fast-to slow fiber type conversion (Rasbach et al., 2010). In these conditions the PCG-1α protein can bind a transcription factor myocyte enhancer factor 2 (MEF2). The PGC-1α/MEF2 complex then induces a remodeling of the fiber composition of skeletal muscle with greater involvement of oxidative fibers, which confer resistance to fatigue (Handschin et al., 2003). Zhang et al. (2007) showed that the amino acid change Gly482Ser causes an alteration in the binding site for MEF2, thus affecting the formation of the PGC-1α/MEF2 complex and inhibiting the switch from fast to slow muscle fibers with a greater involvement of glycolytic fibers. On the other hand overexpression of the PGC-1α protein increases the proportion of oxidative fibers (Lin et al., 2002), while knock-out models for PGC- 1α show a change from oxidative to glycolytic fibers (Handschin et al., 2007a). It has also been observed that increased levels of PGC-1α are associated with reduced levels of blood lactate (Wende et al., 2007).

The aim of this study was to evaluate whether or not the PCG-1α SNP 1444 G > A (Gly482Ser) could affect oxidative stress markers, both in basal conditions and during an exercise fatiguing test in ALS patients.

### MATERIALS AND METHODS

### Subjects

A total of 197 sporadic ALS patients (mean age ± standard deviation (SD): 62.7 ± 11.7 years old) and 197 healthy controls (mean age ± SD: 63.4 ± 9.7 years old) were enrolled in this study. Both groups were recruited at the Department of Clinical and Experimental Medicine, Neurological Clinic, University of Pisa (see **Table 1**).

Diagnosis of ALS was conducted according to the criteria of El Escorial, at the onset of the disease. ALS patients showed the ''upper limbs'' and/or ''lower limbs'' clinical form with mild or moderate functional disability, as assessed by the clinical scale ALS-FRS-R (The ALS Functional Rating Scale, 1996) and respiratory function was satisfactorily documented by an FVC >70% predicted. FALS ALS cases based on a positive family history of the disease were excluded. Control group included subjects without a medical history of neurological disorders (**Table 2**). All recruited subjects were of Caucasian origin. Both patients and controls gave their informed consent. The study protocol was approved by the Ethics Committee of the Great North West Area of Tuscany. The study was conducted according to Good Clinical Practice (GCP).

The association study between the polymorphism Gly482Ser (1444 G > A) in the PGC-1α gene and ALS disease, was investigated using PCR-RFLP method in 197 patients (82 F/115 M) and 197 healthy controls (99 F/98 M).

The cellular redox state was assessed on 74 sALS patients (37 F/37 M) and 65 healthy controls (30 F/35 M), by analyzing plasma levels of various oxidative stress biomarkers. In particular we evaluated the advanced oxidation protein products (AOPPs), as a marker of oxidative damage. We used the ferric reducing antioxidant power (FRAP) and total plasmatic thiol groups as markers of non-enzymatic antioxidants. In addition, in 35 patients (19 F/16 M) we evaluated the total activity of superoxide dismutase (SOD) and catalase enzymes, which are antioxidant enzymatic markers.

Plasmatic levels of oxidative stress markers (AOPP, FRAP, thiols and lactic acid) during exercise, were


sALS, sporadic ALS patients; SD, standard deviation.



<sup>∗</sup>Clinical data available for 136 sALS patients. ALS-frs, ALS-functional rating scale.

measured in 28 patients with sALS (15 F/13 M), undergoing an incremental exercise test on the forearm muscles.

### Exercise Protocols

The exercise test on the forearm muscles was conducted on sALS patients with a myometer connected to a ''hand-grip'' (Digital Multi-Myometer, MIE Medical Research Ltd., Leeds, UK). The exercise protocol consisted in steps in which the contractile force increased incrementally at 30%, 50%, 70% of the maximum voluntary contraction (MVC), calculated as the mean of three sequential contractions at the myometer, expressed in Newton. Each step includes one minute of intermittent contractions (about 1/s) and two minutes of rest. The venous blood samples for determination of oxidative stress markers were performed at baseline, at the end of each step and 15 min after the end of the exercise (**Scheme 1**). This type of exercise was mainly erobic at the beginning of the test and it became progressively anerobic, as the force increased, with rapid motor unit recruitment.

### Analytical Procedures

Peripheral venous blood samples were collected both for 1444 G > A (Gly482Ser) single nucleotide polymorphism (SNP) analysis and to monitor the oxidative stress.

#### SNP 1444 G > A Genotyping in PGC-1α Gene

1444 G > A SNP in PGC-1α gene was determined by restriction enzyme analysis using genomic DNA extracted from peripheral blood lymphocytes and the restriction fragment length polymorphism (RFLP) method.

Genomic DNA was isolated from whole blood samples with the QIAamp<sup>r</sup> Blood Mini Kit (Qiagen).

PCR reaction was performed in a total volume of 25 µl consisting of 1.5 pmol each primer: forward 5<sup>0</sup> -TGCTAC CTGAGAGAGACTTTG-3<sup>0</sup> and reverse 5<sup>0</sup> -CTTTCATCTTCGC TGTCATC-3<sup>0</sup> , 250 µmol/l dNTPs, Buffer (10X), with Mg2<sup>+</sup> (25mM), Fast Start Taq [2.5 U] (Roche s.p.a, Italy), and 10 ng of genomic DNA. Reactions were performed in a Perkin Elmer thermal cycler for one cycle at 94◦C for 6 min, 30 cycles at 94◦C for 30 s, 57◦C for 30 s, 72◦C for 30 s, and a final extension at 72◦C for 7 min.

After digestion with 3 U of MspI (HpaII) restriction enzymes, the amplified fragments were separated by 3.5% metaphor-agarose agarose gel electrophoresis, and the restriction patterns were examinated by ethidium bromide staining and UV light.

#### Evaluation of Oxidative Stress Biomarkers

Peripheral venous blood samples were collected to measure the oxidative stress parameters. Blood samples were immediately centrifuged at 3000 rpm for 10 min and plasma was stored at −80◦C until AOPP, FRAP, thiol levels and total SOD and catalase activity were determined. Briefly, AOPPs, a marker of oxidative damage to proteins, were determined by mixing plasma with

phosphate buffered saline (PBS), acetic acid and potassium iodide. The absorbance was read spectrophotometrically at 340 nm and compared with a solution of chloramine T dissolved in the same buffer. The data were expressed as nmol/ml of chloramine equivalents (Witko-Sarsat et al., 1996).

In order to measure non-enzymatic antioxidant properties, FRAP was assessed according to Benzie and Strain (1996): the FRAP reagent (sodium-acetate, tripyridyltriazine in hydrochloric acid, and ferric chloride) was pre-warmed to 37◦C and mixed with plasma; the absorbance was read after 4 min at 620 nm. A calibration curve was established by substituting the sample with a solution of iron sulfate in hydrochloric acid; the data were expressed in mmol/l.

The content of plasmatic total thiols, which are endogen antioxidants, was estimated by evaluating the sulphydryl groups (-SH) present in the molecules, following the protocol described by Hu (1994).

At the time of determination, 50 µl of plasma were added to 150 µl of Tris—EDTA, 10 µl of 2, 2-dithiobis nitrobenzoic acid (DTNB) and 800 µl of absolute methanol. After incubation at room temperature for 15 min the samples were centrifuged at 3000 g for 10 min. The absorbance of the supernatant was assessed at 412 nm and by subtracting the value of a blank consisting of DTNB. The values were expressed in µmol/l.

The total enzymatic activity of SOD and catalase was detected respectively by superoxide dismutase Assay Kit and catalase assay kit (Cayman, France), respectively, according to the manufacturer's instructions.

The SOD assay kit uses a tetrazolium salt to detect the superoxide radicals generated by xanthine oxidase and hypoxanthine. Briefly, 10 µl of plasma were mixed with 200 µl of the diluted radical detector, the reaction was initiated by adding 20 µl of diluted xanthine oxidase, and the absorbance was read, after 20 min, at 450 nm. The data were expressed as U/ml of SOD activity. One unit is defined as the amount of enzyme needed to exhibit 50% dismutation of the superoxide radical.

Regarding catalase activity, the kit uses the peroxidatic function of catalase to determine the enzyme activity. Twenty microlitres (20 µl) of plasma were mixed with 100 µl of diluited assay buffer, 30 µl of methanol and 20 µl of diluted hydrogen peroxide. After an incubation for 20 min at room temperature on a shaker, were added 30 µl of catalase purpald and 30 µl of potassium hydroxide. Followed an incubation for 10 min at room temperature and further addition of 10 µl of catalase potassium periodate.

After a final incubation for 5 min, the absorbance of each sample was read at 540 nm. The data were expressed as nmol/min/ml.

The assessment of lactic acid levels, carried out by Kit (Sentinel, Milan-Italy), was conducted in the laboratory of the Department of Clinical Chemistry of the Santa Chiara hospital in Pisa. The kit allows an enzymatic-colorimetric determination of lactate. Lactate is oxidized by lactate oxidase to pyruvate and hydrogen peroxide, which, in presence of peroxidase, reacts with N-ethyl-N-(2-hydroxy-3-sulfopropyl)-3 methylaniline forming a compound, whose color intensity is proportional to the concentration of lactate in the examined sample. Plasma (3 µl) were mixed with 300 µl of reagent1, that consists of reagent1a and reagent1b (provided with this kit). After an incubation at 37◦C for 10 min, the absorbance of each sample was read at 540 nm. The data were expressed as mg/dl.

#### Data Analysis

To verify whether the allele frequencies were in Hardy-Weinberg equilibrium, and to assess differences in the genotype and allele distributions between groups, we used the Chi-square (χ 2 ) analysis using Microsoft Excel. To evaluate any possible alteration in the redox balance in sALS patients and the control group, we used the unpaired two-tailed t testing; and data were expressed as mean ± SD. The multi-factor ANOVA test, using the Software StatGraphics, was used to correlate the oxidative stress data with polymorphism G1444A, considering each other biomarker studied as covariates. Only for the ANOVA test, were the values of the considered markers that did not have a normal distribution, prior to performing the assay, were normalized by transforming them into logarithmic data.

The ANOVA test for repeated measures, using the Software SPSS 20 was used to assess any differences in the redox balance during the exercise test. The values of lactic acid, for each exercise step, were normalized compared to the baseline.

A p value < 0.05 was considered significant.

### RESULTS

### Analysis of 1444 G > A Polymorphism in PGC-1α Gene

Results obtained from the analysis of the 1444 G > A polymorphism in the PGC-1α gene are shown in **Table 3**. The distributions of genotypes and allelic frequencies were in agreement with the Hardy-Weinberg equilibrium. We did not observe any difference in the genotype and allelic distribution between the sALS patients and healthy controls.

### Evaluation of Oxidative Stress Biomarkers

sALS patients had significantly higher AOPP plasma levels (397.91 ± 206.97 nmol/ml; p < 0.001) than the healthy controls (**Figure 1A**). Likewise, patients with sALS had a significantly lower plasma levels of FRAP (**Figure 1B**) and thiols (**Figure 1C**) than the controls.

### Correlation Between Oxidative Stress Biomarkers and 1444 G > A SNP in PGC-1α Gene

The study population (ALS patients and controls) was divided into three groups according to genotypes in GG, GA and AA. We did not observe any difference in the levels of the oxidative stress biomarkers AOPP (**Figure 2A**), lactate (**Figure 2B**), FRAP (**Figure 2C**), thiols (**Figure 2D**), SOD



<sup>∗</sup>Reference value for OR (odds ratio); IC, confidence interval; p value obtained by χ <sup>2</sup> analysis.

(**Figure 2E**) and catalase activity (**Figure 2F**) between the three groups. The same results were obtained by comparing each of these ALS subgroups with the control group (data not shown).

### Effects of Exercise on Oxidative Stress Biomarkers

Plasma levels of AOPP, FRAP and thiols remained stable during exercise until 15 min after physical activity (data not shown). Variations were observed in lactic acid levels during and after exercise [F(2.182, 34.911) = 15.5, p < 0.001]. In particular, in sALS patients we found increased lactate levels at 30% (p < 0.01), 50% (p < 0.01) and 70% (p < 0.001) of maximal voluntary contraction (MVC) compared to baseline levels and at 70% (p < 0.05) vs. 30% of MVC. At the recovery we observed a decrease in lactate at 50% (p < 0.05) and 70% (p < 0.001) of MVC although we did not observe any difference comparing the curves of ALS patients and healthy subjects (**Figure 3**).

### Relationship Between the Variations in Oxidative Stress Biomarkers Levels During Exercise and G1444A SNP in PGC-1α Gene

No changes were observed in the levels of FRAP and thiols in the three groups of patients divided according to genotypes in GG, GA and AA (data not shown). However, changes were observed in AOPP and lactic acid levels. In fact, subjects harboring the 482Ser allele (AA) showed increased AOPP levels during and after exercise than the other two groups (GG and GA genotype) which had stackable levels. This variation was significant at 50% of MVC (p < 0.05, GG vs. AA; p = 0.05, GA vs. AA) and at recovery (p = 0.01, GG vs. AA; p < 0.01, GA vs. AA; **Figure 4A**).

Finally, higher levels of lactic acid were observed during exercise in patients harboring AA genotype, compared to the other two groups. In particular, subjects with AA genotype show significantly increased levels of lactic acid at 30% (p < 0.01) and 50% (p < 0.001) of the MVC, compared to subjects with the GG genotype (**Figure 4B**).

## DISCUSSION AND CONCLUSIONS

In vitro studies support the hypothesis that PGC-1α plays a crucial role in protecting the mitochondria from oxidative stress through the reduced accumulation of ROS and reduced apoptotic cell death (Valle et al., 2005; St-Pierre et al., 2006). These data prompted us to evaluate the role of 1444 G > A polymorphism in response to oxidative stress, in basal conditions and during physical activity in sALS patients. ALS is in fact a motor neuronal disorder of a degenerative origin where oxidative stress seems to be one of the pathogenic players and muscle exercise appears to be an important phenotype modulator. We believe that this is the first study investigating PGC-1α SNPs in ALS patients. No differences were found between patients and controls as far as the distribution of 482Ser allele

is concerned, suggesting that the PGC-1α Gly482Ser variant is not associated with the disease. This in line with previous data on mice models showing no direct effect of PGC-1α protein expression on ALS onset (Liang et al., 2011; Da Cruz et al., 2012).

To assess the oxidative stress conditions in sALS patients, we have measured plasma levels of the AOPPs, as a marker of oxidative damage to proteins, and the FRAP and total thiol groups, as non-enzymatic antioxidants markers. We also investigated whether Gly482Ser polymorphism affects the levels of the analyzed oxidative stress biomarkers. Patients had significantly higher AOPP levels than the controls, thus confirming the presence of oxidative damage in ALS. These results are in agreement with previous studies indicating that oxidative stress is implicated in this disease (Bogdanov et al., 2000; Simpson et al., 2004; Siciliano et al., 2007; Kabuta et al., 2015). Bogdanov et al. (2000) detected increased levels of 8-OHdG in biological fluids of ALS patients compared to controls. Studies conducted by Simpson et al. (2004) and later confirmed by Kabuta et al. (2015) highlight increased levels of 4-Hydroxynonenal, which is a product of lipid peroxidation, in the spinal cord, motor cortex, cerebrospinal fluid, plasma/serum and urine of sALS patients compared to patients with other neurodegenerative diseases, controls with diseases of non degenerative origin or healthy controls.

Our results also show a reduction in antioxidant defenses, FRAP and thiols, in sALS patients, again confirming previous studies showing decreased antioxidant reserves in ALS. Keizman et al. (2009) showed low serum levels of uric acid, which represents about 60% of the FRAP value, in ALS patients compared to healthy subjects. Other studies have shown decreased total thiols (Baillet et al., 2010) and GSH levels in sALS patients compared to controls (Weiduschat et al., 2014). Overall, our findings of increased AOPPs levels and decreased FRAP and thiol levels in sALS patients seem to reflect an altered redox balance, thus confirming the key role of oxidative stress in ALS pathogenesis.

Under rest conditions, on the other hand, plasma levels of oxidative stress markers (AOPP, lactate, FRAP, total thiol groups, SOD and catalase activity) between ALS patients with GG genotype and A carriers, showed no differences. Thus, the allelic variant A does not seem to exert a negative role in conditions of oxidative stress. It could be speculated that in the PGC-1α protein, the 482 amino acid location, whether or not Gly or Ser, is not a critical domain for interaction with antioxidant gene promoters. However, under rest conditions, the PGC-1α protein is in the cytoplasm in an inactive conformation state and it is activated only in response to stimuli such as fasting, cold, physical activity and high elevation of ROS levels (Park et al., 2014; we did not measure ROS levels in this study).

Physical activity plays a key role in breaking or rebalancing the cellular redox balance (Childs et al., 2001; Iorio, 2003; Waring et al., 2003), and augmenting ROS levels can increase the

oxidative damage (Bloomer et al., 2005). Based on this evidence, we investigated possible alterations in the redox balance in response to exercise in a subgroup of sALS patients performing erobic exercise. While lactic acid increased, as expected for this type of exercise, changes in plasma levels of AOPP, FRAP and thiols were not detected during or after muscle exercise in sALS patients. These markers remained roughly constant for up to 15 min after the end of the test. The absence of oxidative stress responses to exercise could nonetheless still be in line with what is known, in healthy subjects, regarding the time relapsing curve of the variations in oxidative stress biomarkers levels after anerobic or erobic exercise. This occurs several hours after the end of the exercise, such as 6, 9, 24 h after exercise, which is a latency considered as compatible with the induction of enzyme machinery behind the oxidative stress markers we considered (Lewin, 2010). For ethical reasons, to avoid excessive discomfort to the patient, we have decided to limit the duration of experimental session and to do not collect samples several hours after the end of the exercise.

PGC-1α regulates the expression of various genes coding for several enzymes involved in fatty acid oxidation. Oxidative phosphorylation and intensive training lead to an increase in PGC-1α levels, which in turn can increase the oxidative capacity in skeletal muscle. A few studies have been conducted to test the possible effect of PGC-1α on athletic performance (Eynon et al., 2010). Lucia et al. (2005) reported an association between exercise and Gly482Ser polymorphism (1444 G > A), the minor A allele being significantly less frequent in athletes than in controls. The 482Ser allele is associated with poor physical capacity, while the Gly482 allele appears to be a genetic factor that increases beneficial muscular endurance (Nishida et al., 2015).

In this work we subdivided sALS patients according to genotype, GG genotype, GA genotype and AA genotype, in order to assess whether the 482Ser allele affects oxidative stress biomarker levels before, during and after exercise. While FRAP and thiol levels did not differ between the three groups, patients harboring the AA genotype showed, during exercise, higher levels of AOPP compared to the other two groups, specifically at 50% of maximal voluntary contraction (MVC) and at recovery after exercise. In addition, subjects with the AA genotype showed significant greater exercise-related lactic acid levels at 30% (GG vs. AA) and at 50% (GG vs. AA) of MVC, compared to subjects harboring GG genotype. The patients with the GA genotype showed intermediate levels of lactic acid, however these data were not statistically significant.

### REFERENCES


In patients with the AA genotype, the increase in lactate production during the incremental exercise test, may induce an increase in free radicals which, in turn, could damage proteins; in fact we observed an increase in AOPP levels in A carrier subjects. Therefore, although there were no variations in basal oxidative stress biomarkers levels in the 482Ser carriers compared to the Gly482 patients, our results showed that the occurrence of 482Ser allele may contribute to increased oxidative damage during physical activity.

Mechanisms such as the impairment of the MEF2 binding site in PGC-1α, as a result of changes in Gly482Ser amino acid, impede the fast-to-slow fiber type conversion during the endurance response in Gly482Ser carriers (Steinbacher et al., 2015). This can be taken into account to provide a pathogenic substrate to our results. This would then determine a greater involvement of anerobic glycolytic fibers to exercise, with the consequent accumulation of lactic acid, this in turn being one of the responsible factors of increased hydroxyl radical and further oxidative damage. Studies on mRNA expression profiles in targeted tissues would be useful in confirming this. Considering the PGC-1α protective role against muscle atrophy (Da Cruz et al., 2012), we can assume that synthetic molecules able to induce PGC-1α expression, delivered systemically or applied specifically to the muscle, could be used as adjuvant treatments in ALS patients to improve mitochondrial function, reduce atrophy and improve daily physical activity, thus contributing to a better quality of life in this disease.

In conclusion, the results obtained in this work suggest that the Gly482Ser SNP in the PGC-1α gene is related to the lower resistance to exercise-related oxidative stress in ALS. Knowing the genotype of each patient might be helpful in order to design appropriate training protocols aimed at maintaining the functionality of the skeletal muscle for as long as possible.

### AUTHOR CONTRIBUTIONS

AP and LC wrote the manuscript and performed experiments. ALG and L Petrozzi designed the study. L Pasquali, MM and GS supervised the study. CB, MF, ECI and LM performed clinical evaluation of patients.

### ACKNOWLEDGMENTS

The authors acknowledge the financial support from the CariLucca Foundation.

Parkinson's disease. Neurochem. Res. 35, 1530–1537. doi: 10.1007/s11064-010- 0212-5


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Pasquinelli, Chico, Pasquali, Bisordi, Lo Gerfo, Fabbrini, Petrozzi, Marconi, Caldarazzo Ienco, Mancuso and Siciliano. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Impaired Autophagy and Defective Mitochondrial Function: Converging Paths on the Road to Motor Neuron Degeneration

#### Brittany M. Edens 1,2 , Nimrod Miller 1,2 and Yong-Chao Ma1,2 \*

<sup>1</sup> Departments of Pediatrics, Neurology, and Physiology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA, <sup>2</sup> Lurie Children's Hospital of Chicago, Chicago, IL, USA

Selective motor neuron degeneration is a hallmark of amyotrophic lateral sclerosis (ALS). Around 10% of all cases present as familial ALS (FALS), while sporadic ALS (SALS) accounts for the remaining 90%. Diverse genetic mutations leading to FALS have been identified, but the underlying causes of SALS remain largely unknown. Despite the heterogeneous and incompletely understood etiology, different types of ALS exhibit overlapping pathology and common phenotypes, including protein aggregation and mitochondrial deficiencies. Here, we review the current understanding of mechanisms leading to motor neuron degeneration in ALS as they pertain to disrupted cellular clearance pathways, ATP biogenesis, calcium buffering and mitochondrial dynamics. Through focusing on impaired autophagic and mitochondrial functions, we highlight how the convergence of diverse cellular processes and pathways contributes to common pathology in motor neuron degeneration.

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Paula I. Moreira, University of Coimbra, Portugal Yu Aaron Tang, University of Texas Southwestern Medical Center, USA

#### \*Correspondence:

Yong-Chao Ma ma@northwestern.edu

Received: 25 November 2015 Accepted: 08 February 2016 Published: 03 March 2016

#### Citation:

Edens BM, Miller N and Ma YC (2016) Impaired Autophagy and Defective Mitochondrial Function: Converging Paths on the Road to Motor Neuron Degeneration. Front. Cell. Neurosci. 10:44. doi: 10.3389/fncel.2016.00044 Keywords: autophagy, mitochondria, calcium homeostasis, protein aggregation, ATP biogenesis, oxidative stress, ALS, motor neuron disease

## INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative disease characterized by loss of large motor neurons in the brain and spinal cord, resulting in progressive voluntary muscle wasting and respiratory failure. Patient death typically ensues 3–5 years following symptom onset (Volonte et al., 2015). With a prevalence of four-to-six per 100,000 people affected worldwide, ALS is one of the most common neurodegenerative disorders (Tan et al., 2007). Its origins can be either familial or sporadic, of which familial forms account for a mere 10% whereas the remaining 90% are sporadic (Rotunno and Bosco, 2013). Although ALS was first described by Charcot as early as 1869 (Jay, 2000), it wasn't until more than a century later that the first casual mutation was identified in copper zinc superoxide dismutase 1 (SOD1; Rosen, 1993). Since then, a multitude of genes associated with ALS pathogenesis have been identified. Nonetheless, mechanisms underlying motor neuron-specific vulnerability in ALS remain largely unknown. At present the disease defies all treatment. Riluzole is the only FDA-approved drug for treating ALS, and it may prolong patient lifespan by mere months (Rowland and Shneider, 2001).

Despite the heterogeneous and multigenic nature of ALS, overlapping pathology and common phenotypes are observed in different forms of the disease. Protein aggregates found in ALS patients suggest that cellular clearance mechanisms, such as the autophagy-lysosome pathway, may be impaired in this disease (Blokhuis et al., 2013). Moreover, increased oxidative stress and compromised mitochondrial function are observed in ALS disease condition (Wang and Michaelis, 2010). Intriguingly, oxidative stress is a potent regulator of autophagy (Scherz-Shouval et al., 2007; Huang et al., 2011; Scherz-Shouval and Elazar, 2011), suggesting the potential for functional interactions between the lysosomal and mitochondrial pathways.

Herein, we will consider in depth the dysregulation of autophagy and mitochondrial pathways, as well as their interactions in the context of ALS pathogenesis. We will put a special emphasis on mitophagy, as it directly connects cellular clearance mechanisms with mitochondrial function. Then, we will review the wealth of information regarding mitochondrial dysfunction in ALS, with particular interest in data derived from various mouse models. Lastly, we will discuss the role of oxidative stress as a critical regulator linking these discrete processes, and consider the therapeutic implications for ALS.

### MECHANISMS OF CELLULAR CLEARANCE IN PHYSIOLOGICAL AND PATHOLOGICAL CONDITIONS

### Autophagy

Autophagy is a catabolic process by which cells degrade and recycle cellular constituents through lysosomes to balance sources of energy and building blocks in order to maintain cellular homeostasis and function (Ryter et al., 2013; Yang et al., 2013). The core autophagy machinery consists largely of autophagy-related (ATG) genes, of which ATG1–10, ATG12–16, and ATG18 are all required (Klionsky et al., 2003; Mizushima et al., 2011). Beginning with induction, autophagy is initiated by intracellular or extracellular stimuli such as nutrient deprivation or stress. The most upstream player in the induction process is ATG1, whose mammalian homologs are unc51-like kinase 1 and 2 (ULK1 and 2; Mizushima, 2010). ULK1 interacts with Atg13, FIP200 (focal adhesion kinase family interacting protein of 200 kD), and ATG101 to form an autophagyinitiating complex (Hara et al., 2008; Hosokawa et al., 2009).

A major regulatory event in autophagy induction is exerted by the initiation complex's interactions with the nutrient-sensing mTOR kinase, and the energy-sensing AMP-activated protein kinase (AMPK). AMPK, activated by a drop in cellular energy, phosphorylates ULK1 on Serine 317 and Serine 777 (Hawley et al., 1996; Egan et al., 2011). These phosphorylation events in turn activate ULK1, which initiates autophagy (Egan et al., 2011). Conversely, the presentation of nutrients activates mTORC1 (through amino acid binding), which phosphorylates ULK1 on Serine 757, leading to the inhibition of autophagy (Kim et al., 2011). Therefore, autophagy initiation is kept in check by both nutrient- and energy-sensing mechanisms.

Following induction, the autophagosome forms and sequesters substrates for degradation. Autophagosome formation is controlled by ATG5 and ATG12/ATG16, which conjugate to recruit LC3 (microtubule-associated protein 1A/1B-light chain 3), a well-established autophagosomal marker (Romanov et al., 2012). In addition, a protein complex consisting of Beclin1, ATG14L and VPS34, a class III phosphoinositide 3-kinase (PI3K), has also been shown to play a critical role in autophagosome formation by serving as a scaffold to recruit autophagy targets to the autophagosome lumen (Volinia et al., 1995; Obara and Ohsumi, 2011). The active ULK1 initiation complex positively regulates autophagy at this level by phosphorylating Beclin-1 on Serine14 to promote the activity of VPS34 (Russell et al., 2013).

The process of autophagy is completed when the mature autophagosome docks to and fuses with the lysosome, where its cargo are degraded to release energy and cellular building blocks (Mizushima, 2007; **Figure 1**). This pathway and many of its key molecular constituents are highly conserved throughout evolution, indicating the vital importance of autophagic functions. The most pertinent functions of autophagy include clearance, which eliminates damaged organelles and long-lived proteins that would otherwise compromise cellular health; and catabolism, which releases energy and building blocks to support the growth and activity of the cell (Singh and Cuervo, 2011).

### Cellular Clearance and Neurodegeneration

Though the process of autophagy is ubiquitous throughout a diversity of cells and species, its importance in neurons has become increasingly apparent (Wang and Hiesinger, 2012; Yang et al., 2013). Degradative pathways are essential to maintain balances of cellular energy and stress to promote homeostasis (Ryter et al., 2013); demands on homeostatic regulation are particularly high in neuronal tissue (Chen et al., 2012; Le Masson et al., 2014). As a hallmark of numerous neurodegenerative diseases, the formation of neuronal aggregates highlights the importance of protein clearance in maintaining neuronal health (Polymenidou and Cleveland, 2012; Lim and Yue, 2015). Common to all patients as well as models of ALS is the presentation of pronounced protein inclusions, despite the complex genetic state and broad spectrum of associated mutations involved in the disease. Proteins associated with these aggregates include FUS, TDP-43, UBQLN2 and SOD1, among others (Deng et al., 2011b; Blokhuis et al., 2013). These aggregates may exhibit toxic properties in addition to afflicting the cell by maintaining long-lived proteins (Sau et al., 2007; Johnson et al., 2008; Xu et al., 2013; Wu et al., 2014), as demonstrated in SOD1 mutant mice, where aggregation was shown to drive toxicity in motor neurons. The potency of this toxicity was exacerbated by proteasomal dysregulation (Bruijn et al., 1998; Kitamura et al., 2014).

Decreased efficiency of autophagy, as suggested by the presentation of aggregates, has long been correlated with neurodegeneration. However, a causative relationship between defective autophagy and neurodegeneration was not established until the independent findings that neuronal dysfunction and pathology follow the loss of either Atg5 or Atg8 in mouse (Hara et al., 2006; Komatsu et al., 2006). In both studies, knockout was restricted to the nervous system. Both Atg5- and

Parkin. PINK1 phosphorylates Parkin, and Parkin ubiquitinates mitochondrial outer-membrane proteins, such as mitofusins. Ubiquitination allows for the docking of adaptor proteins such as optineurin, which recruits LC3 to the targeted mitochondrion to form the autophagosome.

Atg8-deficient mice presented with similar phenotypes: profound motor impairment was noted by 3 and 4 weeks of age, manifesting as impaired coordination, poor balance and reduced grip strength. Moreover, both models exhibited the limb-clasping reflex upon suspension, an aberrant behavior noted previously in rodent models of neurodegeneration. Upon histological analysis, brains from both models revealed a high degree of cellular degeneration, accompanied by ubiquitin-positive inclusions and the accumulation of ubiquitinated proteins. These findings strongly suggest that the reduction of basal protein turnover is sufficient to render cells vulnerable to degeneration.

The link between inefficient protein clearance and neurodegeneration is further illustrated by the nature of the mutations resulting in pathological states. Mutations identified in UBQLN2 have been linked to ALS/FTD (Deng et al., 2011b), as well as a more heterogeneous spectrum of neurodegenerative diseases more recently (Fahed et al., 2014). Aggregation of ubiquilin2 protein is found not only in individuals afflicted with heritable UBQLN2 mutations, but also in unrelated sporadic ALS patients and mouse models, suggesting a general role for UBQLN2 in ALS pathogenesis (Deng et al., 2011a). Ubiquilins are ubiquitin-like proteins implicated in regulating autophagy, as well as the ubiquitin proteasome system (UPS; Zhang et al., 2014). They have been shown to colocalize with autophagosomes and associate with LC3 (N'Diaye et al., 2009); furthermore, reduction of ubiquilin levels correlates with a decrease in autophagosome number, suggestive of a critical role in autophagosome formation (Rothenberg et al., 2010). Precisely how and to what extent ubiquilins participate in autophagy, and what implications this has for ALS pathology are exciting questions requiring further attention.

p62/SQSTM1 is another ALS-associated gene involved in degradative pathways (Teyssou et al., 2013). Like UBQLN2, it is associated with both the UPS and autophagy. In the ubiquitin pathway, p62 delivers polyubiquitinated substrates to the proteasome (Seibenhener et al., 2004). In autophagy, p62 interacts directly with LC3 and ATG8 to confer specificity of targeting by selectively acquiring a subset of polyubiquitinated proteins (Pankiv et al., 2007). Just as ubiquilin2-positive inclusions have become a general observation in both familial and sporadic ALS, so have p62-positive inclusions (Mizuno et al., 2006; Deng et al., 2011a), thus signifying the requirement for protein clearance pathways in the maintenance of neuronal health (Fecto and Siddique, 2012).

In recent years mutations in chromosome 9 open reading frame 72 (C9orf72) have emerged as the most common cause of both sporadic and familial ALS (DeJesus-Hernandez et al., 2011; Renton et al., 2011). Though much about this peculiar gene awaits characterization, it is proposed to function as a small GTPase Rab GEF (Guanine Nucleotide Exchange Factor), and has been recently implicated in endosomal trafficking (Zhang et al., 2012; Farg et al., 2014). Furthermore, C9orf72 colocalizes with ubiquilin2 and LC3-positive vesicles, as well as autophagy-related small GTPase Rabs, suggesting a potential role in autophagy. Further studies of C9orf72 will be necessary to elucidate its function in this process, as well as suggest a mechanism of pathogenesis in ALS.

### Mitophagy

Mitophagy refers to the selective process whereby mitochondria are targeted and degraded by the autophagy machinery, primarily for the purpose of eliminating defective organelles. Though the key components and sequence of events are largely conserved between mitophagy and autophagy, mitophagy requires additional mitochondrial-specific adaptors to ensure specificity of targeting (Kim et al., 2007).

The best-studied mitophagy pathway is mediated by PTENinduced putative protein kinase 1 (PINK1) and Parkin. Mutations in PINK1 and PARK2, the gene encoding Parkin, result in pronounced mitochondrial dysfunction, leading to degeneration of muscle and neurons (Poole et al., 2008; Jin and Youle, 2012; **Figure 1**). The finding that overexpression of Parkin is sufficient to mitigate PINK1 mutant phenotypes suggests interaction in the same pathway, with PINK1 upstream of Parkin (Greene et al., 2003; Pesah et al., 2004; Clark et al., 2006). PINK1 is a serine/threonine kinase that associates with mitochondria via a conserved targeting sequence. In healthy mitochondria, PINK1 is imported from the outer to the inner membrane where it is cleaved in a membrane polarization-dependent manner (Deas et al., 2011). When the mitochondrial membrane potential is compromised, PINK1 can no longer be cleaved, thus it accumulates on the outer membrane (Narendra et al., 2010), serving as a signal for recruitment of the E3 ubiquitin ligase Parkin (Vives-Bauza et al., 2010; Shiba-Fukushima et al., 2012). PINK then phosphorylates Parkin and ubiquitin, which is required for Parkin's E3 ligase activity (Koyano et al., 2014). The mechanism underlying phosphorylated ubiquitin activation of Parkin has recently been elucidated (Wauer et al., 2015), following the crystallization of the complex comprising Parkin and phosphorylated ubiquitin. Structural analysis revealed that the binding of phosphorylated ubiquitin to Parkin results in a conformational change within the RING domain, which ultimately results in the activation of Parkin. Importantly, Parkin's ubiquitin binding pocket is commonly mutated in autosomal-recessive Parkinson's, indicating the vital nature of this interaction in promoting mitophagy.

The downstream events that link the targeting of depolarized mitochondria by PINK1-Parkin to the canonical autophagy core machinery are not yet entirely clear. Direct interaction of Parkin with Beclin1 represents one possible explanation (Khandelwal et al., 2011; Lonskaya et al., 2013). Additionally, Parkin has been demonstrated to recruit the Beclin1 regulator AMBRA1 (Activating molecular in BECN1-regulated autophagy protein 1) to the outer membrane of depolarized mitochondria, an event that could also activate the Beclin1 complex and the core autophagy machinery (Van Humbeeck et al., 2011). Interestingly a role for PINK1 was recently discovered in the recruitment of the autophagy receptors optineurin and NDP52 to depolarized mitochondria. This was shown to activate mitophagy in a Parkinindependent manner, whereby optineurin and NDP52 recruit ULK1 and other autophagy factors (Lazarou et al., 2015).

A number of PINK1-Parkin-independent mechanisms of mitophagy have been described. A reduction in iron, for example, has been shown to promote a distinct pathway that does not rely on PINK1-Parkin. This novel pathway involves the transition from oxidative phosphorylation (OXPHOS) to glycolysis and induction of mitophagy without compromising membrane polarization. Importantly, PINK1 stabilization is not required for this process; iron chelation-induced mitophagy is efficiently activated in fibroblasts deficient in Parkin (Allen et al., 2013). Another PINK1-Parkin-independent pathway is suggested by findings on AMBRA1: Parkin is not required for a distinct form of ABMRA1-mediated mitophagy. Though AMBRA1 is known to interact with Parkin, which is thought to contribute to canonical PINK1-Parkin-mediated mitophagy, it has been shown that mitochondrial-targeted AMBRA1 is sufficient to induce mitophagy in a Parkin-free system, therefore representing a novel mitophagic pathway (Strappazzon et al., 2015). Finally, a mechanism of PINK1-Parkin-independent hypoxiainduced mitophagy has been identified. The mitochondrial outer-membrane protein FUNDC1, FUN14 domain-containing protein 1, is a mitophagy receptor requiring LC3 interaction. Either knockdown of FUNDC1 or mutation of its LC3-binidng domain significantly hinders hypoxia-induced mitophagy (Liu et al., 2012). Given our growing knowledge of the roles mitochondrial quality control plays in neuronal maintenance, further elucidation of mitophagic pathways will be necessary to advance our understanding and treatment of neurodegenerative disease.

The strongest evidence supporting the contribution of impaired mitophagy to ALS pathogenesis lies in the associated mutations. The involvement of optineurin, encoded by the OPTN gene, in ALS pathogenesis was identified years after the gene had been implicated in primary open-angle glaucoma (Maruyama et al., 2010). Optineurin is involved in a number of cellular processes including Golgi maintenance and membrane trafficking, but its function as an autophagy receptor is presumably the most relevant to ALS pathogenesis (Turturro et al., 2014). Recently, optineurin was shown to play a significant role in PINK1-Parkin-mediated mitophagy. Damaged mitochondria are initially targeted by the E3 ubiquitin ligase Parkin, which ubiquitinates outer membrane proteins such as mitofusins, the outer membrane-embedded GTPases responsible for mediating mitochondrial fusion (Poole et al., 2010). Optineurin then binds to the ubiquitinated mitochondrial outer membrane proteins with its ubiquitin-binding domain. Hereafter, optineurin induces nucleation of the autophagosome by recruitment of LC3. ALS-causing mutations in OPTN disable this process, implicating inefficient mitochondrial clearance in ALS (Wong and Holzbaur, 2014). It is intriguing that, much like ubiquilin2 and p62, optineurin has been localized to inclusions in both familial and sporadic ALS (Blokhuis et al., 2013), suggesting a broader role for optineurin, and mitophagy, in ALS pathogenesis.

Valosin containing protein (VCP) mutations have been linked with ALS and other degenerative diseases, including dementia (Johnson et al., 2010; Koppers et al., 2012). VCP is a type II ATPase involved in a broad spectrum of biological processes ranging from proteasomal degradation and mitochondrial quality control via mitophagy, to endoplasmic reticulum (ER) associated degradation (Dai and Li, 2001; Rabinovich et al., 2002; Wojcik et al., 2006; Ju et al., 2009). VCP lies downstream of the E3 ubiquitin ligase Parkin and is recruited to the outer membrane of damaged mitochondria. Degeneration-associated mutant VCP or loss of VCP results in the failure of PINK1-Parkin-mediated mitochondrial clearance. Moreover, VCP has been demonstrated to regulate the proteasomal degradation of mitofusins (Kim et al., 2013). Altogether, the role of optineurin and VCP in mitochondrial quality control supports dysregulation of mitophagy as a critical mechanism in ALS pathogenesis.

Very recently, two independent studies identified a definitive link between TBK1, TANK Biding Kinase 1, and ALS, citing a loss-of-function mechanism-of-action (Cirulli et al., 2015; Freischmidt et al., 2015). Substrates of TBK1 include optineurin as well as p62, thus strengthening the connection between mitophagic function and motor neuron degeneration in ALS. Indeed, mutations resulting in disruption of the C-terminal coiled-coiled optineurin-interacting domain of TBK1 were linked with pathogenesis, suggesting a vital role for the kinase in the regulation of mitophagy. Accordingly, a recent study has elucidated the mechanism by which TBK1 acts in mitophagy (Heo et al., 2015). PINK1-Parkindependent phosphorylation of TBK1 activates the kinase to recruit autophagy receptors optineurin and NDP52 to the depolarized mitochondrion. TBK1 phosphorylates these receptors, and in turn, optineurin binding to polyubiquitin chains on the mitochondrial outer membrane enhances TBK1 activation, thus amplifying the mitophagic signal. Importantly, TBK1-dependent phosphorylation of optineurin increases its affinity for polyubiquitin binding, and this step is required for efficient mitophagy. Altogether, these findings put forth a novel framework within which TBK1 plays an integral role in amplifying the mitophagic signal via enhanced recruitment and activation of autophagy receptors on depolarized mitochondria, linking dampened mitophagy with ALS pathogenesis.

### MITOCHONDRIAL FUNCTION: IMPAIRMENTS IMPLICATED IN ALS

#### ATP Generation

ATP is produced in mitochondria through OXPHOS of glucose via the electron transport chain (ETC). In this process, electrons pass along a sequence of protein complexes (I-IV) located in the inner mitochondrial membrane (Milstein and Swaiman, 1968). A flow of protons is pumped from the matrix into the inter-membrane space, generating an electrochemical gradient. When protons flow from high to low concentration through ATP synthase on the inner mitochondrial membrane, ADP is converted to ATP, storing high-energy in a phosphate bond for later use. Linking ATP biogenesis with ALS, a computational modeling endeavor suggested a framework within which mitochondrial dysfunction driving reduced ATP availability underlies motor neuron susceptibility to degeneration. According to this model, ATP reduction drives hyperexcitability through mechanisms such as Na+/K<sup>+</sup> dyshomeostasis, and intracellular calcium hikes (Le Masson et al., 2014).

Reduced ETC activity has likewise been described in both patients and models of ALS, and may significantly contribute to pathogenesis. Compared to healthy controls, ALS-derived fibroblasts show altered mitochondrial bioenergetics: membrane potential is significantly reduced in correlation with age of onset, and an overall decrease in mitochondrial content is noted when compared to controls (Kirk et al., 2014). In line with these findings, a mutant SOD1-mediated switch from OXPHOS to glycolysis has been reported, indicative of inefficient ATP biogenesis in ALS (Allen et al., 2014). Direct genetic evidence of this comes from the recent identification of mutations in CHCHD10 causing ALS and frontotemporal dementia (Bannwarth et al., 2014; Chaussenot et al., 2014). Though the functional repertoire of CHCHD10 awaits further elucidation, it is predicted to play a significant role in energy production through OXPHOS (Bannwarth et al., 2014).

The wealth of data leading to our current understanding of mitochondrial dysfunction in ALS originates from reports on the SOD1 class of mutations (Bendotti and Carri, 2004). These mutations are the most widely studied in ALS, as SOD1 was the first identified FALS-linked gene (Rosen et al., 1993). Wild-type SOD1 is found predominantly in the cytoplasm where it neutralizes the damaging effects of reactive oxygen species (ROS; Fukai and Ushio-Fukai, 2011). A small amount of SOD1 is localized to the mitochondrial inter-membrane space (Jaarsma et al., 2001) where it is suggested to play a protective role in motor neurons (Waterman-Storer et al., 1997). ALS-associated mutations lead to increased localization of misfolded SOD1 protein to the mitochondrial inter-membrane space (Liu et al., 2004; Deng et al., 2006), and mutant SOD1-mitochondrial interactions lead to the alteration of mitochondrial redox potential (Ferri et al., 2006). Accumulation of SOD1 mutant aggregates has been shown to lead to dysfunctional mitochondria with decreased ATP production, calcium buffering and motility defects (Cozzolino and Carri, 2012). Furthermore, OXPHOS activity is impaired in both ALS patients and SOD1 mutant transgenic mice (Bacman et al., 2006). Reduced levels of respiratory chain activity in complexes I-IV were observed in spinal cord tissue collected from ALS patients (Wiedemann et al., 2002). In the SOD1G93<sup>A</sup> mouse model, a pre-symptomatic decrease in Complex I function was noted, while Complex IV was found to be impaired following ALS-like symptom onset (Jung et al., 2002).

### Calcium Buffering

In motor neurons, mitochondria must meet not only high energy demands but also considerable calcium buffering requirements. Healthy mitochondria have enormous calcium buffering capacity and are able to effectively travel from areas of low calcium to areas of high calcium to restore normal levels (Wang and Schwarz, 2009). When mitochondria sense an increase in calcium they become stationary by arresting mitochondrial movement (Rintoul et al., 2003; Szabadkai et al., 2006). As one of the most eminent second messengers in neuronal cells, calcium is required by neurons for neurotransmitter release, modulation of synaptic efficiency, and signal transduction. Unregulated accumulation of calcium proves toxic to cells, however. In neurons, when excitatory glutamate receptors become excessively stimulated, high levels of calcium flow into the cell, leading to cell damage and death (Carriedo et al., 2000; Corona and Tapia, 2007). How calcium mediates this effect is not entirely understood, but its role in enzymatic and signaling cascade activation is one possible explanation.

Dysregulation of calcium homeostasis has been extensively reported in ALS, particularly as it pertains to SOD1 mutation (Jaiswal and Keller, 2009). There have been reports of chronic calcium overload in mitochondria at nerve terminals (Siklós et al., 1996), and SOD1 mutant mouse nerve terminals were shown to be depolarized as a result (Nguyen et al., 2009). Additionally, decreases in calcium loading capacity occur in the spinal cord and brain of mutant SOD1 mice (Damiano et al., 2006). Calcium dysregulation was also linked to SOD1 mutant protein aggregation and impaired mitochondrial movement (Tradewell et al., 2011). Therefore, excitotoxicity following calcium dysregulation is posited to be a prominent mechanism in ALS pathogenesis (Van Den Bosch et al., 2006). While widely supported, this view faces opposition from a study identifying early (presymptomatic), but not late (endstage) hyperexcitability in the context of altered calcium handling (Fuchs et al., 2013). Such findings therefore call into question the role of hyperexcitability in motor neuron cell death.

Motor neurons exhibit a low calcium buffering capacity, largely the result of sparse calcium buffering protein expression (Celio, 1990; Ince et al., 1993; Palecek et al., 1999). Moreover, calcium handling and buffering capacity are impaired in models of ALS (Jaiswal et al., 2009; von Lewinski et al., 2008). Whereas proteins such as parvalbumin and calbindin are abundant in degeneration-resistant neuronal subtypes, conferring more stable calcium homeostasis (Vanselow and Keller, 2000), expression is comparatively limited in the motor neurons affected in ALS (Elliott and Snider, 1995). What is more, the intracellular free calcium increase noted in SOD1G93<sup>A</sup> transgenic mice at both late presymptomatic and symptomatic stages was correlated with decreased expression of calcium-buffering proteins such as SERCA1, SERCA2, and parvalbumin (Chin et al., 2014). Additionally, calciumpermeable α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) expression, especially those lacking the GluR2 subunit, is particularly high in motor neurons, permitting heightened calcium influx (Lips and Keller, 1998; Van Damme et al., 2002). This may contribute to the susceptibility of cultured motor neurons to AMPAR-mediated excitotoxicity upon overstimulation (Carriedo et al., 1996, 2000; Corona and Tapia, 2007). Accordingly, the anti-excitotoxic drug, Riluzole, is at present the only agent approved for the treatment of ALS, which functions by blocking AMPA receptors, thus inhibiting glutamate release and regulating calcium levels in the cytosol (Bellingham, 2011). When tested in the SOD1 mutant mouse model, Riluzole showed an increase in survival (Gurney et al., 1996), but in patients it has been shown to expand life by only several months on average (Bensimon et al., 1994; Lacomblez et al., 1996). It is worth noting that, despite the well-established role of altered calcium handling in ALS, enhanced buffering as achieved by elimination of cyclophilin D was insufficient to extend survival in SOD1 mouse models, in spite of improved motor neuron survival and reduced SOD1 aggregation (Parone et al., 2013).

### Mitochondrial Transport

Motor neurons innervate muscle fibers great distances away from the soma, which requires mitochondria to be continually transported along the axon to areas of high energy and calcium buffering demand. Defects in mitochondrial transport render neurons especially vulnerable to degeneration (De Vos et al., 2008; Wang and Schwarz, 2009). Mitochondria utilize many different motor proteins to move in a saltatory manner, which enables their recruitment to areas of low ATP where they remain docked due to high levels of ADP that halt movement (Mironov, 2007). Generally, about one-third of mitochondria are in motion while two-thirds are stationary (Sheng, 2014). Mitochondria travel along microtubules either away from the soma (anterograde) or towards the soma (retrograde). Kinesins are generally responsible for anterograde travel whereas dynein facilitates retrograde transport (Forman et al., 1983). Other key players in mitochondrial transport include dynactin (Waterman-Storer et al., 1997), syntaphilin (Kang et al., 2008), and the Milton-Miro complex (Stowers et al., 2002; **Figure 2**). Intriguingly, neuronal-restricted loss of Miro culminates in brain stem and spinal cord atrophy resembling motor neuron disease, suggesting a requirement for mitochondrial mobility in motor neuron health and maintenance (Nguyen et al., 2014).

Mitochondria in the ALS-linked SOD1 mutant mouse model experience axonal transport deficits before ALS symptoms arise, resulting in a deficiency of axonal mitochondria (De Vos et al., 2007). Multiple studies have shown impairment in anterograde and retrograde transport (Bilsland et al., 2010; Magrané et al., 2012; Marinkovic et al., 2012). Mutant SOD1 is thought to aggregate in mitochondria and bind to dynein, thus disrupting axonal transport (Ligon et al.,

noted, as is a depletion of the mitochondrial population at the neuromuscular junction. Mitochondrial morphology is also irregular in diseased motor neurons,

2005; Zhang et al., 2007). This can prevent mitochondria from being recruited to areas of high energy or calcium buffering demand, and can also stop defective mitochondria from being trafficked back to the soma. Importantly, such an interaction would affect all axonal transport pathways. Recent studies in the SOD1G93<sup>A</sup> model reveal important effects on dynein-mediated transport of late endosomes. Lysosomal dysfunction occurs in a progressive manner and impairs autophagy-mediated degradation prior to symptom onset in the SOD1G93<sup>A</sup> mode. This occurs through mutant SOD1 binding with dynein, which interferes with retrograde trafficking of late endosomes, resulting in autophagic failure (Xie et al., 2015).

indicating poor health and inefficient turnover.

Mutations in TARDBP, the gene encoding TDP-43, are causal for ALS and like SOD1, have been shown to affect mitochondrial quality and transport. TDP-43 is a highly conserved, ubiquitously expressed DNA/RNA binding protein involved in a wide range of processes important for RNA metabolism (Baralle et al., 2013; Ling et al., 2013). In ALS, however, it was found to be ubiquitinated, phosphorylated and cleaved into insoluble C-terminal fragments that accumulate in cytoplasmic inclusions (Neumann et al., 2006; Johnson et al., 2009). TDP-43 inclusions are found in neurons and glia in both FALS and SALS, with the exception of patients and models presenting with SOD1 mutations (Mackenzie et al., 2007; Tan et al., 2007). Transgenic mice overexpressing wild type human TDP-43 under the control of the Thy1.2 promoter, which drives expression specifically in neurons, displayed motor defects and shortened life span. Spinal motor neurons exhibited cytosolic inclusions and accumulation of fragmented mitochondria. Furthermore, mitochondria were absent from the neuromuscular junction (Shan et al., 2010). In another study, in vivo imaging of mitochondrial movement in mutant TDP-43 transgenic mice revealed a pre-symptomatic impairment of mitochondrial transport in motor neurons, followed by mitochondrial morphological abnormalities (Magrané et al., 2014). Mitochondrial transport defects in TDP-43 transgenic mice can be attributed to the key role that TDP-43 plays in neurofillament stability (Wang et al., 2008; Volkening et al., 2009). Together, SOD1 and TARDBP mutations support a crucial requirement for mitochondrial activity and axonal transport in motor neuron function and maintenance.

### CROSSTALK BETWEEN MITOCHONDRIAL AND AUTOPHAGIC PATHWAYS

### Oxidative Stress and Autophagy Regulation

ROS are by-products of mitochondrial OXPHOS and include superoxide anion and peroxide, which, as a result of unpaired electrons, are unstable and reactive. ROS can readily react with virtually any macromolecule, thereby posing threat of damage to somatic or mitochondrial DNA, lipids, enzymes and other proteins. High levels of ROS lead to elevated oxidative stress within the cell, which is proposed to underlie a number of pathological states (Afanas'ev, 2005; Jang and Van Remmen, 2009). Nonetheless, oxidative stress is posited to serve a pertinent function physiologically, as increasing lines of evidence suggest a chief role in autophagy regulation (Scherz-Shouval et al., 2007; Huang et al., 2011; Scherz-Shouval and Elazar, 2011). ROS have been demonstrated to induce autophagy in starvation conditions through a mechanism requiring AMPK activation (Li et al., 2013), which drives a ROS-dependent decrease in mTOR activity. The upstream regulation of ROS-induced autophagy entails the cytoplasmic activation of ataxia telangiectasia mutated (ATM) kinase, leading to the activation of AMPK. AMPK phosphorylation of TSC2 ultimately results in inhibition of mTOR kinase (Alexander et al., 2010). Therefore, AMPK activity is required for the inhibition of mTOR, which thereby removes repression from the autophagy initiation complex to allow ULK1 phosphorylation by AMPK, triggering autophagy induction.

Dysfunctional and long-lived mitochondria, which are far more likely to engage in reduced-quality energy production, tend to generate excessive ROS as compared to healthy mitochondria (Chistiakov et al., 2014), thus triggering autophagic induction. This is likely a mechanism evolved to restrict damages caused by oxidative stress. Under physiological conditions, ROS are a vital component of autophagy regulation; however, when the autophagic machinery is not competent to respond, ROS levels go unchecked, presenting oxidizing damage to the cell while failing to elicit the appropriate catabolic response. It has been suggested that oxidative stress-induced dysfunction of the autophagy pathway drives the accumulation of longlived and dysfunctional mitochondria (Luo et al., 2013). Moreover the accumulation of damaged mitochondria results in elevated oxidative stress, further challenging the autophagylysosome pathway in a vicious circle (**Figure 3**). This appears to be a prominent mechanism at play in ALS, as afflicted patients and models alike show elevated ROS and signs of unchecked oxidative stress (Weiduschat et al., 2014; Ikawa et al., 2015).

### Cell Death: The Autophagy Connection

Though oxidative stress is a signal of vital importance in promoting homeostasis and inducing autophagy, it also has a chief role in regulating cell death (Ryter et al., 2007). In fact, in models of ALS, ROS has been shown to drive motor neuron cell death (Rojas et al., 2015). Dysfunction of the autophagy-lysosome pathway under oxidative stress is a driving force for the accumulation of damaged mitochondria (Luo et al., 2013), which in turn results in heightened oxidative stress. In instances where this signal is excessive or prolonged, it may elicit a form of cell death independent from apoptosis, termed autophagic cell death (Chen et al., 2008; Li et al., 2013). One form of autophagic cell death, autosis, is characterized by independance from caspases and other constituents of pro-death pathways, as well as unique

Motor neurons' highly polarized morphology and extensive calcium buffering and energy demands likely underlie this vulnerability. (B) Demands on calcium buffering and energy production are fulfilled by mitochondria. ROS are generated as a byproduct of oxidative phosphorylation (OXPHOS). The generally high metabolic requirements of neuronal cells may contribute to elevated ROS, which in turn results in cellular stress and the potential for mitochondrial dysfunction if mitophagic efficiency is not high. (C) Demands of neuronal mitochondria necessitate frequent turnover through mitophagy pathways. Mitochondrial-derived ROS positively regulate autophagy/mitophagy. If the autophagy/mitophagy machinery becomes impaired, ROS is not reduced and mitochondria are not recycled, thus leading to increases in ROS and cellular stress. The accumulation of ROS has potential to damage DNA, proteins and mitochondria, thereby presenting further challenge to the autophagy/mitophagy machinery. As such, the regulation of autophagy/mitophagy through mitochondrial-derived ROS and mitophagy-dependent mitochondrial quality control illustrate two pathways converging onto selective motor neuron death.

morphological features including membrane rupture, shrinkage of the nuclear membrane, electron-dense mitochondria and fragmented ER, and strengthened substrate adhesion. The autotic cell death pathway also features a unique reliance on Na+, K+-ATPases, and can be triggered pharmacologically or by starvation (Liu et al., 2013; Liu and Levine, 2015). This pathway is relevant in vivo, as rodent hippocampal neurons die through autosis following hypoxic-ischemic injury. That treatment with cardiac glycosides, identified in a high-throughput compound screen for autosis inhibitors, robustly rescues autotic cell death in this model suggests therapeutic potential for targeting this pathway (Liu et al., 2013). Though contributions of this pathway to motor neuron death in ALS are uncertain, it is intriguing that Na+/K<sup>+</sup> dyshomeostasis in disease progression has been posited by computational modeling of ALS pathogenesis (Le Masson et al., 2014).

Further evidence concerning additional roles for constituents of the autophagy machinery in traditional caspase-dependent pathways implicates even broader contributions to cell death. For example, cleavage of the autophagy-related gene Atg5 by calpain regulates apoptosis. The cleavage fragment translocates to mitochondria, where it blocks anti-apoptotic Bcl to drive cytochrome-c release and thereby promote caspase-dependent apoptosis (Yousefi et al., 2006). A similar mechanism is supported for Beclin1 in apoptotic induction (Wirawan et al., 2010). Caspase-dependent cleavage of Beclin1 yields a Cterminal fragment that localizes to mitochondria where it potentiates apoptosis, potentially through the release of proapoptotic factors. A role for the autophagy-associated protein ULK1 in promoting cell death, independent of autophagy, has recently emerged. Treatment with H2O<sup>2</sup> facilitates the nuclear localization of ULK1 in an activation-dependent manner. In the nucleus, ULK1 interacts with Poly (ADP-Ribose) Polymerase 1 (PARP1), thereby increasing its activity and potentiating PARP1-induced cell death (Joshi et al., 2016). Interestingly, ULK1 is a transcriptional target of p53; its expression is upregulated in response to DNA damage, and it contributes to cell death following prolonged autophagy (Gao et al., 2011), suggesting both autophagydependent and—independent forms of ULK1-mediated cell death.

Motor neuron cell death in both familial and sporadic forms of ALS is attributable to necroptosis (Re et al., 2014). Necroptosis is a regulated necrotic form of cell death, independent of caspases, thereby distinguishing this pathway from apoptosis. TNF (tumor necrosis factor) induces necroptosis in a manner requiring the inactivation of caspase-8. This promotes the interaction of RIPK1 and RIPK3 (receptor interacting protein kinases), which is vital for the formation of the necrosome. The MLKL (mixed lineage kinase domain-like) pseudokinase is then recruited to execute necroptosis through an unknown mechanism (Linkermann and Green, 2014; Vanden Berghe et al., 2014). It is intriguing that some lines of evidence point to a chief role for the autophagy machinery in necroptosis. In particular, the Bcl-2 inhibitor GX15-070 promotes cells death through RIP1-dependent necroptosis, which requires autophagosome accumulation to serve as sites of necrosome assembly. Inhibition of autophagosome formation precludes this mechanism of cell death, highlighting the importance of constituents of the autophagy-lysosome pathway in necroptotic cell death. Interestingly, RIP1 inhibition does not limit GX15-070-mediated autophagosome accumulation, indicating the role of autophagy induction upstream of RIP1 activity in necroptosis (Basit et al., 2013). Altogether, the roles of autophagy in cell death are complex and incompletely understood, and contributions of these pathways to motor neuron death in ALS require further investigation. Nonetheless, that autophagic dysfunction and cell death frequently co-occur in ALS is perhaps no coincidence.

## THERAPEUTIC PROSPECTIVE

## Boosting Autophagy

That protein aggregates are characteristic of ALS and other neurodegenerative diseases suggests that the autophagy pathway may be an attractive therapeutic target for the prevention and treatment of neurodegeneration (Vidal et al., 2014). Indeed, enhancing autophagy has been shown to effectively target ALS-associated pathological aggregates for clearance to reduce toxicity (Hung et al., 2009; Wang et al., 2012; Barmada et al., 2014). The mTORC1 inhibitor rapamycin can effectively induce autophagy by eliminating mTOR-mediated inhibition of the ULK1 autophagy initiation complex (Jung et al., 2009). Rapamycin administration improved prognosis in TDP-43 mutant mice (Wang et al., 2012), but findings from trials with SOD1 or VCP-associated inclusion body myopathy mutant models did not suggest a similar beneficial effect (Zhang et al., 2011; Ching et al., 2013). Interestingly, in SOD1 mutant mice lacking mature lymphocytes, a modest survival increase was noted following rapamycin treatment (Staats et al., 2013). This finding suggests the benefit of the autophagypromoting effect of rapamycin, as well as its detrimental effect on protective immune function. In support of this finding, preclinical trials with the autophagy-inducing agent trehalose proved successful in a SOD1 mutant model (Castillo et al., 2013). Trehalose promotes autophagy by transcriptional upregulation of ATG genes through an mTOR-independent pathway.

Although direct defects in autophagy genes may not be the causal link in all cases of ALS, dysfunction of autophagy exacerbates disease phenotypes (Bruijn et al., 1998; Kitamura et al., 2014). Therefore, boosting the efficiency of autophagy could ameliorate the toxic effects of aggregates characteristic of ALS and other neurodegenerative diseases. Nonetheless, it is with some caution that we interpret these results, as there are indications that enhanced autophagy may in some instances further exacerbate axonal degeneration and disease phenotype. For example, autophagy has been implicated as an early requirement for axonal degeneration phenotypes in a number of models, including an excitotoxic neurodegeneration model (Wang et al., 2006). Moreover,


ALS-associated genes have functions relevant to degradative and mitochondrial pathways. Specifically, a number of associated genes are involved in the degradative pathways of autophagy/mitophagy and the ubiquitin proteasome system (UPS). Additionally, genes involved in mitochondrial function and quality control indicate the critical importance of mitochondrial health in neuronal maintenance and protection.

the neurite retraction phenotype characteristic of the Parkinsonian LRRK2G2019<sup>S</sup> mutation implicates autophagic induction as an early event. Accordingly, RNAi-mediated knockdown of autophagy factors LC3 or ATG7 is sufficient to rescue neurite length (Plowey et al., 2008). In addition, it was shown using Atg7-deficient mice that negatively regulating autophagy is sufficient to promote cell survival of hippocampal pyramidal neurons following ischemic insult (Koike et al., 2008). These results, taken together with the expanding roles of autophagy in promoting cell death, indicate that it is the regulated balance of autophagy, not merely increased induction, that promotes neuronal health and protection.

### Enhancing Mitochondrial Health

As increasing evidence suggests a prominent role for mitochondrial dysfunction and oxidative stress in motor neuron degeneration, enhancing mitochondrial health represents a promising strategy for treating ALS and related diseases. In line with this, long-term users of the antioxidant vitamin E show a decreased risk for ALS (Wang et al., 2011), consistent with findings of elevated ROS levels (Ikawa et al., 2015) and decreased antioxidant levels (Weiduschat et al., 2014) in ALS-afflicted patients. However, most proposed antioxidant treatments for ALS are not beneficial in clinical trials (Graf et al., 2005; Orrell et al., 2008). This lack of success may be attributed to the low efficiency of targeting antioxidants to mitochondria. To circumvent this issue, the cell permeable Szeto-Schiller antioxidant peptide (SS-31) was investigated (Petri et al., 2006; Szeto, 2006). SS-31 is readily taken up by the cell and localizes to the mitochondrial inner membrane. The specific targeting of SS-31 was shown to inhibit ROS generation and protect against oxidative damage (Zhao et al., 2004, 2005). Presymptomatic treatment with SS-31 in a genetic mouse model of ALS improved disease prognosis significantly (Petri et al., 2006). It remains to be determined whether this therapeutic method offers equal potential for additional models of ALS, and how these findings will translate in human trials.

### CONCLUSION

Protein aggregates are chief characteristics of many neurodegenerative diseases, including ALS (Blokhuis et al., 2013). Gain-of-function properties of disease-associated mutations can directly lead to protein aggregation (Sau et al., 2007; Johnson et al., 2008; Xu et al., 2013; Wu et al., 2014). In addition, loss-of-function of Atg genes is sufficient to cause neurodegeneration in mice, suggesting a prominent role for autophagic clearance pathways in neurodegenerative disease (Hara et al., 2006; Komatsu et al., 2006). Consistent with this, many ALS-disease genes are linked to cellular clearance pathways (Deng et al., 2011b; **Table 1**), and phenotypic improvements have been achieved by enhancing autophagy in some ALS mouse models (Hung et al., 2009; Wang et al., 2012; Barmada et al., 2014). Besides impaired autophagy, defective mitochondrial function, which leads to increased oxidative stress and compromised calcium buffering, also plays a critical role in motor neuron degeneration and ALS pathogenesis. In particular, impaired autophagy and dysfunctional mitochondrial pathways may engage in crosstalk in the onset and progression of ALS. As such, therapeutic strategies targeting these two pathways and their interactions will hold great promise for alleviating disease symptoms and rescuing motor neuron degeneration in ALS and related diseases.

### AUTHOR CONTRIBUTIONS

BME, NM, and YCM conceived of, drafted, edited, and approved the corresponding review.

### ACKNOWLEDGMENTS

This work was supported by NIH grants R01NS094564, AG043970 and grants from the Hartwell Foundation and Whitehall Foundation to YCM. YCM is Ann Marie and Francis Klocke MD Research Scholar supported by the Joseph and Bessie Feinberg Foundation. We thank members of the Ma laboratory for critically reading the manuscript.

## REFERENCES


amyotrophic lateral sclerosis patients. Neurobiol. Aging 35, 2884.e1–2884.e4. doi: 10.1016/j.neurobiolaging.2014.07.022


dynein in motor neurons. Neuroreport 16, 533–536. doi: 10.1097/00001756- 200504250-00002


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Edens, Miller and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## CIIA prevents SOD1(G93A)-induced cytotoxicity by blocking ASK1-mediated signaling

#### *Jae Keun Lee1, Sang Gil Hwang1, Jin Hee Shin2, Jaekyung Shim3 and Eui-Ju Choi <sup>1</sup> \**

*<sup>1</sup> Laboratory of Cell Death and Human Diseases, Department of Life Sciences, School of Life Sciences and Biotechnology, Korea University, Seoul, South Korea*

*<sup>2</sup> Department of Health Sciences and Technology, Samsung Advanced Institute for Health Sciences and Technology, Sungkyunkwan University, Seoul, South Korea*

*<sup>3</sup> Department of Molecular Biology, Sejong University, Seoul, South Korea*

#### *Edited by:*

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### *Reviewed by:*

*Philippe Georgel, Strasbourg University, France Daniel Kaganovich, Hebrew University of Jerusalem, Israel Eldi Schonfeld-Dado, Stanford University, USA*

#### *\*Correspondence:*

*Eui-Ju Choi, Laboratory of Cell Death and Human Diseases, Department of Life Sciences, School of Life Sciences and Biotechnology, Korea University, Green Bldg. Rm431, Seoul 136-701, South Korea e-mail: ejchoi@korea.ac.kr*

Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative disease with higher selectivity in degeneration of motor neurons. However, the molecular mechanism by which the ALS-linked mutants of human superoxide dismutase 1 (SOD1) gene induce neurotoxicity remains obscure yet. Here, we show that depletion of CIIA expression by RNA interference (RNAi) promoted cytotoxicity caused by ALS-linked G93A mutant of the SOD1 gene. The RNAi-mediated knockdown of CIIA also enhanced the SOD1(G93A)-induced interaction between ASK1 and TRAF2 as well as ASK1 activity. Furthermore, endogenous silencing of CIIA by RNAi augmented the effects of SOD1(G93A) on reduction of mitochondria membrane potential (ψm), release of cytochrome c into the cytoplasm, and caspase activation. Together, our results suggest that CIIA negatively modulates ASK1-mediated cytotoxic signaling processes in a SOD1(G93A)-expressing cellular model of ALS.

**Keywords: ALS (amyotrophic lateral sclerosis), ROS (reactive oxygen species), ASK1 (apoptosis signal-regulating kinase 1), mitochondria, cytotoxicity**

#### **INTRODUCTION**

Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative disease with higher selectivity in degeneration of upper and lower motor neurons. The majority of ALS are sporadic, while only 10% of the total ALS are familial. Both the sporadic and familial cases share similar clinical characteristics through possible common mechanisms of the disease. Mutation in the gene encoding superoxide dismnutase-1 (SOD1) is the primary genetic cause of familial ALS (fALS) and provokes the disease with acquired toxic gain-of-function (Gurney et al., 1994; Cleveland and Rothstein, 2001). A number of mechanisms have been proposed to elicit degenerative courses of motor neurons in ALS (Cleveland and Rothstein, 2001; Pasinelli and Brown, 2006). However, the molecular processes underlying the neurotoxicity induced by SOD1 mutants are not understood completely yet. It was previously reported that apoptosis signal-regulating kinase 1 (ASK1) may mediate ALS-associated neurotoxicity with genetic ablation of *ASK1* alleviating fALS disease features in SOD1 mutant transgenic mice (Nishitoh et al., 2008). ASK1 was activated in motor neurons in the lumbar spinal cord of SOD1(G93A) mice along with disease progression (Wengenack et al., 2004; Holasek et al., 2005; Veglianese et al., 2006) and mediated cell-autonomous death pathways induced by SOD1(G93A) (Raoul et al., 2002; Nishitoh et al., 2008). Furthermore, ASK1 appears to mediate MST1-dependent neurotoxicity of motor neurons in ALS model mice (Lee et al., 2013).

A caspase-activated DNase inhibitor that interacts with ASK1 (CIIA) was originally discovered as an anti-apoptotic factor that protects cells from apoptosis induced by tumor necrosis factor-α and a variety of cellular stress such as oxidative stress through the inhibition of ASK1 (Cho et al., 2003; Kim et al., 2010). Given that ASK1 contributes to neurodegeneration in ALS, we investigated the possible role of CIIA in the regulation of ASK1-mediated signaling initiated by SOD1(G93A). Here, we showed that CIIA mitigates SOD1(G93A)-induced cytotoxicity with inhibiting the SOD1(G93A)-induced activation of ASK1-p38 MAPK signaling. These results suggest that CIIA functions as a negative modulator of ASK1-mediated cytotoxic signaling processes in SOD1(G93A) expressing cellular model of ALS.

#### **RESULTS AND DISCUSSION**

#### **CIIA SUPPRESSES SOD1(G93A)-INDUCED CYTOTOXICITY**

Given that CIIA has been shown to prevent apoptotic cell death induced by various cellular stresses including oxidative stress (Cho et al., 2003), we examined whether CIIA affects SOD1(G93A)-induced apoptosis. Depletion of CIIA expression by RNA interference (RNAi) enhanced SOD1(G93A)-induced apoptosis in Neuro2A (N2a) cells (**Figure 1A**). We further analyzed the effect of CIIA on SOD1(G93A)-induced apoptosis in NSC34 mouse motor neuron-like cells, and observed similar results (**Figure 1B**).

#### **CIIA INHIBITS SOD1(G93A)-INDUCED ASK1 ACTIVITY**

ASK1, a member of the MAP3K family, has been shown to mediate SOD1(G93A)-induced neurotoxicity in an ALS mouse model (Nishitoh et al., 2008). Indeed, we observed that overexpression of ASK1(K709R), a dominant-negative mutant of ASK1, reduced SOD1(G93A)-induced apoptosis in NSC34

cells (Supplementary Figure 1), suggesting that ASK1 is associated with the mechanism of SOD1(G93A)-induced cytotoxicity. Moreover, the scavengers of reactive oxygen species (ROS) including Trolox and *N*-acetylcysteine (NAC) abrogated the SOD1(G93A)-induced increase in ASK1 activity (Supplementary Figure 2), suggesting that SOD1(G93A) induces ASK1 activation in a ROS-dependent manner. Given that CIIA was initially identified as an inhibitory protein of ASK1 (Cho et al., 2003), we investigated the possible effect of siRNA-mediated CIIA depletion on ASK1 activity in NSC34 cells expressing human SOD1(WT) or SOD1(G93A). SOD1(G93A) increased the kinase activities of ASK1 and its downstream kinase, p38 MAPK, compared to those of the cells expressing SOD1(WT) (**Figure 2A**). These results thus suggested that CIIA prevents the SOD1(G93A)-induced stimulation of ASK1 and p38 MAPK. It is noteworthy that CIIA did not affect ROS generation induced by SOD1(G93A) (**Figure 2B**).

Given that CIIA is able to physically interact with ASK1 (Cho et al., 2003), we examined the CIIA-ASK1 interaction in NSC34 cells expressing either SOD1(WT) or SOD1(G93A). Coimmunoprecipitation analysis revealed that CIIA physically associated with ASK1 in NSC34 cells expressing SOD1(WT), and this association of the CIIA-ASK1 complex was reduced in the cells expressing SOD1(G93A) (**Figure 2C**, lower panel). These results suggest that SOD1(G93A) promotes the dissociation of CIIA from ASK1, thereby facilitating ASK1 activation. We also examined a possible effect of CIIA on TRAF2-ASK1 interaction, because the recruitment of TRAF2 to ASK1 is an integral part of the mechanism for ROS-induced ASK1 activation (Fujino et al., 2007). SOD1(G93A) induced the binding of TRAF2 to ASK1 in NSC34 cells and this binding was potentiated in those expressing a CIIA siRNA (**Figure 2C**, upper panel), suggesting that CIIA suppresses the SOD1(G93A)-induced TRAF2-ASK1 interaction.

#### **CIIA INHIBITS SOD1(G93A)-INDUCED IMPAIRMENT OF MITOCHONDRIA HOMEOSTASIS AND CASPASE ACTIVATION**

Previously, we reported that the ASK1-p38 signaling pathway mediates the SOD1(G93A)-induced activation of caspase-9 and -3 (Lee et al., 2013). We, therefore, decided to examine a possible effect of CIIA on the mitochondrial intrinsic pathway for apoptosis involving reduction of mitochondrial membrane

TRAF2 or to CIIA.

siRNA. **(A,B)** Cell lysates of the stable transfectants were examined for ASK1 and p38 kinase activities by immune complex kinase assay **(A)** or for

potential (ψm), release of cytochrome c from mitochondria, and activation of caspase-9 and -3. SOD1(G93A) reduced ψ<sup>m</sup> in NSC34 cells, compared to the cells expressing SOD1(WT) (**Figure 3A**). Furthermore, depletion of CIIA expression by RNAi promoted the reducing effect of SOD1(G93A) on ψ<sup>m</sup> in the cells expressing SOD1(G93A) (**Figure 3A**). In addition, SOD1(G93A) induced the release of cytochrome c from the mitochondria to the cytoplasm in NSC34 cells and this effect of SOD1(G93A) was enhanced by siRNA-mediated CIIA depletion (**Figure 3B**). We also examined the effect of CIIA on SOD1(G93A)-induced activation of caspase-9 and -3. SOD1(G93A) induced the increase in the enzymatic activities of caspase-9 (**Figure 4A**) and -3 (**Figure 4B**) as well as cleavage (activation) of the caspases (**Figure 4C**). The RNAi-mediated depletion of CIIA potentiated the effects of SOD1(G93A) on caspase-9 and -3 (**Figures 4A–C**). Collectively, these results suggested that CIIA negatively regulates the effects of SOD1(G93A) on reduction of ψm, release of cytochrome c from the mitochondria into the cytoplasm, and activation of caspase-9 and -3.

Previous studies using SOD1G93A/ASK1−/<sup>−</sup> mice have demonstrated that ALS-linked mutant SOD1 induces ER stress and that ASK1 mediates neuronal death induced by SOD1 mutant-induced ER stress (Nishitoh et al., 2008). In this study, we have shown that CIIA physically associates with ASK1 in the cells expressing SOD1(WT), thereby interfering TRAF2-ASK1 interaction, which is a part of mechanism for ASK1 activation. In contrast, SOD1(G93A) induces the release of ASK1 from CIIA and promotes the association of ASK1 with TRAF2, which leads to ASK1 activation. SOD1(G93A)-induced ASK1 activation results in the stimulation of p38 MAPK and impaired mitochondrial homeostasis as well as the activation of caspase-9 and -3, all of which contribute to apoptotic cell death. Depletion of CIIA expression by RNAi potentiated these effects of SOD1(G93A). In relation to these results, our study is under way to investigate whether CIIA overexpression may improve the severity of disease manifestations in the fALS model mice. It would be noteworthy that overexpressed CIIA has been shown to prevent apoptotic cell death mediated by ASK1 activation (Cho et al., 2003). Taken together, these results suggest that CIIA negatively regulates SOD1(G93A)-induced cytotoxicity through inhibiting ASK1. Our findings thus implicate CIIA as a potential modulator of ALS-associated neurotoxicity.

**FIGURE 3 | CIIA inhibits SOD1(G93A)-induced reduction of mitochondria membrane potential (****ψm) and the release of**

**cytochrome c.** NSC34 cells were stably transfected with plasmid vectors encoding Flag-tagged SOD1(WT) or SOD1(G93A) along with vectors for GFP or CIIA siRNA. **(A)** mitochondria membrane potential (ψm) was analyzed by measuring the fluorescence intensity of tetramethyl rhodamine methyl ester (TMRM, a mitochondria potential sensor). Quantitative data are mean <sup>±</sup> s.e.m. from three independent experiments. <sup>∗</sup>*<sup>P</sup>* <sup>&</sup>lt; <sup>0</sup>.05 vs. WT; †*<sup>P</sup>* <sup>&</sup>lt; <sup>0</sup>.05 vs. G93A. **(B)** Cell lysates were subjected to the subcellular fractionation to obtain the mitochondrial (mito) and the cytosolic (cyto) fractions. Each fraction was subjected to immunoblot analysis with antibodies to cytochrome c, to α-tubulin (cytosolic marker), or to COX IV (mitochondrial marker).

## **MATERIALS AND METHODS**

#### **CELL CULTURE AND DNA TRANSFECTION**

NSC34 motor neuron–like cells and Neuro2a (N2a) cells were cultured under a humidified atmosphere of 5% CO2 at 37◦C in Dulbecco's modified Eagle's medium (Gibco, Carlsbad, CA) supplemented with 10% heat-inactivated fetal bovine serum (Gibco). For DNA transfection, cells were plated on glass coverslips in six-well culture dishes or in a 100-mm dish, grown for 24 h, and transfected for 48 h with appropriate vectors with the use of polyethyleneimine (Sigma-Aldrich).

### **RNA INTERFERENCE**

For depletion of CIIA, NSC34 cells were transfected with a pSuper-retro vector (Oligoengine) for GFP (control) or CIIA siRNA with the use of RNAiMAX (Invitrogen), and the stably transfected cells were selected with puromycin (3μg/ml). The target sequences for CIIA and GFP were 5 -AAGGCCTACATCAAGGACTGT-3 and 5 -GGCTACGTCC AGGAGCGCACC-3 , respectively. For transient transfection of siRNA oligonucleotides with RNAiMAX (Invitrogen), we used GFP (control) siRNA (5 GCTGGAGTACAACTACAACAGCCA CAACG-3 ) and CIIA siRNA (5 -CCTGGGAACAAGCCGGAG CTGTATGAGGA-3 ).

### **IMMUNE COMPLEX KINASE ASSAY**

Immune complex kinase assays were performed as described (Park et al., 2001). GST fusion proteins of MAP kinase kinase 6(K82A) [MKK6(K82A)] and activating transcription factor 2 (ATF2) were used as substrates specific for ASK1 and p38, respectively.

### **CO-IMMUNOPRECIPITATION ANALYSIS**

Cell lysates were incubated for 3 h at 4◦C with appropriate antibodies and then for 1 h in additional presence of protein G-conjugated agarose. The resulting precipitates were subjected to immunoblot analysis with the indicated antibodies.

### **PREPARATION OF THE MITOCHONDRIAL AND CYTOSOLIC FRACTIONS**

NSC34 cells were washed twice with cold PBS and harvested in a cytosolic lysis buffer [250 mM sucrose, 137 mM NaCl, 4.3 mM Na2HPO4, 1.4 mM KH2PO4 (pH 7.2), 70μg/ml digitonin, 0.1 mM phenylmethylsulfonyl fluoride, 1μg/ml aprotinin, and 1μg/ml leupeptin] for 30 min on ice. The harvests were subjected to centrifugation for 5 min at 600 × g and the resulting pellets were indicated as the nuclear fractions. The supernatants were centrifuged at 10,000 × g for 10 min and the resulting pellets were designated as the mitochondrial faction and resuspended with a mitochondria lysis buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 0.2% (v/v) Triton X-100, 0.3% NP-40, 0.1 mM phenylmethylsulfonyl fluoride, 1μg/ml aprotinin, and 1μg/ml leupeptin]. The supernatants were collected and re-centrifuged at 20,000 × g for 20 min to yield the cytosolic fraction. To verify the fractionation, each fraction was subjected to immunoblot analysis with antibodies to α-tubulin (a cytosolic marker) and COX IV (a mitochondria marker).

### **APOPTOSIS ASSAY**

N2a cells were trypsinized, stained with annexin V-FITC and propidium iodide (PI), and analyzed by flow cytometry (FacsCalibur, Becton-Dickinson) with CellQuest software (BD Bioscience) for the detection of apoptotic cells (annexin V-positive and PI-negative). NSC34 cells were fixed, permeabilized, and incubated with TUNEL reaction mixture (Roche, Penzberg, Germany). TUNEL-positive cells were analyzed for apoptosis under Olympus BX53 fluorescence microscope. Data were expressed as the percentage of TUNEL-positive cells.

### **MEASUREMENT OF INTRACELLULAR ROS PRODUCTION**

Intracellular ROS generation was analyzed by dichlorofluorescin (DCF) assay as described (Lee et al., 2013). All data were normalized with respect to the fluorescence intensity of oxidized DCDHF in SOD1(WT)-expressing cells stably transfected with GFP (control) siRNA.

#### **MEASUREMENT OF MITOCHONDRIA MEMBRANE POTENTIAL (**ψ**M)**

**(A)** or Ac-DEVD-amc **(B)**, and then assayed for caspae-9-like **(A)** or

NSC34 cells were incubated for 30 min with tetramethylrhodamine methyl ester (TMRM, Invitrogen) at 37◦C in the dark. The intensity of TMRM fluorescence was analyzed with FL-600 microplate fluorescence reader (excitation, 360 nm; emission, 460 nm). All data were normalized with respect to the fluorescence intensity of the SOD1(WT)-expressing cells stably transfected with GFP (control) siRNA.

#### **CASPASE ASSAY**

NSC34 cells were incubated with 20μM of fluorogenic substrates (Ac-LEHD-amc and Ac-DEVD-amc) in an assay buffer (50 mM Tris–HCl, pH 7.4, 4 mM DTT, 2 mM EDTA, 10% glycerol, and 0.1% Triton X-100) for 1 h at 37◦C in the dark. Ac-LEHD-amc and Ac-DEVD-amc are the fluorogenic substrates specific for caspase-9 and caspase-3, respectively. The relative fluorescence of the cleaved substrate was monitored with FL-600 microplate fluorescence reader (excitation, 360 nm; emission, 460 nm). All data were normalized with respect to the monitored fluorescence in SOD1(WT)-expressing cells stably transfected with GFP (control) siRNA.

#### **STATISTICAL ANALYSIS**

Data are expressed as mean ± s.e.m. Comparisons among more than two groups were conducted with analysis of variance (ANOVA) followed by the Student Newman-Kelus test. All analyses were conducted with SPSS version 12.0 software. A *P*-value of < 0.05 was considered statistically significant.

#### **ACKNOWLEDGMENTS**

CIIA, to Flag, or to α-tubulin.

We thank S. Kang (Korea University, Seoul, Korea) for providing Flag-SOD1 and Flag-SOD1(G93A) cDNAs, and H. Ichijo (University of Tokyo, Tokyo, Japan) for providing ASK1 cDNA. This work was supported by the National Research Foundation Grant (2011-0030141, 2006-0093855) funded by the Ministry of Science, ICT & Future Planning of Korea and by a Korea University grant (Eui-Ju Choi). Jae Keun Lee was supported by the National Research Foundation Grant (2014R1A1A1002076).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fncel.2014. 00179/abstract

#### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 09 February 2014; accepted: 10 June 2014; published online: 26 June 2014. Citation: Lee JK, Hwang SG, Shin JH, Shim J and Choi E-J (2014) CIIA prevents SOD1(G93A)-induced cytotoxicity by blocking ASK1-mediated signaling. Front. Cell. Neurosci. 8:179. doi: 10.3389/fncel.2014.00179*

*This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2014 Lee, Hwang, Shin, Shim and Choi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## GNX-4728, a novel small molecule drug inhibitor of mitochondrial permeability transition, is therapeutic in a mouse model of amyotrophic lateral sclerosis

#### **Lee J. Martin1,2,3\*, Daniele Fancelli <sup>4</sup> , Margaret Wong<sup>1</sup> , Mark Niedzwiecki <sup>1</sup> , Marco Ballarini <sup>4</sup> , Simon Plyte<sup>4</sup> and Qing Chang<sup>1</sup>**

<sup>1</sup> Department of Pathology, Division of Neuropathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA

<sup>2</sup> Pathobiology Graduate Program, Johns Hopkins University School of Medicine, Baltimore, MD, USA

<sup>3</sup> Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD, USA

<sup>4</sup> Congenia Srl-Genextra Group, Milan, Italy

#### **Edited by:**

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### **Reviewed by:**

Ezia Guatteo, Fondazione Santa Lucia IRCCS, Italy Joseph Beckman, Oregon State University, USA

#### **\*Correspondence:**

Lee J. Martin, Department of Pathology, Division of Neuropathology, Johns Hopkins University School of Medicine, 558 Ross Building, 720 Rutland Avenue, Baltimore, MD 21205-2196, USA e-mail: martinl@jhmi.edu

Amyotrophic lateral sclerosis (ALS) is a fatal neurological disorder in humans characterized by progressive degeneration of skeletal muscle and motor neurons in spinal cord, brainstem, and cerebral cortex causing skeletal muscle paralysis, respiratory insufficiency, and death. There are no cures or effective treatments for ALS. ALS can be inherited, but most cases are not associated with a family history of the disease. Mitochondria have been implicated in the pathogenesis but definitive proof of causal mechanisms is lacking. Identification of new clinically translatable disease mechanism-based molecular targets and small molecule drug candidates are needed for ALS patients. We tested the hypothesis in an animal model that drug modulation of the mitochondrial permeability transition pore (mPTP) is therapeutic in ALS. A prospective randomized placebo-controlled drug trial was done in a transgenic (tg) mouse model of ALS. We explored GNX-4728 as a therapeutic drug. GNX-4728 inhibits mPTP opening as evidenced by increased mitochondrial calcium retention capacity (CRC) both in vitro and in vivo. Chronic systemic treatment of G37R-human mutant superoxide dismutase-1 (hSOD1) tg mice with GNX-4728 resulted in major therapeutic benefits. GNX-4728 slowed disease progression and significantly improved motor function. The survival of ALS mice was increased significantly by GNX-4728 treatment as evidence by a nearly 2-fold extension of lifespan (360 days– 750 days). GNX-4728 protected against motor neuron degeneration and mitochondrial degeneration, attenuated spinal cord inflammation, and preserved neuromuscular junction (NMJ) innervation in the diaphragm in ALS mice. This work demonstrates that a mPTPacting drug has major disease-modifying efficacy in a preclinical mouse model of ALS and establishes mitochondrial calcium retention, and indirectly the mPTP, as targets for ALS drug development.

**Keywords: motor neuron disease, therapeutics, motoneuron, mitochondrial permeability transition pore, mitochondria, mitochondrial calcium uptake, neuromuscular junction**

#### **INTRODUCTION**

Mitochondrial mechanisms have been implicated in the pathogenesis and progression of amyotrophic lateral sclerosis (ALS; Wong et al., 1995; Bendotti et al., 2001; Beal, 2005; Martin, 2010a; Reddy and Reddy, 2011; Muyderman and Chen, 2014). Mitochondrial-based mechanisms of disease in ALS might include failure of intracellular Ca2<sup>+</sup> homeostasis, oxidative stress propagation, energy depletion, perturbed fission-fusion dynamics, and cell death initiation (Beal, 2005; Reddy and Reddy, 2011), but it is possible that mitochondrial changes are indirectly related to disease etiology and are secondary or bystander events (Morais and De Strooper, 2010; Muyderman and Chen, 2014). Previous therapeutic studies have approached the mitochondrial role in ALS using antioxidants, metabolic boosters, and dietary manipulations, mostly in mouse models (Reddy and Reddy, 2011), but these approaches generally lacked specificity for establishing mitochondria or specific components or properties of mitochondria as disease targets. Recently, drugs with putative mitochondrial mechanisms of action, such as dexpramipexole and olesoxime, have failed in human ALS clinical trials (Cudkowicz et al., 2013; Lenglet et al., 2014), but these drugs have pleiotropic actions or unclear mitochondrial mechanisms of action. Well-defined, unequivocal and specific mitochondrial mechanisms and targets for directed therapeutics and disease prevention in ALS have remained elusive.

The mitochondrial permeability transition pore (mPTP) is emerging as a critical player in neurodegenerative disease and in acute neuropathology (Martin et al., 2009; Martin, 2010a,b). While the definitive core components of the mPTP have been fleeting (Bernardi et al., 2006; Halestrap, 2009), but now thought to involve the c-subunit ring of the F1F<sup>0</sup> ATP synthase (Bonora et al., 2013; Alavian et al., 2014), a consistent critical regulator of the mPTP *in vivo* and *in vitro* is cyclophilin D (Bernardi et al., 2006; Halestrap, 2009; Alavian et al., 2014). A cyclophilin D knockout study was important in establishing mitochondria as having a direct role in the mechanisms of disease in preclinical mouse models of ALS (Martin et al., 2009). The mPTP as a target of therapeutics in ALS (Martin, 2010b) needs to be validated and then translated to preclinical animal models using meaningful pharmacologic approaches rather than genetic approaches. Very few drugs have been validated as compounds specifically targeting putative components or functions of the mPTP such as CRC. A class of cinnamic anilide derivatives has been recently synthesized and identified as mPTP inhibitors endowed with *in vivo* therapeutic activity in protecting heart mitochondria from calcium overload and rabbit heart from ischemia (Fancelli et al., 2014). These compounds are able to inhibit mPTP opening in response to calcium overload, oxidative stress, and chemical cross-linkers in isolated mitochondria (Fancelli et al., 2014). We studied GNX-4728, a cinnamic anilide compound from the same series, which inhibits the mPTP and protects mitochondria from calcium overload by increasing CRC. We then tested GNX-4728 for therapeutic actions in a transgenic (tg) mouse model of ALS. This study shows that chronic treatment of G37R-human mutant superoxide dismutase-1 (hSOD1) tg mice with GNX-4728 strongly protects against onset of ALS and robustly extends survival with preservation of motor neuron number, motor neuron mitochondria, and neuromuscular junction (NMJ) integrity.

#### **MATERIALS AND METHODS**

#### **MICE**

Adult wildtype non-tg C57BL/6 mice and tg mice were used. Tg mice were hemizygous for a low copy number of hSOD1-G37R mutant allele driven by the endogenous human promoter (line 29) derived from a founder B6.Cg-Tg SOD1-G37R 29Dpr/J (stock # 008229, The Jackson Laboratory, Bar Harbor, MA) as described (Gertz et al., 2012; Wong et al., 2013). Mice were used with approval from the institutional Animal Care and Use Committee.

#### **DRUG**

GNX-4728 is a substituted cinnamic anilide (**Figure 1A**) which belongs to a novel series of potent inhibitors of the mPTP (Fancelli et al., 2014).

#### **MITOCHONDRIAL CALCIUM RETENTION CAPACITY (CRC) ASSAY**

CRC assays were performed on freshly isolated mitochondria from adult non-tg mouse brain and heart (*n* = 6) after GNX-4728 was administered intravenously by tail vein injection (15 mg/kg in 20% DMSO and 40% PEG400) followed by a survival of 5 min. Control mice (*n* = 6) were injected with vehicle. Brain and heart mitochondria were isolated using a similar procedure as described (Wong et al., 2013). Mitochondrial CRC was assessed fluorimetrically in the presence of the fluorescent Ca2<sup>+</sup> indicator Calcium Green 5N (Invitrogen Molecular Probes) using a temperature controlled Perkin-Elmer LS 55 spectrofluorimeter as described

(Fancelli et al., 2014). Briefly, purified organ mitochondria were pulse-loaded with 10 mM calcium and then challenged with increasing concentrations of calcium until mitochondrial permeability transition was triggered as evidenced by complete release of mitochondrially-stored calcium due to mPTP opening.

and brain (p < 0.01) compared to vehicle (combined organ mitochondria).

#### **TG MICE AND DRUG TREATMENT PROTOCOL**

Cohorts of tg mice expressing mutated G37R-hSOD1 were bred and identified by genotyping of tail DNA as described (Martin et al., 2007, 2009; Wong and Martin, 2010). All mice were housed in the institutional vivarium with generally 4–5 mice per cage and *ad libitum* food and water. Starting at 6 months of age, before the onset of overt symptoms, male G37R-hSOD1 mice were treated with 300 µg (100 µl) of GNX-4728 or vehicle (DMSO/cyclodextrin/saline) every other day by intraperitoneal injection. Only male mice were used because of known genderdifferences in the involvement of the mPTP regulator cyclophilin D in ALS pathobiology (Martin et al., 2009) and to minimize burden to the operators treating mice over long-term with individual injections. This dosage and treatment regimen was in part based on preliminary data showing GNX-4728 protection of ALS mouse spinal cord mitochondria from Ca2+-induced swelling (Martin et al., unpublished observations). The treatment group size totals were 18–20 mice, but these mice were divided among different experiments, including survival studies. The initial randomization was accomplished using the envelope method with knowledge of litter origin so that individuals from the same litter could be distributed across groups. Mice were maintained in the colony and received vehicle or GNX-4728 treatment. For survival experiments, the vehicle and GNX-4728 group sizes were 10 and 12, respectively. The mice were observed once a day. The disease progresses usually from hindleg paraplegia with some function of the forelimbs remaining. When animals developed paraplegia, *ad libitum* nutrigel and chow were placed in the cage, and water was available at a bottom-cage level drinking spout. At this time locomotor activity is compromised but the animals can still ambulate to access food and water. At this stage, the mice are observed three or more times a day. Endstage disease was defined as complete lack of locomotor activity, detected within 2–4 h after onset, at which time the mice were euthanized. A subgroup of mice was killed before endstage to assess the efficacy of GNX-4728 using histological endpoints. Disease onset was assessed by running wheel activity and hind-limb paresis. Vehicle and GNX-4728 treated mice and age-matched non-tg littermate mice were evaluated for neurologic deficit. They were assessed at 12 months of age using a voluntary activity wheel (Harvard Apparatus).

#### **SPINAL CORD AND NEUROMUSCULAR PATHOLOGY**

Vehicle and GNX-4728 treated tg mice and naïve mice (*n* = 5 per group) at 12 months of age were deeply anesthetized and perfused by cardiac puncture with ice-cold 100 mM phosphatebuffered normal saline (PBS) followed by 4% paraformaldehyde in PBS. After perfusion-fixation, mouse bodies were stored at 4◦C overnight and then the spinal cord and diaphragm were removed from each mouse. Spinal cords were cryoprotected (20% glycerol) before they were frozen-sectioned (40 µm) transversely using a sliding microtome. Serial tissue section arrays were stored individually in 96-well plates in anti-freeze buffer. The diaphragm was removed as a complete tissue sheet (Comerford and Fitzgerald, 1986) and placed in PBS at 4◦C until processed for NMJ visualization.

Nissl-stained transverse sections of spinal cord were used to count the number of motor neurons in vehicle and GNX-4728 treated tg mice and in age-matched littermate non-tg mice. Spinal cord sections were selected with a random start and then systematically sampled (every 10*th* section) to generate a subsample of sections from each mouse lumbar spinal cord that was mounted on glass slides and stained with cresyl violet for cell counting. Nissl-stained motor neurons in ventral horn were counted by individuals blinded to experimental treatment, using strict morphological criteria, in digital images acquired with a Nikon microscope at 200x magnification. These criteria included a round, open, pale nucleus (not condensed and darkly stained), globular Nissl staining of the cytoplasm, and a diameter of ∼20– 40 µm. With these criteria, astrocytes, oligodendrocytes, and microglia were excluded from the counts, but these counts are likely to estimate the combined populations of α- and γ-motor neurons.

Mitochondrial pathology was assessed specifically within spinal cord motor neurons using immunohistochemistry. Freefloating spinal cord sections from GNX-4728- and vehicle-treated G37R-hSOD1 tg mice and age-matched non-tg mice were stained by an immunoperoxidase method with antibodies to the mitochondrial matrix marker superoxide dismutase-2 (Stressgen) and diaminobenzidine as described (Martin et al., 2007). Mitochondrial diameters within motor neurons were measured by individuals blinded to experimental history using ocular filar micrometry as described (Martin et al., 2007).

Diaphragm motor endplates were visualized with Alexa 594 conjugated α-bungarotoxin (BTX, Invitrogen, Molecular Probes) as described (Martin and Liu, 2007). Dual labeling was done to visualize motor neuron distal axons and their terminals in whole diaphragm preparations by immunofluorescent detection of neurofilament protein using a monoclonal antibody (SMI-32, Convance) and confocal microscopy as described (Martin and Liu, 2007). The immunofluorescent labeling for neurofilament was used to determine whether the BTX-labeled motor endplates were innervated. Confocal microscope images of the typical band distributions of motor endplates in diaphragm (Comerford and Fitzgerald, 1986) were scored as innervated (normal) if there was overlap with the axon terminal or denervated (unoccupied) if the endplate was not associated with an axon. NMJ imaging and scoring were performed by individuals unaware of mouse treatment.

#### **DATA ANALYSIS**

The values shown in the graphs represent the mean ± standard deviation. For histological data, group means and variances were evaluated statistically by one-way ANOVA and a Student's *t*-test. Time-to-event measures (disease onset and survival duration) were analyzed using Kaplan-Meier survival fit analysis. The Cox proportional hazards model was used to analyze the effect of GNX-4728 on survival and to determine hazard ratios. There was no censoring of mice due to drug- or treatment-related deaths. A one-way ANOVA followed by Tukey *post-hoc* test were used for statistical comparisons for time-to-event measures.

#### **PHOTOGRAPHY AND FIGURE CONSTRUCTION**

The original images used for figure construction were generated using digital photography. Digital images were captured as TiFF files using a SPOT digital camera and SPOT Advanced software (Diagnostic Instruments) or a Nikon digital camera (DXM1200) and ACT-1 software. Images were altered slightly for brightness and contrast using ArcSoft PhotoStudio 2000 or Adobe Photoshop software without changing the content and actual result. Figure composition was done using CorelDraw X5 software with final figures being converted to TiFF files. Files of composite figures were adjusted for brightness and contrast in Adobe Photoshop.

#### **RESULTS**

#### **GNX-4728 ENHANCES MITOCHONDRIAL CRC**

The mPTP blocking activity and blood brain barrier permeability of GNX-4728 *in vivo* was assessed by measuring its ability to prevent isolated calcium pulse-loaded organ mitochondria from mPTP opening after systemic drug treatment. Mice were treated with 15 mg/kg GNX-4728 or vehicle iv and, 5 min later, mitochondria were isolated from heart and brain and then assayed for mPTP opening induced by calcium (**Figure 1B**). Systemic treatment with GNX-4728 significantly increased mitochondrial CRC in heart and brain (**Figure 1B**). The increase in brain mitochondrial CRC is indicative of the ability of GNX-4728 to cross the blood-brain barrier (BBB). Other compounds of this series were able to protect heart, liver and kidney, but not brain mitochondria after *in vivo* administration which supports the notion that GNX-4728 actively crosses the BBB (data not shown).

#### **GNX-4728 INCREASES LIFESPAN OF ALS MICE**

GNX-4728 treatment of G37R-hSOD1 mice strongly protected against clinical onset of disease and robustly extended survival compared to littermate vehicle-treated ALS mice (**Figure 2**). Disease onset (mean ± SD), as assessed by running wheel activity and hind-limb paresis, was significantly (*p* < 0.01) delayed with GNX treatment (381 ± 45 days) vs. vehicle (159 ± 39 days). Median lifespan (mean ± SD) was significantly (*p* < 0.01) increased with GNX treatment (686 ± 120 days) vs. vehicle (366 ± 29 days). Hazard ratio determinations of the two treatment groups revealed significantly different hazard rates for the GNX-4728 treated mice as 0, 0.1, 0.11, 0.12, 0.29, 0.8, and 1.0 at 420, 490, 560, 630, 700, 770, and 840 days of age, respectively (**Figure 2**). For some ALS mice treated with GNX-4728 their lifespan was almost doubled (**Figure 2**).

#### **GNX-4728 PROTECTS AGAINST SPINAL CORD, MITOCHONDRIAL, AND DIAPHRAGM PATHOLOGY IN ALS MICE**

Twelve-month-old vehicle-treated and GNX-4728-treated ALS mice and age-matched non-tg mice were evaluated histologically for motor neuron numbers, mitochondrial swelling, and inflammatory changes in spinal cord and for NMJ innervations in diaphragm (**Figure 3**; **Table 1**). Nissl-staining was used to visualize motor neurons in spinal cord (**Figures 3A–C**). In preparations of non-tg mice, large spinal motor neurons were prominent and inflammatory changes were not present (**Figure 3A**). At 12 months of age, vehicle-treated G37R-SOD1 tg mice were symptomatic and had about 80% loss of motor neurons in lumbar spinal cord (**Figures 3B,D**) and prominent small cell infiltration and reactive inflammatory changes (**Figure 3B**). Iba1 immunoreactivity, a marker for inflammation (Zhao et al., 2013), was significantly elevated in vehicle-treated ALS mouse spinal cord compared to non-tg control (**Figure 3A**). In contrast,

indicated. See text for description of hazards ratios (at graph bottom).

**diaphragm neuromuscular junctions (NMJs) in G37R-hSOD1 tg mice**. **(A–C)** Brightfield microscope images of cresyl violet (Nissl)-stained lumbar spinal cord sections from a 12-month-old non-tg mouse **(A)** and G37R-hSOD1 tg mice that received vehicle **(B)** or GNX-4728 **(C)** treatments. Scale bar (in **A**) = 40 µm (same for **B,C**). **(D)** Graph showing the number of (Continued)

#### **FIGURE 3 | Continued**

lumbar spinal cord motor neurons in G37R-hSOD1 tg mice that received vehicle or GNX-4728 (GNX) treatments. Values are mean ± SD (N = 5/group). Significant differences ∗∗p < 0.001 or <sup>∗</sup>p < 0.05 from non-tg mice. **(E–G)** Brightfield microscope images of Iba1 (microglial marker)-immunostained lumbar spinal cord sections from a 12-month-old non-tg mouse **(E)** and G37R-hSOD1 tg mice that received vehicle **(F)** or GNX-4728 **(G)** treatments. Scale bar (in **E**) = 100 µm (same for **F,G**). **(H)** Graph showing the immunodensity of spinal cord Iba1 immunoreactivity in non-tg mice and G37R-hSOD1 tg mice that received vehicle or GNX-4728 (GNX) treatments. Values are mean ± SD (N = 5/group): significant difference <sup>∗</sup>p < 0.05 from non-tg mice; significant difference <sup>+</sup>p < 0.05 from vehicle-treated mice. **(I–K)** Confocal microscope images of diaphragm NMJs stained for skeletal muscle motor endplates (α-bungarotoxin, red) and motor neuron axons (neurofilament, green) from a 12-month-old non-tg mouse **(I)** and G37R-hSOD1 tg mice that received vehicle **(J)** or GNX-4728 **(K)** treatments. Scale bar (in **I**) = 100 µm (same for **J,K**). **(L)** Graph showing the percent innervation of motor endplates in diaphragm. Significant difference <sup>∗</sup>p < 0.01 from non-tg mice.

#### **Table 1 | GNX-4728 protection of spinal cord motor neuron mitochondria**.


<sup>1</sup>Values are mean ± standard deviation. <sup>∗</sup>GNX4728 vs. non-tg and vehicle, p < 0.05. ∗∗Vehicle vs. non-tg, p < 0.01.

12-month-old GNX-4728-treated ALS mice had significant preservation of motor neurons (**Figures 3C,D**) and significantly attenuated inflammation as evidenced by the diminished Iba1 immunoreactivity (**Figures 3F–H**).

To assess the ability of GNX-4728 to protect mitochondria directly within motor neurons we used immunohistochemistry to detect the mitochondrial marker SOD2 and measured mitochondrial diameters. In non-tg mouse spinal motor mitochondria are about 0.4–0.5 µm in diameter (**Table 1**), consistent with previous observations (Martin et al., 2007). In 12-month-old vehicle-treated G37R-hSOD1 tg mice, mitochondrial diameters were increased significantly compared to age-matched non-tg mice (**Table 1**). GNX-4728 significantly protected against mitochondrial swelling in G37R-hSOD1 tg mouse spinal motor neurons (**Table 1**).

To assay for whether GNX-4728 protects NMJs in ALS mice, a whole-mount diaphragm preparation was used. In non-tg mice, motor endplate innervation of diaphragm was near 100% (**Figures 3I,L**), while in vehicle-treated ALS mice endplate innervation was only about 20% (**Figures 3J,L**). In contrast, in GNX-4728-treated ALS mice, NMJ innervation was restored to about 80% of non-tg control (**Figures 3K,L**).

#### **DISCUSSION**

Our study demonstrates that a small molecule cinnamic anilide derivative, GNX-4728, has several major therapeutic benefits in a mouse model of ALS. GNX-4728 increased brain mitochondrial CRC *in vivo* after systemic administration. Importantly from a preclinical perspective, chronic systemic treatment of ALS mice with GNX-4728 resulted in the following: (1) a delay in disease onset; (2) dramatically increased lifespan; (3) protection of spinal cord motor neurons; (4) protection of spinal cord motor neuron mitochondria; (5) block of spinal cord inflammatory changes; and (6) preservation of NMJs in diaphragm. These results are particularly exciting because they support the concept of the mPTP as a practical drugable therapeutic target in ALS *in vivo* and demonstrate that cinnamic anilides could be a future avenue to the effective treatment of ALS.

In this study we used a tg mouse model of ALS to test the therapeutic efficacy of GNX-4728. These mice express a low copy number of hSOD1-G37R mutant allele in a non-conditional expression pattern throughout the body (Wong et al., 1995). This mouse model is very different from the hSOD1-G93A tg mouse model that expresses the mutant allele at a very high copy number and consequently have an aggressive disease and a lifespan of only about 4 months (Gurney et al., 1994). Because of the rapidity of the disease course, the hSOD1-G93A highexpresser tg mouse has been the most commonly used mouse model to assess ALS therapeutics. However, although these mice develop a fatal paralysis and show a very prominent lower motor neuron disease (Gurney et al., 1994; Bendotti et al., 2001; Martin et al., 2007), the details of the cellular and molecular pathology appear to be distinct from that seen in human ALS (Martin, 1999, 2010a; Martin and Liu, 2004; Martin et al., 2007). These differences might contribute to the lack of success in the clinical translation of drugs shown to have therapeutic efficacy in the hSOD1-G93A tg mouse model. Therefore, we used, instead of the high-expressing hSOD1-G93A tg mouse line, the hSOD1-G37R mouse line with a much longer lifespan. The major downsides of using the longer-lived mouse line is the time commitment for drug treatments and the modeling of only a subtype of familial ALS caused by a mutated hSOD1 gene.

We found that GNX-4728 improved neurologic function and survival of ALS mice. The effect on lifespan was robust. A limitation of the current study is the relative small group sizes for the survival analysis, as it would have been better to have doubled the number of animals in each group. Few pharmacological studies have been done using the hSOD1-G37R mouse line, and the most notable study showed that mice fed chow containing a cocktail of riluzole, nimodipine, and minocycline had about a 15% increase in maximal lifespan (Kriz et al., 2003). This modest effect might be due to the lack of direct disease mechanism-based targeting. In fact, treatment of human ALS patients with riluzole, a sodium channel antagonist with anti-glutamatergic actions, prolongs life by only about 2–3 months in some patients, but there is no visible functional improvement (Bensimon et al., 1994). Questions persist about the clinical usefulness of riluzole because of its modest effects and high cost (Miller et al., 2012). Another anti-glutamate drug, gabapentin, failed in clinical trials (Miller et al., 2001). Thus, a role for glutamate and excitotoxicity in the direct mechanisms of ALS is appearing tenuous (Martin, 2010a), and the concept has so far failed to bear effective therapeutic fruit after more than 20 years (Rothstein et al., 1992). Other disease-mechanism concepts need to be considered to potentially move the treatment of ALS forward. A role for the mPTP in the mechanism of ALS is an attractive concept that can bring together old and new ideas regarding cell degeneration in ALS including, intracellular calcium dysregulation, mitochondria, oxidative stress, and possibly excitotoxicity (Martin, 2010a,b; Reddy and Reddy, 2011). In the current preclinical animal model study, we attribute the success of GNX-4728 to its actions on mitochondria. It penetrated the BBB and increased mitochondrial CRC in brain, revealed by the increased threshold for calcium-induced permeability transition, suggesting that GNX-4728 is operating by inhibiting the mPTP. Striated muscle mitochondrial CRC was also increased. Genetic studies have revealed the participation of the mPTP in the mechanisms of disease in mouse models of ALS (Martin et al., 2009). Other studies support the role of mitochondrial calcium dysregulation and mPTP activation in the mechanisms of ALS (Bendotti et al., 2001; Jaiswal and Keller, 2009; Barrientos et al., 2011; Nguyen et al., 2011; Bartolome et al., 2013; Muyderman and Chen, 2014). The finding that GNX-4728 bocks mitochondrial swelling directly within motor neurons and protects these cells further points to the mPTP as a target of disease in ALS and that GNX-4728 is possibly providing therapeutic benefit through modulation of mPTP function.

GNX-4728 was delivered systemically in our study. Its therapeutic activity could be within the CNS, because of its BBB permeability, at peripheral locations (skeletal muscle and nerves), or both. It could be acting at skeletal muscle mitochondria to provide therapeutic effects. Skeletal muscle is also a primary site of disease in mouse models of ALS (Dobrowolny et al., 2008; Wong and Martin, 2010; Luo et al., 2013) and possibly in human ALS (Vielhaber et al., 2000). Mitochondria within skeletal muscle of ALS mice show early abnormalities in calcium signaling capacity (Zhou et al., 2010), DNA methylation and autophagy (Wong et al., 2013), and metabolism (Dupuis et al., 2006). Additional studies need to address the potential effects of GNX-4728 and non-BBB permeable cinnamic anilides in skeletal muscle of ALS mice.

Despite evidence derived from human ALS and mouse and cells models of ALS indicating that mitochondria have a role in disease pathogenesis (Beal, 2005; Martin, 2010a; Reddy and Reddy, 2011), recent human ALS clinical trials with putative mitochondrial acting drugs have been unsuccessful. Dexpramipexole and olesoxime both failed to show significant benefits in ALS patients (Cudkowicz et al., 2013; Lenglet et al., 2014). Dexpramipexole can act on brain mitochondria to increase the efficiency of oxidative phosphorylation (Alavian et al., 2012). Olesoxime has pleiotropic actions. It protects neurons from trophic factor and target deprivation and apoptosis (Bordet et al., 2007; Martin et al., 2011). Olesoxime binds proteins in the outer mitochondrial membrane (Bordet et al., 2007, 2010), but it does not increase mitochondrial CRC (Bordet et al., 2010). Thus, the failure of dexpramipexole and olesoxime as therapies for human ALS could be due to their inability to modulate mitochondrial CRC.

To our knowledge, no other experimental drug or experimental manipulation has shown this level of therapeutic efficacy in ALS mice. We have not observed untoward side-effects *in vivo*, even after chronic treatment of mice. This study identifies a new drug class, cinnamic anilides, which might be promising for the treatment of ALS. Future experiments are warranted that test the efficacy of GNX-4728-like compounds in other mouse models of ALS to assess their potential general application for the treatment of different forms of ALS.

#### **AUTHOR CONTRIBUTIONS**

Conceived and designed experiments: Lee Martin, Simon Plyte. Provided reagents: Daniele Fancelli, Simon Plyte. Performed experiments: Lee Martin, Margaret Wong, Marco Ballarini, Qing Chang. Analyzed the data: Lee Martin, Margaret Wong, Qing Chang, Mark Niedzwiecki. Wrote the paper: Lee Martin, Daniele Fancelli, Simon Plyte.

#### **ACKNOWLEDGMENTS**

The authors are grateful to Dr. Federica Draghi for her encouragement for this project and her review of the manuscript. This work was supported by a grant from Congenia Srl-Genextra Group, Milan, Italy. Drs. Martin, Wong, and Chang were supported by grants from the U.S. Public health Service, National Institutes of Health, National Institute of Neurological Disorders and Stroke (NS034100, NS065895).

#### **REFERENCES**


**Conflict of Interest Statement**: The study was funded in part by Congenia Srl-Genextra.

*Received: 30 August 2014; accepted: 01 December 2014; published online: 19 December 2014*.

*Citation: Martin LJ, Fancelli D, Wong M, Niedzwiecki M, Ballarini M, Plyte S and Chang Q (2014) GNX-4728, a novel small molecule drug inhibitor of mitochondrial permeability transition, is therapeutic in a mouse model of amyotrophic lateral sclerosis. Front. Cell. Neurosci. 8:433. doi: 10.3389/fncel.2014.00433*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2014 Martin, Fancelli, Wong, Niedzwiecki, Ballarini, Plyte and Chang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## The role of oxidative stress in degeneration of the neuromuscular junction in amyotrophic lateral sclerosis

#### **Eveliina Pollari 1,2\*, Gundars Goldsteins <sup>1</sup> , Geneviève Bart <sup>3</sup> , Jari Koistinaho <sup>1</sup> and Rashid Giniatullin 3,4**

<sup>1</sup> Molecular Brain Research Laboratory, Department of Neurobiology, A. I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland

<sup>2</sup> Experimental Neurology - Laboratory of Neurobiology, Department of Neurosciences, Vesalius Research Center, KULeuven – University of Leuven, Leuven, Belgium

<sup>3</sup> Cell Biology Laboratory, Department of Neurobiology, A. I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland

<sup>4</sup> Laboratory of Neurobiology, Department of Physiology, Kazan Federal University, Kazan, Russia

#### **Edited by:**

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### **Reviewed by:**

David Marcinek, University of Washington, USA Dan Lindholm, Helsinki university, Finland Andreas H. Kottmann, The Sophie Davis School of Biomedical Education, City University of New York, USA

#### **\*Correspondence:**

Eveliina Pollari, Molecular Brain Research Laboratory, Department of Neurobiology, A. I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Yliopistonranta 1, P. O. Box 1627, Kuopio 70211, Finland e-mail: Eveliina.pollari@uef.fi

Amyotrophic lateral sclerosis (ALS) is characterized by the progressive loss of motoneurons and degradation of the neuromuscular junctions (NMJ). Consistent with the dying-back hypothesis of motoneuron degeneration the decline in synaptic function initiates from the presynaptic terminals in ALS. Oxidative stress is a major contributory factor to ALS pathology and affects the presynaptic transmitter releasing machinery. Indeed, in ALS mouse models nerve terminals are sensitive to reactive oxygen species (ROS) suggesting that oxidative stress, along with compromised mitochondria and increased intracellular Ca2<sup>+</sup> amplifies the presynaptic decline in NMJ. This initial dysfunction is followed by a neurodegeneration induced by inflammatory agents and loss of trophic support. To develop effective therapeutic approaches against ALS, it is important to identify the mechanisms underlying the initial pathological events. Given the role of oxidative stress in ALS, targeted antioxidant treatments could be a promising therapeutic approach. However, the complex nature of ALS and failure of monotherapies suggest that an antioxidant therapy should be accompanied by anti-inflammatory interventions to enhance the restoration of the redox balance.

**Keywords: ALS, neuromuscular junction, ROS, oxidative stress, neurodegeneration**

#### **NMJ AS A VULNERABLE TARGET OF ALS (DYING BACK HYPOTHESIS)**

Temporal analysis of axon and neuromuscular junction (NMJ) degeneration in sporadic ALS (sALS) and mouse mutant SOD1 (mSOD1) cases indicate that motoneuron pathology begins distally from the synaptic area (**Figure 1**) markedly earlier than clinical symptoms and proceeds towards soma in a retrograde dying back manner (Fischer et al., 2004; Rocha et al., 2013). Impaired axonal transport, Ca2<sup>+</sup> imbalance and mitochondria dysfunction drive the axonal degeneration, and eventually lead to dying of the neuron (Fischer-Hayes et al., 2013).

**Figure 1** shows the principal structure of the NMJ including the presynaptic machinery restricted to active zones (AZ) releasing acetyl choline (ACh) in quantal manner and postsynaptic structures consisting of densely packed ACh receptors linked to the muscle-specific kinase (MuSK), agrin and other molecules involved in NMJ maturation and maintenance (reviewed in Shi et al., 2012). Thus, the dysfunction of the neuromuscular transmission can originate from the presynaptic site or from disorganized postsynaptic density. Notably, the motor nerve terminals are covered by the Terminal Schwann Cells (TSC) which can contribute to ALS progression.

In mSOD1 mice many motor terminals of the diaphragm muscle show a number of dysfunctional changes in the early disease stage (Naumenko et al., 2011). Muscle fibers are proposed to initiate the early changes leading to ALS progression (Wong and Martin, 2010). However, our results indicate that in the NMJ of ALS mice the presynaptic machinery is affected first (Naumenko et al., 2011). There is a noticeable variation in the probability of transmitter release between synapses, suggesting different degeneration rates of synapses. At the early symptomatic phase, only a few synapses have compromised function. Presumably, early in the disease course, the proportion of damaged synapses is low allowing compensation of the lost function by the healthy ones. Interestingly, during the pre-symptomatic stage enhanced neuromuscular transmission can be observed before the occurrence of the marked decline during the symptomatic phase, possibly due to compensatory mechanisms against the initial degeneration (Rocha et al., 2013).

Indeed, while some axon branches degenerate in ALS, others show sprouting thus compensating for lost synapses (Schaefer et al., 2005). Supporting the regenerating axons provides a therapeutic opportunity for maintaining innervation. However, as the disease progresses the proportion of damaged synapses increases and the sparse functional synapses cannot mediate synaptic transmission anymore. In mouse models of ALS axons of fast-fatiguable motoneurons are affected synchronously in hindlimbs, long before symptoms appear, whereas axons of slow

motoneurons are more resistant. Thus it is possible that ALS involves predictable, selective vulnerability patterns of NMJs by physiological subtypes of axons, where NMJs of the resistant axons partially assume compensatory functions (Pun et al., 2006; Dibaj et al., 2011).

In some mSOD1 mouse models, oxidative stress appears to originate from distal muscles before the disease onset (Kraft et al., 2007). Reactive oxygen species (ROS) affect synaptic transmission by inhibiting transmitter release. Increasing ROS levels further inhibit NMJ function in spite of already elevated level of oxidative stress (Naumenko et al., 2011). This suggests that oxidative damage could start in peripheral tissues and proceed retrogradely to neurons. Skeletal muscle targeted expression of mSOD1 provokes motor deficits, but at a rather late age and without evident effect on the life expectancy (Wong and Martin, 2010). In this particular model, the muscle pathology is accompanied by NMJ abnormalities and distal motoneuron axonopathy. Initiation of the motoneuron degeneration by muscle cells supports the hypothesis of dying-back pathogenesis where the neurodegeneration starts from deficits in muscle and NMJs and proceeds from distal axons towards soma leading to apoptosis of motoneurons (Fischer et al., 2004; Dupuis and Loeffler, 2009).

### **PRESYNAPTIC PART OF THE NMJ AS THE MAIN SENSITIVE PART REACTING TO OXIDATIVE STRESS**

Measurements from the diaphragm muscle of G93A-SOD1 mice have revealed a dramatic reduction in the frequency of miniature end-plate potentials (MEPPs) during the early symptomatic stage (Naumenko et al., 2011). Remarkably, no changes in the amplitude of MEPPs were observed, indicating purely presynaptic decline in the synaptic transmission. The amplitude of single evoked EPPs remained unchanged suggesting vulnerability of spontaneous quantal transmitter release from nerve terminals.

This phenotype (selective depression of MEPPs with little affected EPPs) largely resembles the inhibitory action of ROS on transmitter release at the frog NMJ: exogenous H2O<sup>2</sup> elicits a strong inhibition of spontaneous release with limited effect on EPPs (Giniatullin and Giniatullin, 2003). NMJ impairment in ALS could therefore be produced by mechanisms similar to those, which affect synapses damaged by oxidative stress. Recent studies revealed distinct mechanisms underlying spontaneous versus evoked transmitter release (Maximov et al., 2007; Pang et al., 2011; Melom et al., 2013). Soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNARE) protein, Snap25, was identified as the main targets of ROS at the presynaptic level (Giniatullin et al., 2006) and could be one of the targets of ROS inhibiting transmitter release (**Figure 1B**).

Apart from ROS, another interesting candidate contributing to the damage of NMJ in ALS is extracellular ATP. ATP, the major co-transmitter of ACh at the NMJ (Redman and Silinsky, 1994), can produce a strong inhibitory action on transmitter release (Giniatullin and Sokolova, 1998) via ROS induction (Giniatullin et al., 2005; Sciancalepore et al., 2012). This mechanism could contribute to the motor nerve terminal dysregulation in ALS (**Figure 1B**) or, when applied persistently, to ATP-driven neurodegeneration of NMJ (**Figure 1C**). This view is consistent with recent data showing that extracellular ATP, operating via cytotoxic P2X7 receptors could largely regulate immune function and inflammatory responses (Volonté et al., 2012). Notably, TSC also express P2X7 receptor (Grafe et al., 1999; Colomar et al., 2003; Nobbio et al., 2009). Whereas accumulating evidence suggest that Schwann cells can contribute to ALS (De Winter et al., 2006; Gorlewicz et al., 2009; Lobsiger et al., 2009; Chen et al., 2010), the role of myelinating versus non-myelinating TSCs in ALS however requires, further studies (Turner et al., 2010).

We propose that the early damage to the NMJ in ALS is due to intraterminal dysregulation of nerve terminals without essential changes in their morphology (**Figure 1B**). Underlying mechanisms probably include dysfunctional mitochondria, intracellular Ca2<sup>+</sup> and ROS. Elevated intraterminal Ca2<sup>+</sup> can eventually support enhanced Ca2+-dependent evoked release during early stage of ALS (Rocha et al., 2013). The other model (**Figure 1C**), applicable to the later stage of ALS, suggests that the main damage results from the accumulation of toxic ROS, inflammatory factors, including glial transmitters from local Schwann cells and invasive immune cells, and absence of neuroprotective trophic factors. However, these two scenarios most likely co-exist within the same muscle during ALS progression providing a heterogeneous picture of morphological and functional changes (Rocha et al., 2013) and resulting in the pitfalls of the monotherapy in this disease.

#### **AXONAL TRANSPORT, PRESYNAPTIC MITOCHONDRIA AND ROS-INDUCED ROS RELEASE**

Correct spatial distribution of mitochondria within a cell is an instrumental prerequisite for normal physiology. In neurons, mitochondria are subjected to both anterograde and retrograde axonal transport, which in case of motoneurons covers substantial distances. The transport of mitochondria in axons is driven along microtubules by kinesin and dynein motors (Pilling et al., 2006).

Accumulation of mitochondria at presynaptic nerve terminals of motoneurons is thought to support synaptic function through ATP production and partially take part in Ca2<sup>+</sup> buffering during neurotransmission (**Figure 1;** Chouhan et al., 2010). Mitochondria are connected to the presynaptic membrane by a complex cytoskeletal superstructure, which is connected with nerve terminal filamentous linkages between synaptic vesicles, providing polarized organization for mitochondrial crista structures (Perkins et al., 2010). Defects in neuronal mitochondrial morphology and axonal transport have been demonstrated in primary neuronal cultures from ALS model animals (De Vos et al., 2007; Magrané et al., 2012). Importantly, these abnormalities are also observed in vivo in both SOD1 and TDP43 ALS mouse models, indicating that they are common denominators of different genetic forms of ALS (Magrané et al., 2014).

The high order of mitochondrial organization at presynaptic nerve terminals implies their participation in coordinated responses to various stimuli. One of the fundamental oxidative stress responses in mitochondria is mitochondrial permeability transition (MPT) pore opening, followed by sudden collapse of membrane potential and burst of ROS production, which might contribute to the spreading of MPT in bystanding mitochondria, and lead to the effect known as ROS-induced ROS release (Zorov et al., 2000). The latter can contribute to the functional dysregulation within the nerve terminal during the early stage of ALS (**Figure 1B**). Our studies have demonstrated that SOD1 activity is increased in ALS animal spinal cord mitochondria, and causes elevated hydroperoxide production (Ahtoniemi et al., 2008; Goldsteins et al., 2008). Augmented hydroperoxide flux from presynaptic mitochondria might contribute not only to reduced probability of quatal ACh release but also to the desynchronization of neurotransmitter release at NMJ (Tsentsevitsky et al., 2013) which would additionally diminish synaptic efficacy (**Figure 1C**). Apart from presynaptic location, muscle mitochondria and activity of NADPH oxidase in TSCs can serve as an additional source of ROS (**Figure 1**).

#### **NEUROINFLAMMATION, IMMUNE CELLS AND OXIDATIVE STRESS IN SPINAL CORD IN ALS**

Oxidative stress, such as free radical damage and abnormal free radical metabolism, is evident in sALS and fALS patients (Shaw et al., 1995; Ferrante et al., 1997; Smith et al., 1998; Chang et al., 2008). The aberrant activity of mSOD1 leads to oxidative damage (Wiedau-Pazos et al., 1996; Crow et al., 1997) and other ALSlinked proteins, such as mutant TDP-43, promote oxidative stress in a motoneuron cell line (Duan et al., 2010). Excitotoxicity and oxidative stress caused by astrocytes arises from aberrant glutamate receptor function which leads to misregulated glutamate homeostasis (Rothstein et al., 1992). Oxidative stress promotes tissue damage by exacerbating and interacting with other pathological events that promote motoneuron degeneration.

Inflammation, which is an additional source of ROS, is evident in ALS patients and mSOD1 mice; microglia are activated and proliferating whereas the T cells and dendritic cells infiltrate into the spinal cord (Engelhardt et al., 1993; Henkel et al., 2004, 2006). Moreover, there is marked increase in proinflammatory cytokines and enzymes, such as interleukin-6 (IL-6), monocyte chemoattractant protein-1 (MCP-1), IL-8, and cyclooxygenase-2 (Cox-2) (Sekizawa et al., 1998; Almer et al., 2001, 2002; Elliott, 2001; Hensley et al., 2002; Kuhle et al., 2009). Astrocytes expressing mSOD1 are also prone to exhibit an activated pro-inflammatory state (Hensley et al., 2006; Di Giorgio et al., 2008; Marchetto et al., 2008). Activated pro-inflammatory M1 microglia cause ROS and glutamate excitotoxicity induced motoneuron injury and death (Zhao et al., 2004). MSOD1 induced oligodendrocyte dysfunction drives demyelination in the spinal cord and accelerates motoneuron degeneration (Kang et al., 2013).

Immune responses are also activated in peripheral tissues of ALS patients (Mantovani et al., 2009). Regulatory T (Treg) cells lower neuroinflammation through microglia by inducing secretion of anti-inflammatory cytokines IL-10 and transforming growth factor-β (TGF-β; Kipnis et al., 2004; Mantovani et al., 2009). In ALS patients, elevated levels of Treg cells and CD4 T cells in blood correlate with slow disease progression (Beers et al., 2011). T cell infiltration in the spinal cord in mSOD1 mice is amplified during the presymptomatic stage and the number of T cells in the spinal cord increases as the disease progresses (Beers et al., 2008; Chiu et al., 2008). The spinal cord T cell population is mainly composed of helper CD4 cells. The proportion of cytotoxic CD8 becomes prominent at the end-stage. This supports the assumption that during the early stages of the disease, there is a systemic combat to maintain neuroprotective responses, but as the disease aggravates, the immune response shifts towards cytotoxic.

Macrophages infiltrate ventral spinal roots, peripheral motor nerves and skeletal muscles in ALS mouse models (Chiu et al., 2009; Graber et al., 2010). The role of macrophages in affected tissues in ALS mice appears to be the phagocytic removal of debris from axonal degeneration. Thus, activated macrophages could contribute to ROS production in axons and muscle in ALS and along with other inflammatory agents participate in triggering of sprouting in nerve terminals. However, in ALS mice the majority of activated macrophages accumulated within fascicles of motoneurons in the peripheral tissues and were only rarely found adjacent to end-plate of NMJs. It is therefore unlikely, that macrophages directly contribute to oxidative damage of NMJs in ALS.

Interestingly, in ALS motoneurons in the brainstem oculomotor nuclei and Onuf's nucleus in the sacral spinal cord are preserved and selective vulnerability seems to be related to oxidative stress. Reduced capability to maintain calcium homeostasis and disturbed mitochondrial function predispose specific motoneurons to degeneration in ALS (Vanselow and Keller, 2000; Jaiswal and Keller, 2009). In addition, the most vulnerable motoneurons are more prone to endoplasmic reticulum stress and exhibit increased susceptibility to excitotoxicity (Saxena et al., 2009; Brockington et al., 2013).

#### **GENDER DEPENDENCE OF ALS AND OXIDATIVE STRESS**

ALS affects men more than women, with earlier age of onset for men as well a tendency for spinal initiation of the disease whereas in women it is more commonly bulbar (McComb and Henderson, 2010). Most of these features were also observed in mSOD1 animals (Veldink et al., 2003; Suzuki et al., 2007). Gender specific differences are also detected at the synapses: Synaptic vesicle release being more frequent in females' endplate with impairment only observed in males (Naumenko et al., 2011). Specific interneurons control motoneuron excitability via specialized cholinergic synaptic boutons: C-boutons. In ALS there is no gender difference in the number of C-boutons, but their size is bigger in male mSOD1 mice (Herron and Miles, 2012).

The most obvious explanation for gender differences is a protective effect of estrogen. ROS damage in muscle is limited in young women. Even aging women have significantly less lipid peroxidation, protein carbonylation and mitochondrial DNA damage than men (Pansarasa et al., 2000). Several estrogencontrolled pathways might protect females against fast neuromuscular degeneration in ALS, for instance estrogen-mediated cyclophilin D prevention of mitochondrial calcium overload (Kim et al., 2012). However, experiments with ovariectomized mSOD1 mice or rats with and without supplemental 17β-estradiol do not support the idea that estrogen could explain the gender differences in ALS (Choi et al., 2008; Hayes-Punzo et al., 2012).

Additional evidence for gender differences in the ROS balance are coming to light, for instance, lower blood level of uric acid (UA) were observed in ALS patients (Keizman et al., 2009). UA, a scavenger of NO radical and superoxide, reduces damage to cells, by preventing protein nitration on tyrosine residues by peroxynitrites, and higher level of UA in blood increases likelihood of longer survival in men (Paganoni et al., 2012).

Gender differences are also striking in the response to treatments. Examples of therapeutic approaches with gender bias are specific inhibition of spinal cord microglial P2X7, which appears to increase the duration of life without affecting the age of symptom onset in male (Cervetto et al., 2013) and G-CSF treatment which delays disease progression in male mSOD1 mice (Pitzer et al., 2008; Naumenko et al., 2011; Pollari et al., 2011).

#### **PROMISING THERAPEUTIC APPROACHES AND CHALLENGES OF THE ANTIOXIDANT THERAPY IN ALS**

Several molecules with antioxidant capabilities have failed in clinics after showing promise in animal models (Gordon, 2013; Musarò, 2013; Pandya et al., 2013; Sreedharan and Brown, 2013). Still, riluzole is the only approved drug that delays the progression of ALS.

The translational failures in ALS can be explained by the same arguments as in other neurodegenerative diseases: (a) animal models represent only a fraction of genetic variations among ALS patients and do not model well sALS; (b) the preclinical studies are characterized by inadequate randomization, blinding, statistical power, control cohorts and consideration of comorbidities; and (c) optimal dosing and administration route in clinics are unknown, flaws in patient stratification or identification of proper patients, and insufficient samples size (Ubogu, 2012; Planas, 2013).

Perhaps the most important reason for the translation block is that by the time of diagnosis, ALS has already progressed too far, making prevention of further deterioration challenging. The late diagnosis allows multiple disease mechanisms to accelerate their contributions to motoneuron death overriding the regenerative mechanisms and preconditions. Also, nerve terminal retraction, axonal degeneration and eventual neuronal death may take weeks or months before their deteriorating effects become clinically noticeable (Coleman and Freeman, 2010; Sreedharan and Brown, 2013). The late time of diagnosis is an especially important concern for protection of NMJ functions, as NMJ degeneration is among the earliest pathological alterations in ALS.

During the presymptomatic phase of the ALS, oxidative stress may be triggered by increased production of superoxide and nitric oxide in neurons, central and peripheral glia and even in muscle cells. Also, levels of a major intracellular antixodiant, glutathione, reduce early in ALS tissues. At the same time, bloodspinal cord barrier appears to become leaky and infiltration of inflammatory cells into the spinal cord, motor nerves and muscles contributes to oxidative stress present prior to the disease onset (Henkel et al., 2004; Chang et al., 2008; Chiu et al., 2009; Chen et al., 2010; Halter et al., 2010; Dibaj et al., 2011; Drechsel et al., 2012; Winkler et al., 2014). In fact, several findings favor the idea of linking ALS therapy to the oxidative stress-related degeneration of NMJ. First, normal SOD1 activity is required for maintenance of NMJ function in aged rodents (Sakellariou et al., 2014) and, in a zebrafish model expressing mSOD1 in physiological levels, NMJ has increased susceptibility to oxidative stress showing early morphological alterations (Da Costa et al., 2014). Second, even though pathological changes in synapses and axons occur early during the ALS pathogenesis, these selfdestructive mechanisms could be delayed by correcting molecular environment (Sreedharan and Brown, 2013). Third, the NOXmediated increase in superoxide production takes place in neural cells in mSOD1 models of ALS (Harraz et al., 2008; Jaronen et al., 2013). Even though is it not known whether NOX is expressed in Schwann cells around NMJs, NOX could well contribute to oxidative deterioration of NMJ in ALS, as muscle cells express various isoforms of NOX. While it is unclear whether ALSlinked mutations or conditions in sALS could trigger activation of NOX in skeletal muscles, inhibitors of NOX activation are known to provide protection in ALS models. Considering that NOX activation pathway is a readily druggable target, the role of NOX in NMJ degeneration in ALS models is worth exploring (Sullivan-Gunn and Lewandowski, 2013). Finally, it is of interest that Vitamin D, an essential dietary vitamin with multiple physiological functions, has been demonstrated to influence several aspect of ALS pathology, including skeletal muscle strength and oxidative stress (Gianforcaro and Hamadeh, 2014).

Overall, recent research on NMJ, oxidative stress and inflammation in ALS models warrant further preclinical investigation of the possibility of developing an ALS therapy by targeting the signaling pathways of NMJ dysfunction, provided that early diagnosis of ALS and biomarkers for NMJ dysfunctions become available (**Figure 1**). While keeping in mind the previous failures in clinical trials for ALS, it is evident that multiple mechanisms, including oxidative stress in the center, contribute to ALS pathogenesis. The concept of combination therapy is not novel in the field of neurodegenerative diseases, but it is still a valid approach once most of the key targets of the disease mechanisms, including oxidative stress become identified.

#### **ACKNOWLEDGMENTS**

This project was supported by Photonic program from the Finnish Academy (ROSim Grant 135179). Rashid Giniatullin was supported by the Program of Competitiveness of Kazan University.

#### **REFERENCES**


of amyotrophic lateral sclerosis. *Brain* 131, 3335–3347. doi: 10.1093/brain/ awn243


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; accepted: 27 April 2014; published online: 13 May 2014*.

*Citation: Pollari E, Goldsteins G, Bart G, Koistinaho J and Giniatullin R (2014) The role of oxidative stress in degeneration of the neuromuscular junction in amyotrophic lateral sclerosis. Front. Cell. Neurosci. 8:131. doi: 10.3389/fncel.2014.00131 This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2014 Pollari, Goldsteins, Bart, Koistinaho and Giniatullin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Activation of the endoplasmic reticulum stress response in skeletal muscle of G93A\*SOD1 amyotrophic lateral sclerosis mice

#### Dapeng Chen<sup>1</sup> , Yan Wang<sup>2</sup> and Eva R. Chin<sup>1</sup> \*

<sup>1</sup> School of Public Health, University of Maryland, MD, USA, <sup>2</sup> Proteomics Core Facility, College of Computer, Mathematics and Natural Sciences, University of Maryland, MD, USA

Mutations in Cu/Zn superoxide dismutase (SOD1) are one of the genetic causes of Amyotrophic Lateral Sclerosis (ALS). Although the primary symptom of ALS is muscle weakness, the link between SOD1 mutations, cellular dysfunction and muscle atrophy and weakness is not well understood. The purpose of this study was to characterize cellular markers of ER stress in skeletal muscle across the lifespan of G93A\*SOD1 (ALS-Tg) mice. Muscles were obtained from ALS-Tg and age-matched wild type (WT) mice at 70d (pre-symptomatic), 90d and 120–140d (symptomatic) and analyzed for ER stress markers. In white gastrocnemius (WG) muscle, ER stress sensors PERK and IRE1α were upregulated ∼2-fold at 70d and remained (PERK) or increased further (IRE1α) at 120–140d. Phospho-eIF2α, a downstream target of PERK and an inhibitor of protein translation, was increased by 70d and increased further to 12.9-fold at 120–140d. IRE1α upregulation leads to increased splicing of X-box binding protein 1 (XBP-1) to the XBP-1s isoform. XBP-1s transcript was increased at 90d and 120–140d indicating activation of IRE1α signaling. The ER chaperone/heat shock protein Grp78/BiP was upregulated 2-fold at 70d and 90d and increased to 6.1-fold by 120–140d. The ER-stress-specific apoptotic signaling protein CHOP was upregulated 2-fold at 70d and 90d and increased to 13.3-fold at 120–140d indicating progressive activation of an apoptotic signal in muscle. There was a greater increase in Grp78/BiP and CHOP in WG vs. the more oxidative red gastrocnemius (RG) ALS-Tg at 120–140d indicating greater ER stress and apoptosis in fast glycolytic muscle. These data show that the ER stress response is activated in skeletal muscle of ALS-Tg mice by an early pre-symptomatic age and increases with disease progression. These data suggest a mechanism by which myocellular ER stress leads to reduced protein translation and contributes to muscle atrophy and weakness in ALS.

Keywords: amyotrophic lateral sclerosis, skeletal muscle, endoplasmic reticulum stress, misfolded proteins, muscle atrophy, protein synthesis, unfolded protein response

**Abbreviations:** ALS, amyotrophic lateral sclerosis; SOD1, Cu/Zn superoxide dismutase; ER, endoplasmic reticulum; UPR, unfolded protein response; PERK, protein kinase RNA-activated (PKR)-like ER kinase; IRE1α, inositol-requiring kinase 1-alpha; eIF2α, eukaryotic initiation factor 2 alpha; PDI, protein disulfide isomerase; Grp78/BiP, 78 kDa glucoseregulated protein and immunoglobulin binding protein; CHOP, C/EBP-homologous protein; SDS-PAGE, dodecyl sulfate polyacrylamide gel electrophoresis; WG, white gastrocnemius; RG, red gastrocnemius; DIA, diaphragm; HRT, heart; LIV, liver.

#### Edited by:

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Jari Koistinaho, University of Eastern Finland, Finland Thomas J. Hawke, McMaster University, Canada P. Bryant Chase, The Florida State University, USA

#### \*Correspondence:

Eva R. Chin, School of Public Health, University of Maryland, SPH Bldg. Rm 2134b, College Park, MD 20742, USA erchin@umd.edu

> Received: 31 December 2014 Accepted: 16 April 2015 Published: 18 May 2015

#### Citation:

Chen D, Wang Y and Chin ER (2015) Activation of the endoplasmic reticulum stress response in skeletal muscle of G93A\*SOD1 amyotrophic lateral sclerosis mice. Front. Cell. Neurosci. 9:170. doi: 10.3389/fncel.2015.00170

## Background

Amyotrophic Lateral Sclerosis (ALS) is a fatal motor neuron disease characterized by degeneration of motor neurons and progressive paralysis of skeletal muscle. ALS is inevitably fatal, with patients generally dying due to respiratory failure within 2–5 years of diagnosis (Rowland and Shneider, 2001). Although the majority of ALS cases are sporadic without family history, 5–10% of the total cases of ALS have a known genetic basis (Rowland and Shneider, 2001). Mutations in human Cu/Zn superoxide dismutase 1 (SOD1) account for ∼20% of familial ALS (fALS) cases (Rosen et al., 1993). Mice generated to express a human Cu/Zn SOD1 mutation found in fALS patients (Gly<sup>93</sup> to Ala; G93A) develop a rapidly progressive and fatal motor neuron disease similar to the clinical phenotype of ALS (Gurney et al., 1994). There is evidence that the SOD1 mutations exert their deleterious effects through a ''gain-of-function'' mechanism rather than through a loss of superoxide dismutase activity (Yim et al., 1996). The nature of this toxic ''gain-of-function'' is not known, although a number of putative mechanisms have been proposed, including oxidative stress, glutamate-mediated excitotoxicity, mitochondrial dysfunction, protein aggregation and endoplasmic reticulum (ER) stress (Robberecht and Philips, 2013).

Neurodegenerative diseases, including ALS, that result from unfolded/misfolded proteins have been linked to ER stress (Kaufman, 1999). Most newly synthesized proteins are folded properly in the ER, but unfolded and misfolded proteins accumulate in the ER lumen, causing cellular stress, activation of unfolded protein response (UPR) and an ER stress response (Xu et al., 2005). The ER stress response involves activation of three ER-resident stress sensors: protein kinase RNAactivated-like ER kinase (PERK), inositol-requiring kinase 1 alpha (IRE1α), and activating transcription factor 6 (ATF6; Xu et al., 2005). Normally, these ER stress sensors physically interact with the ER chaperone immunoglobulin binding protein (Grp78/BiP) which suppresses their activation (Xu et al., 2005). However, when unfolded/misfolded proteins accumulate, Grp78/BiP preferentially binds to unfolded/misfolded proteins, resulting in activation of the ER stress response, including an upregulation of genes encoding Grp78/BiP, protein disulfide isomerase (PDI) and down regulation of protein synthesis (Kaufman, 1999; Bertolotti et al., 2000). PERK activation induces the eukaryotic initiation factor 2 alpha subunit (eIF2α) kinase and phosphorylation of eIF2α resulting in inhibition of protein translation (Xu et al., 2005). Activation of IRE1α leads to the alternative splicing of the transcription factor X-box binding protein 1 (XBP1) to the spliced XBP1 form to induce genes that regulate protein quality control in the ER (Xu et al., 2005). Although ER stress is usually a short term homeostatic event essential for cell survival, prolonged and severe ER stress can trigger apoptosis by ER stressspecific cell death signals, including C/EBP homologous protein (CHOP) and caspase-12 (Nakagawa et al., 2000; Kaufman, 2002).

It has previously been shown that mutant SOD1 accumulates inside the ER, where it forms insoluble high molecular weight aggregates and interacts with Grp78/BiP in spinal cord microsomal fractions (Kikuchi et al., 2006). Markers of ER stress activation have been shown in spinal cord sections of ALS patients and in mouse models of ALS (Kikuchi et al., 2006; Atkin et al., 2008). Pathology studies show that ER stress is evident in spinal cords of ALS patients suggesting that ER stressinduced apoptosis may contribute to motor neuron death (Ilieva et al., 2007; Atkin et al., 2008). The ER stress response is also activated in mouse models of ALS, although the time course is controversial (Atkin et al., 2006, 2008; Kikuchi et al., 2006; Ilieva et al., 2007; Nishitoh et al., 2008; Saxena et al., 2009; Wang et al., 2011).

The targets of mutant SOD1-induced toxicity in ALS pathology are the motor neurons and the skeletal muscle that it innervates (Cleveland and Rothstein, 2001). While the primary focus has been on selective defects in the motor neuron causing muscle weakness and atrophy, it has been shown that musclerestricted SOD1 mutations also recapitulate the hallmark signs of ALS, albeit at a slower rate of progression (Dobrowolny et al., 2008; Wong and Martin, 2010). Thus, it has been proposed that defects in skeletal muscle, leading to muscle cell dysfunction, also contribute to the motor neuron pathology via a ''dying-back'' phenomenon. It has previously been shown that early markers of muscle adaptation in the G93A\*SOD1 mouse (i.e., by 49d, prior to atrophy) include metabolic enzymes, particularly down regulation of enzymes of oxidative metabolism and upregulation of enzymes of glycolytic metabolism and increases in proteins involved in protein synthesis (Capitanio et al., 2012). Putative markers of disease progression, which were altered at 98d when atrophy was evident, include glycolytic enzymes which decrease and cell stress markers (i.e., heat shock proteins) and transport proteins (i.e., albumin), which increase.

The intracellular mechanisms leading to altered gene/protein expression in skeletal muscle with disease-induced plasticity are not fully understood. However, there are reports of mitochondrial depolarization leading to reduced mitochondrial Ca2<sup>+</sup> buffering and increased cytosolic Ca2<sup>+</sup> which may trigger events in the muscle atrophy process (Zhou et al., 2010; Yi et al., 2011). Recently we reported impaired intracellular Ca2<sup>+</sup> regulation in the sarcoplasmic reticulum (SR) in muscle fibers from G93A\*SOD1 mice (Chin et al., 2014). The changes intracellular Ca2<sup>+</sup> occurred prior to the decline in motor function (by 90d) and were associated with decreases in myocellular Ca2<sup>+</sup> buffering proteins SERCA1, SERCA2 and parvalbumin. Based on the known association between SR/ER Ca2<sup>+</sup> regulation and protein folding (Glembotski, 2012; Prell et al., 2013), and the putative contribution of skeletal muscle defects to the progression of ALS, we hypothesized that ER stress would be induced in skeletal muscle. Thus, the primary aim of this study was to investigate the ER stress signaling pathway in skeletal muscle at three different ages across the lifespan of the G93A\*SOD1 mouse model of ALS. A secondary aim was to compare key markers of ER stress in skeletal muscles of varying fiber type composition and metabolic capacities as well as to nonmuscle tissue. Our findings indicate that ER stress is activated in skeletal muscle of G93A\*SOD1 mice as early as 70d, with ER stress pathways leading to inhibition of protein translation. Our data further suggest that defects in myocellular protein handling and activation of apoptosis may contribute to the muscle atrophy and weakness observed in ALS.

## Methods

### Ethics Statement

All procedures were conducted under a protocol approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Maryland, College Park.

#### Animals

Control C57BL/6 SJL hybrid female and transgenic ALS B6SJL-Tg (SOD1-G93A)1Gur/J (G93A\*SOD1) male mice were obtained from The Jackson Laboratory. Wild-type control (WT) and transgenic G93A\*SOD1 heterozygote (ALS-Tg) mice were bred to establish a colony at our animal care facility at the University of Maryland. Mice were weaned at postnatal day 21 and genotyped. Male and female ALS-Tg mice along with their wild-type littermates were investigated at a range of ages from the pre-symptomatic to the symptomatic stages of the disease: (i) early pre-symptomatic at postnatal day 70 (70d); (ii) late presymptomatic at postnatal day 90 (90d); and (iii) end stage at postnatal day 120–140 (120–140d) (see Supplementary Material Table S1). Early signs of disease such as muscle tremors can be detected between 65 and 90d but overt muscle weakness and limitations in mobility do not occur until 100–120d (Rosen et al., 1993). We chose the 70d and 90d time points based on differences observed in single muscle fiber resting intracellular Ca2<sup>+</sup> concentration that we now have reported (Chin et al., 2014). The final time point (120–140d) was based on symptom progression, with the date of use determined by the inability of the mouse to right itself after 30 s of being placed on its side as previously described by others (Deforges et al., 2009).

At time of use, animals were euthanized by CO<sup>2</sup> inhalation followed by cervical dislocation. Skeletal muscles, cardiac muscle, and liver were harvested, quickly frozen in liquid nitrogen and stored at −80◦C for subsequent analysis. Various skeletal muscles were harvested in order to assess differences in ER stress between muscles of varying fiber type and of different oxidative and glycolytic capacities. White gastrocnemius (WG) has primarily fast glycolytic fibers (97% type IIB, 1.5% IIX/B and 1.5% IIX), red gastrocnemius (RG) primarily fast oxidative glycolytic fibers (22% type IIB, 3% IIX/B, 20% IIX, 42% IIA and 8% and type I) (Bloemberg and Quadrilatero, 2012) and diaphragm (DIA) has a mixed fiber type including both slow oxidative, fast oxidative and fast glycolytic fibers (39% type IIX, 23% type IIX/B, 23% IIA/X and 10% type I (Guido et al., 2010). Tibialis anterior (TA) has primarily fast glycolytic fibers (50% type IIB, 40% type IIX and 10% IIA), with reports of a fiber type shift to more oxidative (20% IIB, 10% IIX and 70% IIA) at 115d in the G93A\*SOD1 mouse (Deforges et al., 2009). In WT mouse muscle, glycolytic capacity is greatest in type IIB > IIB/X > IIX = type I > IIA/X > IIA (Bloemberg and Quadrilatero, 2012) and thus expected to be highest in WG > TA > DIA > RG. Conversely, in WT, oxidative capacity is greatest in type IIA >

#### Protein Extraction

The superficial (white) and deep (red) gastrocnemius, DIA, cardiac muscle, and liver tissues were used for assessment of protein levels using western blot technique. Tissue samples were homogenized on ice using a polytron at 50% maximum power for three 10 s bursts, separated by 30 s in ice cold lysis buffer (20 mM Hepes, pH = 7.5, 150 mM NaCl, 1.5 mM MgCl2, 0.1% Triton X-100, 20% Glycerol) containing 1 mM DTT and protease inhibitor cocktail (cOmplete mini EDTA-free Protease Inhibitor Cocktail, Roche). After 20 min of incubation at 4◦C followed by centrifugation for 5 min at 20,000× g, the supernatant was collected, quick frozen in liquid nitrogen and stored at −80◦C until required.

### Western Blot Analyses

Total protein concentration in the samples was determined using a BCA protein assay kit (Thermo Scientific). Samples were then prepared with loading buffer and denatured by incubating samples at 100◦C for 5 min. For western blot analyses, 30 µg total protein was loaded on bis-acrylamide gels and separated using polyacrylamide gel electrophoresis (PAGE). Samples were then transferred to PVDF membrane (Millipore) and blocked with 5% (w/v) non-fat dry milk in Tris-buffered saline (pH 8.0) for 1 h. The appropriate primary antibodies were added: Grp78/BiP (BD Biosciences) (PERK, phospho-PERK (Thr980), IRE1α, eIF2α, phospho-eIF2α (Ser51), PDI, and CHOP; 1:1000, Cell Signaling Technology) and membranes were incubated at 4◦C overnight, washed and then and subsequently probed with HRP-linked anti-rabbit IgG or anti-mouse IgG antibodies (1:1000, Cell Signaling Technology) 1 h at room temperature. Secondary antibodies were detected using HRP-linked chemiluminescence with SuperSignal West Dura Chemiluminescence Substrate (Thermo Scientific) and imaged using the chemiluminescence imaging system (GeneGnome, Syngene). The signal for the target protein of each sample was quantified using densitometry (Image J Software) and expressed in arbitrary unit (AU). GAPDH (1:2000, Thermo Scientific) or β-actin (1:1000, Cell Signaling Technology) was used to confirm equal protein loading across samples.

### Mass Spectrometry based Protein Relative Quantification

To confirm differential expression of ER stress proteins using a non-antibody based method, we completed in-gel digestion of proteins in the molecular weight range of the Grp78/BiP protein. We focused on the Grp78/BiP protein based on preliminary work with an antibody that gave us a divergent response (Chen et al., 2012) to the one we report here. Skeletal muscle total protein samples were prepared as described previously. For the purpose of protein separation, 30 µg of total protein was loaded onto 8% one-dimensional SDS-PAGE gel. After gel electrophoresis, protein bands were stained using a protein blue stain kit (Thermo Scientific). The targeted bands (∼80 kDa) were carefully excised and in-gel tryptic digestion carried out following standard procedure. Briefly, proteins were reduced with 5 mM DTT, alkylated with 55 mM iodoacetamide, and digested with 20 ng/µL trypsin (Life Technologies™) at 37◦C overnight. All reagents were dissolved in 50 mM ammonium bicarbonate (pH 8.5).

After trypsin digestion, peptide products were collected and analyzed by nano LC-MS/MS analysis using LTQ Orbitrap mass spectrometer coupled to a Shimadzu 2D Nano HPLC system. Peptides were loaded with an autosampler into an Zorbax SB-C18 trap column (0.3 × 5.0 mm) (Agilent Technologies, Palo Alto, CA, USA) at 10 µL/min with solvent A (97.5% water, 2.5% ACN, 0.1% formic acid) for 10 min, then eluted and separated at 300 nL/min with a gradient of 0–35% solvent B (2.5% water, 97.5% ACN, 0.1% formic acid) in 30 min using a Zorbax SB-C18 nano column (0.075 × 150 mm). The mass spectrometer was set to acquire a full scan at resolution 60,000 (m/z 400) followed by data dependent MSMS analysis of top 10 peaks with more than one charge in the linear ion trap at unit mass resolution. The resulting LC-MS/MS data were searched against a mouse protein database generated from uniprot and a common contaminant database using Mascot (v2.3) and Sequest search engines through Proteome Discoverer (v1.4). Carbomidomethylation at Cys was set as fixed modification. Methionine oxidation and asparagine and glutamine deamidation were set as variable modification. Spectral counting with normalized total spectra was carried out using Scaffold software, (Proteome Software, Inc). Protein probability >99% and at least one unique peptide with a probability score >95% were set to as minimum requirement for protein identification.

#### Gene Expression

In order to investigate transcriptional events involved in ER stress pathway, we isolated mRNA from TA muscle (TA) and examined transcript levels of XBP-1, GRP78/BiP, and CHOP. Briefly, total RNA was isolated using TriPure Reagent (Roche) and RNA content was determined by using a NanoDrop spectrophotometer and mRNA was diluted to 5 ng/µL. Reverse transcription from mRNA to cDNA was conducted by using One-Step RT-PCR System (Life Technologies™). Semi-quantitative PCR (sqPCR) was used to determine gene transcriptional levels and the primer information such as XBP-1, CHOP, and 18 s was acquired from a previous study (Rosen et al., 1993). The band intensity of PCR products were quantified using densitometry (Image J Software) and expressed in arbitrary unit (AU). For XBP-1, two variants of XBP1 mRNA are expressed in cells. Under normal conditions, un-spliced XBP1 mRNA (XBP1 u) is expressed. However, as ER stress is induced, a spliced form of XBP1 mRNA (XPB1-s) will be expressed. Thus, upregulation of XBP1-s mRNA is a marker of ER stress activation downstream of IRE1α (Wu et al., 2011).

(120–140d; n = 3 each for WT and ALS-Tg) mice. (B) Protein samples were incubated either with or without PERK peptides and then PERK protein levels were detected using western blots technique. (C) Analysis of average arbitrary units (AU) obtained by densitometry for PERK. (D) Analysis of average ratio of phospho-PERK to total PERK. Data in (C) and (D) are presented as mean ± S.E; \*,p < 0.05; \*\*,p < 0.01 WT vs. ALS-Tg in the same age group. #p < 0.05 vs. 70d.

#### Data Analysis

To determine statistical differences in protein and mRNA expression level between genotype (WT vs. ALS-Tg) and Age (70d, 90d and 120–140d) data were analyzed using two-way ANOVAs. Where interaction effects (genotype × age) were observed, the main effects are not indicated in the Results section but can be found in the Supplementary material (Table S1). For significant interaction effects, Tukey post hoc tests were used to determine differences across time points for ALS-Tg (i.e., 70d vs. 90d). T-tests were used to determine differences between WT and ALS-Tg at each time point. Statistical significance was accepted as p < 0.05.

### Results

#### ER Stress Pathway is Induced in Skeletal Muscle of ALS Mice

PERK and IRE1α are two ER stress sensors known to be upregulated when ER stress is induced. In our study, PERK protein level was upregulated 2.6-fold in WG muscle of ALS-Tg vs. WT mice at 70d (p = 0.01), 5.4-fold at 90d (p = 0.025) and 5.2-fold at 120–140d (p = 0.001; **Figures 1A,C**). There was no difference in PERK level of ALS-Tg WG between 70, 90 and 120–140d (no main effect for age or genotype × age interaction). To assess the specificity of the antibody for total PERK, the antibody was pre-incubated with a PERK-antibody blocking peptide. Under these conditions, the protein band at ∼140kDa identified as PERK was not visible (**Figure 1B**), confirming the specificity of the PERK antibody. Since activated PERK undergoes auto-phosphorylation, we also assessed the ratio of phospho-PERK to total PERK. The phospho-PERK/total PERK ratio was increased 2-fold in WG of ALS-Tg mice at 120–140d (p = 0.012) indicating greater activation of existing PERK protein at the symptomatic age (**Figures 1A,D**). Previous studies have shown that PERK can activate eIF2α kinase, resulting in phosphorylation of eIF2α at Ser51 and suppression of protein synthesis during ER stress (Wu et al., 2011). We therefore assessed the downstream effects of activation of PERK. In WG muscle, the ratio of phospho-eIF2α/total eIF2α was increased 2.3-fold in ALS-Tg vs. WT at 70d (p = 0.005), remained elevated at 90d (p = 0.048) and increased further to 12-fold at 120–140d (p = 0.011; **Figures 2A,B**). For phospho-eIF2α/total eIF2α there was a genotype × age interaction effects with the increase in ALS-Tg only increasing between 70d and 120–140d (70d vs. 90d; p = 0.227; 70d vs. 120–140d p = 0.018; 90d vs. 120–140 p = 0.062). Total eIF2α was not altered, just phosphorylation at Ser51, indicating inhibition of protein translation at these ages.

In addition to PERK, up-regulation of ER stress sensor IRE1α was also observed in WG of ALS-Tg mice by 70d. IRE1α protein levels were increased 2.5-fold at 70d (p = 0.043) remained elevated at 90d (p = 0.0005) and showed a further increase to 4.9-fold WT levels at 120–140d (p = 0.0002 vs. WT by ttest; p = 0.008 for 70d vs. 120–140d for ALS-Tg by Tukey post hoc; **Figures 3A,B**). XBP1 mRNA splicing is commonly used to indicate upregulation of IRE1α since activation of

IRE1α leads to mRNA splicing. Thus, we investigated transcript levels of the un-spliced (XBP-1u) and spliced XBP1 (XBP-1s) forms of XBP-1. In TA muscle, XBP-1s mRNA was increased to 1.3- and 1.4-fold at 90d (p = 0.001) and 120–140d (p < 0.0001). There was a genotype × age interaction effect with XBP1-s being higher at 90d vs. 70d (p = 0.002) and at 120–140d vs. 70d (p = 0.002). XBP1-u mRNA was not altered (**Figures 3C,D**).

symptomatic (120–140d; n = 3 each for WT and ALS-Tg) mice. (B) Analysis of ratio of phospho-eIF2α to total eIF2α. Data in (B) is presented as mean ± S.E;

\*,p < 0.05; \*\*,p < 0.01 WT vs. ALS-Tg in the same age group.

Since cellular stress results in induction of ER chaperone proteins to handle misfolded and unfolded proteins, we examined changes in PDI and Grp78/BiP protein levels, two proteins involved in post-translational modification and known to be up-regulated with ER stress activation (Xu et al., 2005). Grp78/BiP was upregulated 2-fold in WG of ALS-Tg vs. WT mice at 70d (p = 0.006), remained elevated at 90d (p = 0.025) and increased further to 6.7-fold at 120–140d (p = 0.005) (**Figures 4A,B**). There was a genotype × age interaction effect with Grp78/BiP in ALS-Tg being different at 70d vs. 120–140d (p < 0.001) and at 90d vs. 120–140d (p < 0.001). Increased expression of PDI was also observed (2.2-fold but only at 120–140d (p = 0.001; **Figures 4A,C**). There was a genotype × age interaction effect for PDI with differences across all age groups for ALS-Tg (70d vs. 90d, p = 0.004; 70d vs. 120–140d, p = 0.001; 90d vs. 120–140d, p < 0.001). Taken together, these data show evidence of ER stress in skeletal muscle as early as 70d and further augmented at the symptomatic age in ALS-Tg mice.

To confirm antibody-based findings of the Grp78/BiP increase in skeletal muscle of ALS-Tg mice, we carried out

FIGURE 3 | IRE1α protein level and spliced Xbp-1 transcriptional level are up-regulated in skeletal muscle of G93A\*SOD1 ALS-Tg mice. (A) Protein was isolated from WG muscle of different ages of wild-type (WT) and transgenic G93A\*SOD1 (ALS-Tg) mice and western blot was performed by using antibody specific for IRE1α. GAPDH was used as the total protein loading control. Three postnatal ages were examined as follows: early pre-symptomatic (70d; n = 3 each for WT and ALS-Tg), late pre-symptomatic (90d; n = 5 each for WT and ALS-Tg), and symptomatic (120–140d; n = 3 each for WT and ALS-Tg) mice. (B) Analysis of average arbitrary

units (AU) obtained by densitometry for IRE1α. (C) Total mRNA was isolated by using Tibialis anterior (TA) muscle tissues and un-spliced (Xbp-1u) and spliced Xbp-1 (Xbp-1s) transcriptional levels were determined by using semi-quantitative PCR and PCR results are shown by running the products in 1.5% agarose gel. 18S was used as the internal control. Three different ages animals as mentioned previously were used. (D) Analysis of average arbitrary units (AU) obtained by densitometry for Xbp-1s. Data in (B,D) are presented as mean ± S.E; \*,p < 0.05; \*\*,p < 0.01 WT vs. ALS-Tg in the same age group. #p < 0.05 vs. 70d.

and PDI are shown. Three postnatal ages were examined as follows: early

\*\*,p < 0.01 WT vs. ALS-Tg in the same age group. #p < 0.05 vs. 70d and †p < 0.05 vs. 90d.

label-free spectral counting-based mass spectrometry. (A) Workflow of label-free spectral counting-based protein quantitative analysis using LC-MS/MS. Protein samples were separated using SDS-polyacrylamide gel electrophoresis (PAGE) and gel pieces excised at ∼80 kDa for the purpose of in-gel trypsin digestion and LC-MS/MS analysis. Protein quantitative data analysis was conducted using spectral counting and interpreted by normalized total spectra numbers. (B) Grp78/BiP protein identification and peptide coverage using LC-MS/MS. (C) Representative mass-to-charge ratio spectrum and b-/y ions fragmentation of Grp78/BiP peptide (highlighted in B). (D) Protein quantitative data analysis using the ratio of normalized total spectra numbers of Grp78/BiP to a house keeping protein GAPDH. Three independent muscles from 120–140d old wild type (WT) and ALS-Tg mice were analyzed. \* p < 0.05, WT vs. ALS-Tg.

ALS-Tg in the same age group. #p < 0.05 vs. 70d and †p < 0.05

vs. 90d.

pre-symptomatic (70d; n = 3 each for WT and ALS-Tg), late pre-symptomatic (90d; n = 5 each for WT and ALS-Tg), and relative protein quantification using LCMSMS and spectral counting compare Grp78/BiP protein levels between genotypes (**Figure 5A**). Grp78/BiP protein was identified by 11 exclusive unique spectra which contributed to the identification of 10 exclusive unique Grp78/BiP peptides (**Figures 5B,C**). Protein quantitative data analysis showed that Grp78/BiP was more abundant in skeletal muscle of ALS mice as spectral counting numbers were significantly higher in ALS-Tg vs. WT mice (**Figure 5D**). Collectively, our label-free spectral counting-based protein quantitative data is consistent with western blot data, supporting our notion that ER stress is activated in skeletal muscle of ALS mice.

#### ER Stress-Specific Cell Death Signal is Induced in Skeletal Muscle of ALS Mice

Several mechanisms have been suggested to link the ER stress pathway to cell death, including activation of the ER stressspecific cell death signal CHOP (Xu et al., 2005). In WG of ALS-Tg mice, CHOP was upregulated 1.8-fold at 70d (p = 0.041), remained elevated at 90d (p = 0.025) and further increased to 12-fold at 120–140d (p = 0.019; **Figures 6A,B**). There was a significant genotype × age interaction effects with ALS-Tg only being different at 70d vs. 120–140d (p < 0.001) and at 90d vs. 120–140d (p < 0.001). There were no changes in CHOP mRNA (data not shown), indicating that there is post-translational modification and increased stability of CHOP protein (Ohoka et al., 2007). In addition to evaluating CHOP induction in the limb muscle, we also investigated DIA muscle since atrophy of this muscle results in respiratory failure and death in ALS mice (Tankersley et al., 2007). In DIA, CHOP protein expression was increased 1.7-fold in ALS-Tg vs. WT mice at 70d (p = 0.008), remained elevated at 90d (p = 0.001) and then increased further to 9-fold at 120–140d but due to the high degree of variability in CHOP elevation, ALS-Tg vs. WT was not significant at 120–140d (p = 0.10), there was a genotype × age interaction effect with CHOP showing an increase in ALS-Tg at 120–140d vs. 70d (p = 0.004) and 120–140d vs. 90d (p = 0.005) (**Figures 6C,D**).

#### Greater Activation of ER Stress Pathway in Glycolytic vs. Oxidative Muscle of ALS Mice

On dissection we noted that the superficial gastrocnemius muscle was more red than white in the ALS-Tg mice (**Figure 7A**). Previous studies showed that fast type IIb motor units are affected first during ALS disease progression in the G93A\*SOD1 mouse (Hegedus et al., 2009). Thus, we wanted to determine whether ER stress is activated to a greater extent in fast glycolytic vs. fast oxidative skeletal muscle by comparing two typical ER stress markers Grp78/BiP and CHOP between WG (fast glycolytic) and RG (fast oxidative) muscle. At the symptomatic age (120–140d), both Grp78/BiP and CHOP were induced to greater extent in WG vs. RG (2.0- and 5.6-fold, respectively; p < 0.05; **Figures 7B,C**), indicating that ER stress activation is greater in fast glycolytic muscle.

#### ER Stress Markers are not Induced in Cardiac Muscle and Liver Tissues of ALS Mice

ER stress is activated when misfolded proteins accumulate in the ER lumen. One could argue that the ER stress activation

we observed in our study may be a non-skeletal muscle specific event since the animal model we used is a whole-body SOD1 protein mutation and accumulation of mutant SOD1 could activate ER stress in all tissues, including skeletal muscle. Thus, we investigated the ER stress pathway in non-pathological tissues such as cardiac muscle and liver. Two classical ER stress markers, Grp78/BiP and CHOP, were not different between WT and ALS-Tg mice for heart or liver at any age (see **Figure 8**). Therefore, ER stress activation is observed in skeletal but not cardiac muscle or other highly oxidative tissues like liver in ALS mice.

### Discussion

In this study we show that ER stress is activated in skeletal muscle of G93A\*SOD1 mice and thus may play a role in muscle atrophy in ALS. This is based on evidence that: (i) ER stress is activated at 70d, an early pre-symptomatic age and is further upregulated at 120–140d, an age when mice are symptomatic; (ii) skeletal muscle ER stress induces the cell death signal CHOP; (iii) ER stress is activated to a greater extent in highly glycolytic muscles with primarily type IIb fibers which are affected by an early loss of fast fatigable motor axons; and (iv) the ER stress activation is specific to skeletal vs. cardiac muscle. These data support the hypothesis that ER stress plays a role in muscle atrophy in ALS mice.

#### ER Stress Response in ALS

Our lab previously reported impairments in SR Ca2<sup>+</sup> uptake, leading to elevations in resting cytosolic Ca2<sup>+</sup> concentration

in muscle fibers from G93A\*SOD1 mice (Chin et al., 2014). We therefore hypothesized that altered intracellular Ca2<sup>+</sup> in association with increased oxidative stress in muscle cells would lead to misfolded proteins and activate the UPR and ER stress responses. We propose a model (**Figure 9**) where age-dependent activation of ER stress sensors PERK and IRE1α in response to misfolded proteins leads to increases in protein chaperones Grp78/BiP and PDI in skeletal muscle. Prolonged oxidative stress (in this case due to mutant SOD1) and persistent accumulation of misfolded proteins further augments the ER stress response and activates the apoptotic signal CHOP leading to muscle atrophy. At 70d, an age where muscle grip function is still 100% of WT levels (Chin et al., 2014), there is already an ∼2-fold increase in the ER stress markers PERK, IRE1α, Grp78/BiP and CHOP. By 84d grip function is reduced to 77% WT levels and by 120–140d, when grip function is zero (i.e., mice are no longer able to grasp a metal grid) (Chin et al., 2014), there are further increases in IRE1α, Grp78/BiP and CHOP as well as increased phospho-PERK and phospho-eIF2α/total eIF2α. Based on these data we cannot determine if impaired Ca2<sup>+</sup> regulation and ER stress are causative or a consequence of muscle atrophy, but only that they are associated. Using the same animal model, studies using MRI to asses muscle volume show significant muscle loss as early as 8 weeks of age and continuous loss over the lifespan of these mice (Marcuzzo et al., 2011; Mead et al., 2011). Collectively these data show muscle atrophy and weakness over the time frame that we observed increases in markers of ER stress, suggestive of some involvement of protein misfolding, activation of the ER stress response and possibly apoptosis. The impairment in muscle protein translation as

well as apoptosis would contribute to muscle cell atrophy and weakness which, in conjunction with motoneuron degeneration, would contribute to the pathophysiology and disease progression in ALS.

Proteins that require folding and post-translational modification, primarily secretory and membrane bound proteins, are processed in the ER. Proteins that do not fold properly are degraded by the ubiquitin proteasome pathway. If misfolded/unfolded proteins accumulate, the UPR is triggered to promote proper folding and autophagy is activated to support cell survival. However, if there is a persistent increase in misfolded/unfolded proteins that exceeds the capacity of the ER to regulate proper folding, then protein translation is inhibited to reduce protein load and there is activation of an ER stress-induced cell death pathway via upregulation of CHOP. Activation of the ER stress response is increasingly being recognized as a cellular mechanism in neurodegenerative as well as metabolic diseases such as diabetes (Ozawa et al., 2005; Chambers and Marciniak, 2014). The biological role of ER stress activation in motor neurons has been shown in one study in which deletion of an ER stress-induced pro-apoptotic signaling protein (puma) in ALS mice resulted in improved motor neuron survival and delayed disease onset and motor dysfunction (Kieran et al., 2007). Although our study did not directly investigate the cellular consequences of ER stress activation in skeletal muscle of ALS mice, it is the first study to show the associated changes in ER stress markers in skeletal muscles across the lifespan of the G93A\*SOD1 mouse.

Other studies with both ALS patients and transgenic G93A\*SOD1 mice support the role of ER stress-related cellular dysfunction in ALS pathophysiology (Atkin et al., 2006, 2008; Kikuchi et al., 2006; Ilieva et al., 2007; Nishitoh et al., 2008; Saxena et al., 2009; Wang et al., 2011). Studies examining lumbar spinal cord sections of G93A\*SOD1 mice, demonstrated significant upregulation of three ER stress sensors, PERK, IRE1α and ATF-6 as early as 60d, a time point at which these mice do not show any symptoms, suggesting they may trigger disease pathophysiology (Atkin et al., 2008). In addition, the ER stress-specific cell death markers, CHOP and caspase-12, were activated, indicating apoptosis was induced before symptom onset (Atkin et al., 2008). ATF6, IRE1α, CHOP, and caspase-12 were also shown to be up-regulated in the lumbar spinal cord sections of G93A<sup>∗</sup> SOD1 mice at 90d, at a late pre-symptomatic stage of the disease (Atkin et al., 2006). In yet other studies, a broad ER stress response occurred in spinal cords only in the end stage of ALS (∼140d) when the mice were symptomatic which does not support the role of ER stress in ALS pathology (Kikuchi et al., 2006). Levels of ER stress-related proteins including PERK, IRE1α, ATF6, XBP-1, Grp78/BiP, and CHOP were upregulated in motor neurons of the spinal cords of ALS patients (Atkin et al., 2008). However, evidence from these studies is based on the association of ER stress markers with disease presence.

#### Impaired Protein Translation in Muscle Atrophy and ALS

Motor neurons and skeletal muscles are two targets of mutant SOD1-mediated toxicity in ALS. In skeletal muscle cells, ER stress leading to inhibition of protein translation and decreased protein synthesis could explain the reduction in muscle mass of the G93A\*SOD1 mouse (Gurney et al., 1994; Marcuzzo et al., 2011; Mead et al., 2011). We observed increases in PERK protein-kinase activity and increased phosphorylation of eIF2α on serine residue 51. Increased phospho-eIF2α has been shown to inhibit translation of messenger RNA into protein, effectively decreasing the protein load (Harding et al., 1999; Scheuner et al., 2001). Inhibition of protein translation can alter both muscle protein synthesis during growth but also muscle plasticity in disease. We recently reported alterations in intracellular Ca2<sup>+</sup> levels in skeletal muscle of the G93A\*SOD1 mice and reductions in the Ca2<sup>+</sup> buffering proteins SERCA1, SERCA2 and parvalbumin (Chin et al., 2014). SERCA1 is the SR/ER Ca2<sup>+</sup> ATPase isoform expressed in fast glycolytic fibers and SERCA2 is expressed in fast oxidative and slow fibers (Periasamy and Kalyanasundaram, 2007). Due to the shift to more oxidative fibers in WG (Hegedus et al., 2008) and TA (Deforges et al., 2009), we expected an increase in SERCA2 protein expression. However, SERCA2 protein was also decreased in skeletal muscle of ALS-Tg mice (Chin et al., 2014) despite a compensatory upregulation of SERCA2 mRNA (unpublished data). Thus, at least for SERCA2, we have observed transcriptional upregulation with no concurrent increase in protein translation. Our current data showing increased phospho-eIF2α/total eIF2α are consistent with this inhibition of protein translation as early as 70d. While there is an overall decrease in protein translation, this will specifically affect the secretory and transmembrane proteins that are processed in the ER (Chambers and Marciniak, 2014) such as the SERCA pump proteins. However, the muscle is still capable of increased protein synthesis and can upregulate expression of the required ER stress proteins. Future studies will be required to identify other muscle-specific proteins that are misfolded and that may contribute to atrophy and/or disease pathophysiology in ALS.

#### ER Stress, Muscle Metabolic Capacity and Disease Pathology

There are limited reports about ER stress pathways in skeletal muscles of motoneuron disease, although there are clinical reports of ER stress and UPR being activated in inclusion body myositis (Vattemi et al., 2004; Nogalska et al., 2006), autoimmune myositis (Nagaraju et al., 2005; Vitadello et al., 2010), and myotonic dystrophy (Ikezoe et al., 2007). It has also been shown that ER stress response proteins IRE1α, PDI, and other ER chaperones are upregulated in skeletal muscle of mice in response to high fat diet feeding which lead to a decrease in protein synthesis and insulin resistance (Deldicque et al., 2010). ER stress sensors and ER chaperones have also been shown to be upregulated in skeletal muscle following exercise (Kim et al., 2011; Wu et al., 2011). Our study is consistent with these prior reports of ER stress response in diseases and under conditions of various cellular stresses such as altered metabolic substrate supply and changes in metabolic activity. Interestingly, our observation that ER stress was greater in WG compared to RG is consistent with the notion that an energetic stress contributes to activation of the ER stress response as glycolytic fibers are more likely to experience periods of hypoxia or perturbed energy homeostasis.

Previous studies have shown that glycolytic skeletal muscle is more susceptible to atrophy in response to hypoxia (de Theije et al., 2015) and in disease states such as cancer cachexia (Yu et al., 2008; Baltgalvis et al., 2009) and heart failure (Li et al., 2007). The DIA, a muscle of mixed fiber type with high glycolytic capacity, had levels of ER stress comparable to the glycolytic WG. It thus appears that metabolic capacity rather than muscle fiber type per se (i.e., fast vs. slow contracting) is a crucial determinant of susceptibility to ER stress in skeletal muscle cells. Muscles with a higher oxidative capacity have an increased capacity to respond to oxidative stress (Yu et al., 2008) and induce heat shock proteins (i.e., hsp70) (Locke and Tanguay, 1996; Tarricone et al., 2008). This may render protection and provide a greater capacity to refold proteins in highly oxidative muscle fibers. We speculate that skeletal muscles with high glycolytic capacity and those with high repetitive use (i.e., DIA) are most susceptible to ER stress due to the greater metabolic stress with contractile activity and a reduced capacity to deal with misfolded proteins.

#### Potential Limitations

In the current study we have assessed markers of ER stress using commercially available antibodies. The specificity and selectivity of antibodies for the protein ligand are critical to the interpretation of data from western blot analyses. We have confirmed the specificity of some of these protein markers using either a blocking peptide (PERK) or mass spectrometry (Grp78/BiP). Supporting our data, a recent study using proteomics and bioinformatics tools reported activation of stress responses in gastrocnemius muscle of ALS-Tg mice at 98d including Alpha-crystallin B chain (Cryab), Heat shock protein HSP 90-beta (Hsp90ab1) and protein disulfide-isomerase A3 (Pdia3) that suggest abnormalities in the ER protein folding machinery and activation of the UPR (Capitanio et al., 2012). Our observations of increased expression of proteins involved in UPR and ER stress are consistent with this report and illustrate that two independent techniques have been used to confirm the differential expression of ER stress protein markers in skeletal muscle.

#### Mutant SOD1, Mitochondrial Disruption and ER/SR Dysfunction

Previous studies indicated that mutant SOD1 is present within the ER lumen which may account for activation of ER stress in SOD1-linked ALS cases (Karch et al., 2009). Mutant SOD1 also accumulates in mitochondria specifically in motoneurons (vs. sensory neurons) (Sotelo-Silveira et al., 2009). Based on the ubiquitous expression of SOD1 (Hirano, 1991) one would expect activation of the UPR and ER stress pathway from mutant SOD1 in all tissues. However, our results of two typical ER stress markers, Grp78/BiP and CHOP, showed that ER stress is not present in cardiac muscles or liver tissues. We speculate that this is due to the high oxidative capacity, and increased mitochondrial content in these tissues. Motoneurons and glycolytic skeletal muscle, however, have low mitochondrial content and thus mitochondrial impairment will be detrimental to survival of these cells. Decreased mitochondrial inner membrane potential and mitochondrial Ca2<sup>+</sup> buffering have been observed in response to osmotic stress and plasma membrane depolarization in muscle fibers of G93A\*SOD1 mice (Yi et al., 2011). The same group has shown that reduced mitochondrial membrane potential is greatest in muscle fibers at the neuromuscular junction region (Zhou et al., 2010). Thus, changes in mitochondrial function leading to elevated cellular Ca2<sup>+</sup> are specifically impaired at the skeletal muscle membrane nearest to its point of innervation. This may play an important role in the axonopathy associated with ALS (Fischer et al., 2004; Wong and Martin, 2010).

Vulnerable motoneurons also have reduced cytosolic Ca2<sup>+</sup> buffering capacity and disrupted mitochondrial and ER Ca2<sup>+</sup> buffering. Motoneurons with mutant (G93A) SOD1 had reduced mitochondrial membrane potential, lower ER Ca2<sup>+</sup> release in response to a SERCA inhibitor and impaired mitochondrial Ca2<sup>+</sup> buffering (Jaiswal and Keller, 2009; Jaiswal et al., 2009). Thus, impairment of mitochondrial and ER Ca2<sup>+</sup> storage is thought to be central to motoneuron degenerative changes. Others have shown that mutant SOD1 aggregates with voltage dependent anion channels in mitochondria, specifically of motoneurons (Israelson et al., 2010). In the latter study, ADP transport but not Ca2<sup>+</sup> uptake by motoneuron mitochondria was diminished. There is emerging data on the importance of Ca2<sup>+</sup> regulation, mitochondrial function and ER protein misfolding in various diseases (Jaiswal, 2013; Prell et al., 2013), but the tissue-specific mechanisms altered leading to muscle atrophy and weakness are still not clear. Further, the role of myocyte health in axonal survival and the putative retrograde cellular pathophysiology remains unresolved. Future studies will be required to determine the precise mechanisms by which proteins that regulate motoneuron and skeletal muscle interaction are affected by unfolded proteins and ER stress leading to muscle atrophy and weakness.

### Conclusion

In summary, our findings show that ER stress is present in skeletal muscle of transgenic ALS mice starting at an early, pre-symptomatic age. The ER stress sensors (PERK, IRE1α), ER chaperones (Grp78/BiP, PDI), and ER stress-induced apoptotic mediator (CHOP) are activated and associated with disease pathophysiology and inhibition of protein translation in skeletal muscle. Additionally, we show that ER stress activation is greatest in glycolytic skeletal muscle and muscles of highest contractile activity demand. These data suggest that ER stress induces an early cellular pathology in skeletal muscle that may contribute to the atrophy in ALS.

### Authors Contribution

ERC designed the study, established the G93A\*SOD1 animal colony and completed animal dissections. DC assisted

#### References


in maintaining colonies and completed muscle tissue analyses. YW advised on muscle sample preparation for mass spectrometry, completed mass spectrometry analysis and interpretation of MS data. DC, YW and ERC made figures, analyzed the data, and interpreted the findings. DC and ERC wrote the manuscript. DC, YW and ERC edited the manuscript. All authors read and approved the final manuscript.

### Acknowledgments

We thank Brittany Jacobs, Davi A.G. Mazala and Samuel A. English for their contributions to data collection and analyses. The authors also thank Dr. Andrew T. Ludlow for his critical reading of the manuscript. This work was supported by the University of Maryland, College Park new investigator funds to ERC. Funding for Open Access provided by the UMD Libraries Open Access Publishing Fund.

### Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel.2015. 00170/abstract

of amyotrophic lateral sclerosis. Am. J. Physiol. Cell Physiol. 307, C1031–C1038. doi: 10.1152/ajpcell.00237.2013


Cu, Zn superoxide dismutase mutation. Science 264, 1772–1775. doi: 10. 1126/science.266.5190.1586-a


a longitudinal MRI study. Exp. Neurol. 231, 30–37. doi: 10.1016/j.expneurol. 2011.05.007


**Conflict of Interest Statement**: Eva R. Chin and Dapeng Chen are Inventors on a pending patent which includes some of these data. Eva R. Chin is the Founder and Chief Scientific Officer of MyoTherapeutics, a University of Maryland-based startup company. Yan Wang declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Chen, Wang and Chin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Dysfunctional mitochondrial Ca2<sup>+</sup> handling in mutant SOD1 mouse models of fALS: integration of findings from motor neuron somata and motor terminals

#### *Ellen F. Barrett\*, John N. Barrett and Gavriel David*

*Department of Physiology and Biophysics, and Neuroscience Program, University of Miami Miller School of Medicine, Miami, FL, USA*

#### *Edited by:*

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### *Reviewed by:*

*Paolo Bernardi, University of Padova, Italy Eveliina Pollari, KU Leuven, Belgium Gundars Goldsteins, University of Eastern Finland, Finland*

#### *\*Correspondence:*

*Ellen F. Barrett, Department of Physiology and Biophysics, University of Miami Miller School of Medicine, PO Box 016430, Miami, FL 33101, USA e-mail: ebarrett2@med.miami.edu*

Abundant evidence indicates that mitochondrial dysfunction and Ca2<sup>+</sup> dysregulation contribute to the muscle denervation and motor neuron death that occur in mouse models of familial amyotrophic lateral sclerosis (fALS). This perspective considers measurements of mitochondrial function and Ca2<sup>+</sup> handling made in both motor neuron somata and motor nerve terminals of SOD1-G93A mice at different disease stages. These complementary studies are integrated into a model of how mitochondrial dysfunction disrupts handling of stimulation-induced Ca2<sup>+</sup> loads in presymptomatic and end-stages of this disease. Also considered are possible mechanisms underlying the findings that some treatments that preserve motor neuron somata fail to postpone degeneration of motor axons and terminals.

**Keywords: mitochondria, motor neuron, motor nerve terminal, Ca2<sup>+</sup> regulation, mutant SOD1 models of fALS**

#### **MITOCHONDRIA HELP BUFFER LARGE STIMULATION-INDUCED Ca2<sup>+</sup> LOADS IN MOTOR NEURONS**

Mitochondria are important for temporary buffering of large Ca2<sup>+</sup> loads in both the soma and peripheral motor terminals of motor neurons. Whole cell recordings using fluorescent Ca2<sup>+</sup> indicators demonstrate that mitochondria normally buffer about 50% of large Ca2<sup>+</sup> loads in hypoglossal motor neurons (Fuchs et al., 2013). Imaging studies in motor axons injected with these indicators demonstrate that during repetitive action potential stimulation (e.g., 50 Hz) mitochondrial uptake is also the dominant mode of Ca2<sup>+</sup> sequestration in motor nerve terminals (reviewed in Barrett et al., 2011). Consistent with this idea, stimulation increases [Ca2+] in the matrix of motor terminal mitochondria (David et al., 2003; Vila et al., 2003). Mitochondrial Ca2<sup>+</sup> uptake is functionally important, because when this uptake is blocked, phasic evoked transmitter release depresses rapidly during repetitive stimulation, likely due to depletion of the vesicle pool by excessive asynchronous transmitter release caused by the increased cytosolic [Ca2+] (David and Barrett, 2003).

Mitochondrial Ca2<sup>+</sup> uptake is especially important in motor neurons that innervate fast, fatiguable (FF) muscles, because these motor neurons tend to have low levels of cytoplasmic Ca2<sup>+</sup> buffers (e.g., parvalbumin, calbindin, reviewed by Grosskreutz et al., 2010), and tend to discharge in high frequency bursts (Burke, 2004) that deliver large Ca2<sup>+</sup> loads to motor terminals. Motor neurons that innervate limb FF muscles are especially vulnerable in transgenic mice expressing fALS-associated mutations of superoxide dismutase 1 (SOD1) (reviewed by Gordon et al., 2004; Murray et al., 2010).

Effective sustained buffering of Ca2<sup>+</sup> by mitochondria requires multiple steps:


**FIGURE 1 | Diagrams summarizing aspects of mitochondrial handling of stimulation-induced Ca2<sup>+</sup> loads in wild-type (B) and in presymptomatic (C) and symptomatic (D) SOD1-G93A motor neurons, inferred from measurements in somata and motor terminals. (A)** Resting wild-type and SOD1-G93A motor neurons. <sup>m</sup> is inside-negative due to H+ extrusion by the electron transport chain (ETC). Pi represents dynamic phosphate-dependent matrix Ca2<sup>+</sup> buffering capacity. **(B)** Wild-type neuron buffering a large stimulation-induced Ca2<sup>+</sup> influx. At least 50% of this influx enters mitochondria via the mitochondrial Ca2<sup>+</sup> uniporter (MCU). The mitochondrial Ca2<sup>+</sup> influx partially depolarizes m, thus increasing ETC activity. Matrix Ca2<sup>+</sup> buffering limits the increase in matrix [Ca2+]. **(C)** Presymptomatic SOD1-G93A neuron. Much of the incoming Ca2<sup>+</sup> load continues to enter mitochondria. Matrix buffering persists, but the ability to accelerate ETC activity is reduced, so <sup>m</sup> depolarization is greater. **(D)** Symptomatic SOD1-G93A neuron. Total mitochondrial Ca2<sup>+</sup> uptake is reduced due to (1) greater depolarization of <sup>m</sup> due to further reduction in the ability to accelerate ETC activity, (2) reduced ability to buffer matrix Ca2+, resulting in a greater elevation of matrix [Ca2+], even though total mitochondrial Ca2<sup>+</sup> uptake is reduced. The greater elevation of matrix [Ca2+] produces transient openings of the mPTP. In all diagrams, CaB represents non-mitochondrial Ca2<sup>+</sup> buffering/extrusion (including Ca2<sup>+</sup> uptake by endoplasmic reticulum), which increases in symptomatic neurons as mitochondrial uptake diminishes. These simplified diagrams omit the outer mitochondrial membrane and the pH-dependence of the matrix Ca2<sup>+</sup> buffer, and assume that the stimulation-induced Ca2<sup>+</sup> influx across the cell membrane is not altered by the SOD1-G93A mutation.

(3) Mechanisms to extrude matrix Ca2<sup>+</sup> into the cytosol. In motor nerve terminals a major extrusion mechanism is the mitochondrial Na+/Ca2<sup>+</sup> exchanger (distinct from the plasma membrane Na+/Ca2<sup>+</sup> exchanger) (García-Chacón et al., 2006). More rapid discharge of matrix Ca2<sup>+</sup> loads can be achieved by opening the mitochondrial permeability transition pore (mPTP), a high-conductance channel in the inner mitochondrial membrane, likely formed by dimers of ATP synthase (see below, Bernardi and Petronilli, 1996; Bernardi and von Stockum, 2012; Giorgio et al., 2013).

#### **SYMPTOMATIC SOD1-G93A MOTOR NEURON SOMATA AND TERMINALS SHOW REDUCED MITOCHONDRIAL Ca2<sup>+</sup> UPTAKE AND TRANSIENT mPTP OPENING DURING STIMULATION**

For technical reasons, to date most studies of mitochondrial function in fALS mice have used mitochondria isolated from (ventral) spinal cord or brainstem, only a fraction of which originate from motor neurons. However, Fuchs et al. (2013) measured mitochondrial contributions to regulation of stimulationinduced elevations of cytosolic [Ca2+] in vulnerable hypoglossal motor neurons in SOD1-G93A mice at presymptomatic and end-stage disease. Merging their findings with measurements of stimulation-induced changes in <sup>m</sup> and matrix [Ca2+] from vulnerable motor nerve terminals in mutant SOD1 mice is beginning to suggest mechanisms by which disturbances in mitochondrial function and Ca2<sup>+</sup> handling develop.

Abnormal mitochondrial morphology is an early event in fALS mice, and some of the earliest detected changes occur in motor terminals (reviewed in Barrett et al., 2011). At presymptomatic ages Nguyen et al. (2009) found that mitochondria in motor nerve terminals innervating a fast muscle depolarize more in response to repetitive stimulation than those in wild-type terminals. These depolarizations were not altered by inhibiting mPTP opening with cyclosporin A (CsA). These findings suggest an early limitation of the ability to accelerate electron transport in response to <sup>m</sup> depolarization. Damiano et al. (2006) reported reduced ability of ventral horn mitochondria to handle Ca2<sup>+</sup> loads beginning as early as 65 days in SOD1-G93A mice. A Ca2<sup>+</sup> imaging study comparing vulnerable presymptomatic motor neurons (hypoglossal, 70 days) to wild-type motor neurons found similar elevations of cytosolic [Ca2+] at moderate stimulation frequencies (40 Hz), but greater elevations at maximal stimulation frequencies (Fuchs et al., 2013). In this study a mitochondrial uncoupler significantly impaired Ca2<sup>+</sup> clearance. Combining information from motor terminals and cell bodies, it appears that at presymptomatic ages vulnerable motor neurons exhibit limitations in mitochondrial responses and Ca2<sup>+</sup> handling at high frequencies of stimulation, but continue to rely heavily on mitochondria for clearance of large Ca2<sup>+</sup> loads (**Figure 1C**).

Disturbances of mitochondrial function and Ca2<sup>+</sup> handling become much more pronounced as mice become symptomatic and approach end-stage disease (**Figure 1D**). Nguyen et al. (2011) found that the transient <sup>m</sup> depolarizations produced by repetitive stimulation were much larger and longer-lasting in end-stage than in presymptomatic terminals. These <sup>m</sup> depolarizations were reduced by CsA. Vila et al. (2003) reported that in end-stage mice the stimulation-induced elevations of matrix [Ca2+] showed upward ramping behavior in contrast to the plateau recorded in wild-type mice. These findings suggest reduced ability to accelerate ETC activity and/or reduced ability to buffer matrix [Ca2+]. In addition, the CsA sensitivity and transient nature of the endstage <sup>m</sup> depolarizations suggest reversible openings of the mPTP (reviewed in Nicholls, 2009). Temporary reversible openings of the mPTP might provide an alternative, rapid method for removing large Ca2<sup>+</sup> loads from the matrix (Bernardi and Petronilli, 1996; Bernardi and von Stockum, 2012), acting perhaps as a safety valve. In motor neuron somata of end-stage mice stimulation-induced elevations in cytosolic [Ca2+] were no longer affected by mitochondrial uncoupling or blocking the MCU, indicating severely reduced mitochondrial Ca2<sup>+</sup> uptake, even though there was no change in *resting* m. Expression of cDNAs for MCU-associated proteins was not decreased, but rather increased (Fuchs et al., 2013). These findings are consistent with the idea that mitochondrial defects become apparent mainly when mitochondria are called upon to accelerate ETC activity and/or matrix Ca2<sup>+</sup> buffering in response to stimulation.

When mitochondrial Ca2<sup>+</sup> handling becomes dysfunctional in end-stage disease, motor neurons must rely more heavily on alternative mechanisms for handling stimulation-induced Ca2<sup>+</sup> loads. Fuchs et al. (2013) report increased reliance on plasma membrane Ca2<sup>+</sup> extrusion in end-stage motor neuron somata. Another alternative mechanism is Ca2<sup>+</sup> uptake into endoplasmic reticulum via SERCA pumps. In wild-type terminals inhibition of SERCA pumps with cyclopiazonic acid (CPA) has no significant effect on stimulation-induced elevations of cytosolic [Ca2+] (David and Barrett, 2003) or on stimulation-induced changes in end-plate potential (EPP) amplitude (**Figure 2A**). However, in symptomatic mutant SOD1 mice CPA greatly increases EPP amplitudes during tetanic stimulation (**Figure 2B**), suggesting greater elevations of cytosolic [Ca2+] when SERCA pumps are inhibited. This increased reliance on non-mitochondrial modes of Ca2<sup>+</sup> clearance likely subjects cells and terminals to higher peak cytosolic [Ca2+] during stimulation, because these non-mitochondrial clearance mechanisms rely on active transport (primary via ATPases or secondary via plasma membrane Na+/Ca2<sup>+</sup> exchanger) and thus may be slower than the rapid passive uptake of Ca2<sup>+</sup> into mitochondria that becomes possible following opening of the MCU.

#### **WHY DOES KNOCKOUT OF CYCLOPHILIN D (CyPD) PRESERVE MOTOR NEURON SOMATA BUT NOT MOTOR AXONS OR TERMINALS IN MUTANT SOD1 MICE?**

In an effort to understand how mitochondrial dysfunction contributes to pathology in mutant SOD1 mice Parone et al.

(2013) completed a comprehensive study analyzing the effects of knocking out CyPD, a matrix protein, in mice expressing one of three different fALS-producing SOD1 mutations. CyPD is not required for mPTP formation, but does sensitize the mPTP to Ca2<sup>+</sup> (Giorgio et al., 2013), so knockout of CyPD would be expected to inhibit mPTP opening by increasing the matrix [Ca2+] required for opening. CyPD knockout had many beneficial effects within the spinal cord, including retention of normal mitochondrial morphology, improved mitochondrial ATP synthesis, increased mitochondrial Ca2<sup>+</sup> buffering capacity, reduced glial activation, reduced aggregation of misfolded SOD1, and improved motor neuron survival. These beneficial effects suggest an important role for mPTP opening in multiple aspects of motor neuron pathology. Nonetheless, the CyPD knockout had no beneficial effect on muscle denervation, motor axon degeneration or disease progression. This result reinforces evidence from multiple studies indicating that this disease includes a dying-back pathology (Fischer et al., 2004; Fischer and Glass, 2007), such that maintenance of the central soma, though necessary, is not sufficient to sustain the motor axon and motor terminal connections with muscle (reviewed by Murray et al., 2010; Dadon-Nachum et al., 2011).

Multiple possible mechanisms might explain why degeneration of motor axons/motor terminals was not postponed by CyPD knockout. Some examples:


and longer-lasting stimulation-induced elevations of cytosolic [Ca2+] in motor nerve terminals, with or without mPTP opening.

	- (a) While *irreversible* mPTP opening is a trigger for cell death, the *reversible* opening of the mPTP that appears to occur in repetitively stimulated motor terminals of symptomatic SOD1-G93A mice (Nguyen et al., 2011) may serve a beneficial role (e.g., as in heart cells, e.g., Saotome et al., 2009). Temporary mPTP opening might allow the mitochondrion to discharge a dangerously high level of Ca2+, then close the mPTP and re-establish <sup>m</sup> (reviewed by Bernardi and von Stockum, 2012). Such a mechanism might be especially helpful in motor terminals that experience large Ca2<sup>+</sup> loads only episodically, as might be expected from the discharge pattern of motor neurons that innervate FF muscle fibers.
	- (b) CyPD knockout may have preserved motor neuron somata by inhibiting a death mechanism specific to the central nervous system. For example, Re et al. (2014) found that astrocytes cultured from ALS patients killed stem-cell-derived motor neurons via necroptosis, a form of programmed necrosis that is Bax-dependent, suggesting mitochondrial involvement.
	- (c) CyPD knockout may fail to prevent disease-related changes in skeletal muscles, which dominate the environment experienced by motor nerve terminals. Skeletal muscles in mutant SOD1 mice evidence signs of hypermetabolism and oxidative stress even at asymptomatic stages of disease (reviewed in Cozzolino and Carri, 2012), and increase expression of Nogo-A, which inhibits formation of neuromuscular junctions (Bros-Facer et al., 2014).
	- (d) Maintenance of motor nerve terminals may require some activity of SOD1 in addition to its dismutase role. Degeneration of motor nerve terminals is one of the earliest events in all fALS-causing SOD1 mutations studied to date (Parone et al., 2013), but the ability of SOD1 mutations to produce motor neuron disease seems unrelated to their dismutase activity (reviewed by Valentine et al., 2005). SOD1 is necessary to maintain neuromuscular junctions, because SOD1 knockout

mice gradually lose skeletal muscle innervation, even though motor neuron somata remain intact (Fischer et al., 2012). Perhaps fALS-inducing mutations of SOD1 disrupt this neuromuscular junction-maintaining function via a mechanism independent of dismutase activity.

These considerations suggest that the search for mechanisms underlying ALS pathologies and effective ALS therapeutics needs to include a focus on motor axons, motor nerve terminals and/or skeletal muscle.

#### **ACKNOWLEDGMENT**

Supported by the Muscular Dystrophy Association.

#### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 April 2014; accepted: 16 June 2014; published online: 08 July 2014. Citation: Barrett EF, Barrett JN and David G (2014) Dysfunctional mitochondrial Ca*2<sup>+</sup> *handling in mutant SOD1 mouse models of fALS: integration of findings from motor neuron somata and motor terminals. Front. Cell. Neurosci. 8:184. doi: 10.3389/ fncel.2014.00184*

*This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2014 Barrett, Barrett and David. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Seeking homeostasis: temporal trends in respiration, oxidation, and calcium in SOD1 G93A Amyotrophic Lateral Sclerosis mice

#### *Cameron W. Irvin, Renaid B. Kim and Cassie S. Mitchell\**

*Department of Biomedical Engineering, Georgia Institute of Technology – Emory University, Atlanta, GA, USA*

#### *Edited by:*

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### *Reviewed by:*

*Karthik Bodhinathan, Sanford-Burnham Medical Research Institute, USA Adán Dagnino-Acosta, University Center for Biomedical Research/University of Colima (Cátedras CONACYT), Mexico Jia Li, Harvard Medical School, USA*

#### *\*Correspondence:*

*Cassie S. Mitchell, Department of Biomedical Engineering, Georgia Institute of Technology – Emory University, 313 Ferst Drive, Atlanta, GA 30332, USA cassie.mitchell@bme.gatech.edu*

> *Received: 17 April 2015 Accepted: 18 June 2015 Published: 01 July 2015*

#### *Citation:*

*Irvin CW, Kim RB and Mitchell CS (2015) Seeking homeostasis: temporal trends in respiration, oxidation, and calcium in SOD1 G93A Amyotrophic Lateral Sclerosis mice. Front. Cell. Neurosci. 9:248. doi: 10.3389/fncel.2015.00248* Impairments in mitochondria, oxidative regulation, and calcium homeostasis have been well documented in numerous Amyotrophic Lateral Sclerosis (ALS) experimental models, especially in the superoxide dismutase 1 glycine 93 to alanine (SOD1 G93A) transgenic mouse. However, the timing of these deficiencies has been debatable. In a systematic review of 45 articles, we examine experimental measurements of cellular respiration, mitochondrial mechanisms, oxidative markers, and calcium regulation. We evaluate the quantitative magnitude and statistical temporal trend of these aggregated assessments in high transgene copy SOD1 G93A mice compared to wild type mice. Analysis of overall trends reveals cellular respiration, intracellular adenosine triphosphate, and corresponding mitochondrial elements (Cox, cytochrome c, complex I, enzyme activity) are depressed for the entire lifespan of the SOD1 G93A mouse. Oxidant markers (H2O2, 8OH2- dG, MDA) are initially similar to wild type but are double that of wild type by the time of symptom onset despite early post-natal elevation of protective heat shock proteins. All aspects of calcium regulation show early disturbances, although a notable and likely compensatory convergence to near wild type levels appears to occur between 40 and 80 days (pre-onset), followed by a post-onset elevation in intracellular calcium. The identified temporal trends and compensatory fluctuations provide evidence that the "cause" of ALS may lay within failed homeostatic regulation, itself, rather than any one particular perturbing event or cellular mechanism. We discuss the vulnerabilities of motoneurons to regulatory instability and possible hypotheses regarding failed regulation and its potential treatment in ALS.

Keywords: ALS, motor neuron disease, mitochondria, oxidative stress, Ca2**+**, energy metabolism

### Introduction

Amyotrophic Lateral Sclerosis (ALS) is a late-onset neurodegenerative disease consisting of progressive muscle atrophy, muscle paralysis, dysarthria, dysphagia, and dyspnea. While there has been much research conducted on the disease, the precise causes and effective treatments have remained elusive. Transgenic mice, and namely the superoxide dismutase 1 glycine 93 to alanine mutation (SOD1 G93A), have served as the predominant means by which to investigate the underlying cellular pathophysiology (Pfohl et al., 2015). A multitude of categorical disturbances have been identified, which are described in detail in a recent informatics-based systematic review of the SOD1 G93A field (Kim et al., 2015): apoptosis, including changes in pro- and anti-apoptotic signals; energetics, including mitochondrial dysfunction, adenosine triphosphate (ATP) depletion, and calcium homeostasis; excitability, including hypoexcitability, hyperexcitability, and excitotoxicity; genetic transcription, including damage to mRNA and DNA; inflammation, due to reactive microglia and astrocytes; oxidative stress, from the production of free radicals; proteomics, including the build-up of mutant protein aggregates and reduced autophagy or proteasome function; and systemic function, which also includes potential non-neuromuscular contributors.

This Frontiers Research Topic and present study focuses on a unique and highly inter-related triad of the ALS pathophysiology: the role of mitochondria, oxidative stress, and altered calcium homeostasis. Most of the previous experimental work has focused on identifying the presence of impairments in this triad using a wide variety of specific methods and measures, such as recording of the mitochondrial potential, evaluation of intracellular ATP concentration, electrophysiological assessment of calcium entry, and measurement of intracellular free radicals. While the presence of deficiencies in mitochondria, oxidative regulation, and calcium homeostasis has been well established, their timing, as a function of ALS disease initiation and progression, is less understood. The goal of this work is to evaluate their overall temporal trends, pre-natal through death, in the SOD1 G93A transgenic ALS mouse model. In this systematic review of 45 articles, we aggregate *in vitro*, embryonic, and *in vivo* experimental measurements of cellular respiration, mitochondrial mechanisms, oxidative and antioxidative markers, and intracellular calcium in both SOD1 G93A transgenic ALS mice and in wild type control mice. We evaluate the magnitude and statistical trend of these assessments in SOD1 G93A mice compared to wild type mice and examine changes over the course of the SOD1 G93A mouse life span.

### Materials and Methods

The general method includes: (1) identifying and recapturing published experimental data for SOD1 G93A transgenic mouse and wild type control mouse assessments of mitochondrial function, oxidative stress, and calcium homeostasis; (2) Normalizing recaptured data to temporally compare the assessed measures; (3) Performing a Gaussian average to temporally interpolate the values and develop a visual trend line for each measure and category of measures.

#### Inclusion and Exclusion Criteria

Potential articles were identified under the PubMed search criteria of (G93A OR transgenic mouse) AND (ALS OR "Amyotrophic Lateral Sclerosis"). Initial exclusion criteria consisted of: non-English language articles; articles for which fulltext pdf downloads were unavailable (see Data Recapture); and articles labeled as literature reviews.

Keyword searches of recaptured figure captions and within figure text were performed to find relevant articles from the initial literature pool (see Data Recapture) using the following terms: calcium and its permutations, including Ca2+, Ca, etc.; mitochondria and its permutations (mito∗); oxidative stress and its permutations (oxid∗), reactive oxygen species (ROS), free radicals, hydrogen peroxide (H2O2), nitric oxide (NO), malondialdehyde (MDA), 8-hydroxy-2- deoxyguanosine (8OH2- dG), heat shock proteins (HSPs). Study-specific inclusion criteria required the use of both non-treated SOD1 G93A and wild type transgenic mice for a given quantitative experimental measurement. "Non-treated" consisted of controlled experimental assessments of the measured parameter(s) without the application of chemicals or processes meant to intentionally attempt to modify the assessed measures or related physiology.

#### Data Recapture

Articles were either downloaded using PubMed Central or from e-journal subscriptions available from the libraries of Georgia Institute of Technology and Emory University. Data was recaptured from the following article locations (Kim et al., 2015), referred to as entities: article title; abstract; figure captions; within figure text (*x*–*y* axis labels, bar graph categorical labels, legends, etc.); and data series and response values (e.g., quantifiable figure/table data). Data was scraped from the full-text pdf article. Every data point was reviewed by an independent quality control team to insure complete accuracy.

### Categorical Definitions

Given that each study utilizes its own specific measures, aggregation is a requirement in order to obtain a meaningful visualization of categorical trends and meta-analysis results. Experimental measures were aggregated into four main categories: calcium regulation, cellular respiration, mitochondrial mechanisms, and oxidative regulation. Sub-categories of measures are further defined within each category as discussed below.

#### Calcium Regulation

The calcium regulation category includes experimental assessments of intracellular calcium dynamics, including calcium entry, calcium sinks, free calcium concentration, and calcium sensitivity. Calcium entry encompassed electrophysiological measurements of extracellular calcium entering through the membrane of motoneuron cells (e.g., Ca2<sup>+</sup> persistent inward current, Ca2<sup>+</sup> amplitude, Cell Voltage, Cellular Calcium, etc.). Calcium sensitivity included measures of mitochondrial membrane sensitivity to a discretely measured calcium challenge. Calcium sink values encompassed measures of calcium buffering or calcium capacity. Calcium concentration encompassed general cytosolic free calcium levels.

#### Cellular Respiration

The cellular respiration category includes measures of general respiration rate as well as intracellular concentrations of ATP and adenosine diphosphate (ADP). Measures of the general respiration rate include production of heme, oxygen consumption, or respiration control ratio.

#### Respiration Mechanisms

Respiration mechanisms included experimental assessment of complex I activity, COX activity, Cytochrome C levels, and general mitochondrial enzyme activity (including Complex III, IV, V). These parameters are involved the electron transport chain (ETC), which leads to the production of ATP.

#### Oxidative Regulation

Experimentally assessed contributors to oxidative stress include hydrogen peroxide (H2O2) production and the oxidant markers MDA and 8-hydroxy-2- -deoxyguanosine (8OH2- dG). The concentration of HSP was also incorporated in the oxidative regulation category given their neuroprotective effects and potential impact on the slowing of oxidant-induced symptoms. Specific HSPs included HSP 25, 27, and 70.

#### Tissue Sources

*In vivo* tissues primarily consisted of SOD1 G93A mouse brain or spinal cord cells; in fact, 35 articles identified at least one of these two regions as a source. Some articles utilized homogenized tissue or multiple tissue sources (Leclerc et al., 2001; Damiano et al., 2006; Johnson et al., 2011; Miana-Mena et al., 2011), which included spinal cord or brain tissue with other systemic tissues, such as blood, liver, soleus, diaphragm, and liver. Tissues used for *in vitro* assessment were more varied. Cell lines mostly included standard SOD1 G93A transfected mice cells. However, other G93A-transfected sources included SH-SY5Y cells (Carri et al., 1997; Goos et al., 2007; Pesaresi et al., 2011), NSC-34 cells (Liu et al., 2002; Ferri et al., 2008; Mali and Zisapel, 2009; Crippa et al., 2010), yeast (Gunther et al., 2004; Kloppel et al., 2010), and bacteria (Singh et al., 1998).

#### Analysis

The ratio of SOD1 G93A to wild type (e.g., SOD1 G93A/wild type) is used to normalize each assessed measure. Each study was normalized to its own published wild type data. For each included measure or category of measures, the ratio of SOD1 G93A to wild type is plotted versus time. Data from each article is given equal weight. For ease of visualization, all *in vitro* cell line experimental data is plotted as −20 days and embryonic experimental data is plotted as −5 days; *in vivo* data is plotted at its corresponding post-natal day of experimental assessment. A standard Gaussian average was performed to interpolate values in-between the raw experimental data points and to produce trend lines indicative of the general aggregate behavior of each experimentally measured parameter.

### Results

In total, 262 data points from 45 unique papers were collected and normalized for inclusion in this meta-analysis. **Table 1** shows

#### TABLE 1 | Categorization, distribution, and sources of included experimental data.


*There are four major categories of measures: intracellular calcium, mitochondrial mechanisms, cellular respiration, and oxidative regulation). Sub-categories of measures, as defined in the Methods, are listed under each category along with the number of included unique articles and extracted data points. Each data point represents a paired SOD1 G93A and wild type mouse assessment at a discrete time point.*

the experimental data point and article distribution to each of the defined categories and subcategories of measures. The category, mitochondrial mechanisms, which contains measurements of constituents necessary for cellular respiration, includes 67 values from 10 unique articles. Cellular respiration, which includes the assessed respiration rate and intracellular concentrations of ATP and ADP, includes 59 values from 12 unique articles. The oxidative stress category, which includes oxidative markers and anti-oxidative HSPs, has a total of 73 values from 14 unique articles. Intracellular calcium, which contains measures examining intracellular calcium homeostasis, contains 63 data points from 15 unique articles.

#### Cellular Respiration is Depressed for Entire Lifespan

Among the most interesting trends found in this meta-analysis study is that ATP production, along with general respiration rates, were found to be depressed for the entire lifespan of the SOD1 G93A mouse (**Figure 1**). While it has been well documented that respiration rates are lowered in ALS (Kawamata and Manfredi, 2010; Cozzolino and Carri, 2012; Peixoto et al., 2013) even well before physical pathologies develop (Browne et al., 2006), this meta-analysis supports the assertion that this phenomenon is a trend that, at least in the high-copy SOD1 G93A transgenic mouse model, is present since birth. That is, the SOD1 G93A mice have notable depression of cellular energetics well before symptom onset, and this depression remains throughout the course of disease progression.

**Figure 1** illustrates the ratio of transgenic mouse to wild type mouse experimentally measured levels of intracellular ATP, ADP, and respiration rate. All of the *in vitro* cellular data (plotted as <sup>−</sup>20 days in **Figure 1**) collected for intracellular ATP concentration and respiration rate fall well below their wild type counterparts. Examining the mathematical mean of *in vitro* measures of ATP and respiration rate reveal that SOD1 G93A levels are approximately 70% of those seen in wild type, which is equivalent to a 30% reduction.

Post-natal *in vivo* assessment of intracellular ATP and respiration rate in SOD1 G93A mice also shows substantial depression compared to wild type mice. While ATP and respiration rate is depressed throughout the life span of the SOD1 G93A mouse, there appears to be small fluctuations throughout the disease course. However, more data is necessary to determine whether these small fluctuations are statistically or mechanistically meaningful. ATP is at its lowest at the disease end point, where ATP levels approach only 30% of wild type (**Figure 1**). Interestingly, the temporal trend of ADP is slightly different than ATP and respiration rate. For most of the disease course, ADP is depressed in a similar manner to ATP and respiration. However, ADP levels in SOD1 G93A mice show an interesting rise above wild type control mice that occurs near the disease end point. This near-death rise in ADP could be

purposes, *in vitro* data is plotted at −20 days and *in vivo* time points are plotted at their corresponding post-natal day of assessment, 0–150 days. A gray dotted line is provided to show the expected wild type or homeostatic value. Trend lines are generated based on a Gaussian average of the normalized data points.

attributed to the cells' inability to convert ADP to ATP, which would leave ADP in excess. In fact, this same trend in the lowering of the ATP/ADP ratio is seen in clinical patients (Ghiasi et al., 2012).

#### Depression of Mitochondrial Mechanisms

Additionally, measures of mitochondrial mechanisms and signals necessary for respiration and ATP production are similarly depressed throughout the course of the disease. **Figure 2A** illustrates the temporal trends of four different mitochondrial mechanism measurements and signals necessary for cellular respiration: complex I, COX, general enzyme activity, and cytochrome C. *In vitro* measures were only obtainable for enzyme activity, which shows a depression similar to that seen in ATP levels. Post-natal *in vivo* assessment of SOD1 G93A mice reveals that all four of the mitochondrial mechanism measurements are generally depressed compared to wild type. The *in vivo* depression is present at birth and throughout the entire disease duration. Although the mitochondrial mechanism experimental measures remain below wild type, the Gaussian average trend lines identify a potential small bump in mitochondrial mechanism activity near disease onset (around 80 days), which could represent a regulatory compensation mechanism; a larger sample size is necessary to determine if this small bump has possible statistical or mechanistic implications in disease progression. Finally, when the mitochondria's ability to produce ATP is impaired, there is a compensatory increase in mitochondrial enzyme complexes, especially Complex II, III, IV (Nalbandian et al., 2015). This upward trend in mitochondrial enzymes is seen **Figure 2A** near post-onset and the disease end point.

#### Elevation of Oxidants Near Onset

Oxidant levels have been documented as being elevated throughout different stages of ALS (Liu et al., 2002; Mattiazzi et al., 2002; Martin et al., 2007), including pre-onset (around 40 days), onset (80 days), and especially end-stage (120+ days). In **Figure 2B** we plot the temporal trends of three commonly measured oxidants in SOD1 G93A mice (8OH2- dG, MDA, and H2O2). The data presented in **Figure 2B** reveals that SOD1 G93A intracellular oxidant levels are initially similar to slightly above wild type at birth. However, by pre-onset, levels are mildly elevated, and at onset and end point, oxidant levels are substantially increased compared to wild type. *In vitro* assessment of oxidants does not reveal as pronounced of elevation as the *in vivo* assessments. *In vivo* assessment of oxidants reveals a 1.5 factor increase in oxidants compared to wild type around symptom onset (80 days). *In vivo* assessment near the SOD1 G93A disease end point (120+ days) reveals oxidant levels that are a factor of 2–8 times greater than seen in wild type control mice.

Heat shock proteins, which have an anti-oxidative effect, are also plotted in **Figure 2B**. HSPs in SOD1 G93A mice were found to initially be substantially greater than wild type levels, but they exhibited a fluctuating decline as the disease progressed. However, there appears to be a recurrent delayed rise in HSP as the disease enters the symptomatic stage (around 80 days).

FIGURE 2 | Temporal trends of mitochondrial mechanisms, oxidant regulation, and calcium regulation in the SOD1 G93A transgenic mouse model. The ratio of SOD1 G93A to wild type (SOD1 G93A/wild type) is plotted over time for each experimental measure. For visualization purposes, *in vitro* data is plotted at −20 days, embryonic data is plotted at −5 days, and *in vivo* time points are plotted at their corresponding post-natal day of assessment, 0–150 days. A gray dotted line is provided to show the expected wild type or homeostatic value. Trend lines are generated based on a Gaussian average of the normalized data points. (A) Mitochondrial mechanisms (Complex I, COX, enzyme activity, and cytochrome c). (B) Oxidant markers (8OH2- dG, MDA, H2O2) and protective heat shock proteins (HSPs). (C) Calcium regulation (entry, sensitivity, sink, and cytosolic concentration).

This fluctuation has been previously described. It is hypothesized that HSP levels are insufficient to quell the oxidant rise. Thus, decreased HSP levels actually precede motor neuron loss in ALS (Maatkamp et al., 2004).

#### Fluctuations of Intracellular Calcium

Calcium homeostasis is critical for both functional neural excitability and normal cellular signaling. There are four main experimental types of calcium regulatory measurements: calcium entry (incoming calcium through ion channels), calcium sensitivity (measurement of the cell's rate of response to calcium), calcium sink (binding and storage of intracellular calcium, including buffers, transporters, and intracellular stores), and the actual calcium concentration (free intracellular concentration). Each of these measures contributes to the balance or homeostasis of intracellular calcium.

*In vitro* data examining free intracellular calcium concentration is conflicting, with about half of the data points showing elevated calcium and half showing lower intracellular calcium compared to wild type (points plotted at −20 days in **Figure 2C**). These apparent conflicts in *in vitro* intracellular calcium concentration could possibly be explained by the usage of different tissue types for *in vitro* assessment (see Tissue Sources). Measurements of *in vitro* calcium sinks are depressed compared to wild type, ranging from 60 to 80% of that seen in wild type mice, and calcium sensitivity is about 30% of wild type. Post-natal *in vivo* assessment of intracellular calcium and calcium entry at birth reveals levels that are substantially above wild type. However, free calcium and calcium entry appears to dip to near-normal levels during pre-onset. There is limited data for SOD1 G93A mice calcium entry and concentration from onset through end point, but available data revels that calcium appears to rise sharply after onset, resulting in a disease end point intracellular calcium concentration that is a factor 1.5 greater than wild type. *In vivo* assessment of calcium sinks and calcium sensitivity in SOD1 G93A mice show depressed levels compared to wild type from birth through the disease end point. Intuitively, the point at which the calcium sink trend line is highest coincides with the time points when calcium concentration, or free-floating calcium, is lowest.

#### Discussion

The results of our systematic review and meta-analysis of 45 articles shed new light on the temporal trends of cellular respiration, oxidative markers, mitochondrial mechanisms, and calcium regulation in the SOD1 G93A transgenic ALS mouse model. By aggregating data, we show that cellular respiration and corresponding mitochondrial mechanism are impaired for the entire lifespan of the SOD1 G93A mouse. Oxidant markers are initially similar to wild type but are more than double that of wild type by the time of symptom onset despite early post-natal elevation of protective HSPs. All aspects of calcium regulation show early disturbances, although a notable and likely compensatory convergence to near wild type levels occurs between 40 and 80 days, which is followed by a divergence after symptom onset.

This systematic review clearly shows that SOD1 G93A mice exhibit a long-term metabolic deficit, however, these symptoms are also present in other ALS mouse models. Dupuis et al. (2004) performed metabolic experiments on both G93A and G86R mice to demonstrate the similarities in mitochondrial function. G37R mice also show significant reduction in ATP production (Coussee et al., 2011). Finally, substantial metabolic disturbances have also been documented in non-SOD transgenic mice, including mice with mutations in TDP-43, FUS, VCP, among others (Carri et al., 2015).

#### Interactions within the Respiration-Oxidation-Calcium Triad

There are multiple feedback loops between the triad of cellular respiration, calcium regulation, and oxidative regulation. The inter-relationships between the categories and sub-categories of measurements examined in this meta-analysis are illustrated in **Figure 3**. Red boxes indicate parameters, which are lower in SOD1 G93A mice compared to wild type, and green box indicates parameters, which are higher in SOD1 G93A mice compared to wild type. Similarly, the color of the arrows indicates either a positive relationship (green) or a negative relationship (red), and their size indicates the relative strength of the relationship. The biology of these interactions is summarized below.

Mitochondria have a highly interactive dynamic with the endoplasmic reticulum (ER), the main intracellular calcium storehouse. Mitochondria take up calcium via the calciumsensitive mitochondrial uniporter. However, sustained free cytosolic calcium inactivates the uniporter, preventing further calcium uptake (Moreau et al., 2006). Accumulated calcium in the mitochondria can then be released back into the cytosol via the sodium–calcium and hydrogen–calcium exchangers (Fuchs et al., 2013). Once intramitochondrial calcium rises above a certain level, the mitochondrial transition pore opens, initiating apoptotic or necrotic signaling cascades (Leung and Halestrap, 2008). Calcium originating from the activation of AMPA receptors and/or pathologically increased membrane permeability is thought to result in this shift of calcium from the ER to the mitochondria. Ryanodine receptors on the surface of the ER further amplify calcium-mediated calcium release from the ER, which in turn, could further exacerbate AMPA activation (Berridge, 2002). A second receptor that exacerbates calcium release from the ER is the calcium-activated IP3R.

Collectively, intracellular release of calcium from the ER, mitochondria, and other intracellular stores could explain the increase in intracellular cytosolic calcium concentration seen near onset (∼100 days), which is mirrored by a paradoxical decrease in extracellular calcium entry (see **Figure 2C**). Another contributor for this apparent paradox could be a decrease in expression of calcium binders like calbindin D28K and parvalbumin (Celio, 1990), which have been proposed to result in increased cytosolic calcium in ALS mice (Appel et al., 2001). Chin et al. (2014) similarly shows a decrease in parvalbumin in SOD1 G93A mice as well as a reduction in sarcoplasmic/ER Calcium ATPase proteins, including SERCA1. Notably, calcium binders, which fall under the calcium sink category in this meta-analysis, show a slight dip that also corresponds to the timing of the intracellular calcium increase (**Figure 2C**).

Calcium release has a bi-directional relationship with ROS production since ROS homeostasis is maintained via Ca2<sup>+</sup> signaling and Ca2<sup>+</sup> dependent pathways. Calcium stimulates NO synthesis and leads to ROS production at Complex III (Feissner et al., 2009). Moreover, because the ryanodine receptor forms a tetramer with the sarcoplasmic and ERs, the reversible oxidation of endogenous superoxide groups can result in the release of additional calcium from the sarcoplasmic reticulum (Fill and Copello, 2002). Finally, oxidative agents like peroxide directly induce calcium release from the ER via IP3R (Wesson and Elliott, 1995). In summary, free radicals induce calcium leakage into the cytosol via the ryanodine receptor, Ca2+ leak channels, and inositol 1,4,5-trisphosphate receptors, and conversely, intracellular calcium concentration activates NOX and NOS, which then produces additional excess ROS and RNS, respectively (Chin et al., 2014). Ultimately, elevated internal calcium creates a cyclical feed forward mechanism that continually increases calcium and oxidative stress to the point of apoptosis (Chin et al., 2014).

Lower respiration rates, and consequently, lower intracellular ATP concentrations, directly contribute to the lowering of the mitochondrial potential, which can ultimately initiate apoptotic cascades. The impact of lower ATP concentration on oxidative and calcium imbalances is bi-directional, with increases in oxidants and calcium-mediated calcium release further impairing mitochondrial function, especially Complex 1, a key constituent for ATP production (Cassina et al., 2008; Cozzolino and Carri, 2012; Lautenschlager et al., 2013). Through a less direct path, an increase in oxidative stress can also lead to a swelling of mitochondria, which also further inhibits ATP production (Martin et al., 2007). Finally, lower concentrations of ATP impede calcium-ATPase in removing free calcium from the cytosol or shuttling calcium back to the ER for storage (Kaplan et al., 2003; Fuchs et al., 2013).

#### Deciphering the Timing: Cause, Effect, and Instability

Because of their large size and innate emergent properties, motoneurons are susceptible to homeostatic instabilities. It has been previously shown that motoneurons, even in a physiological state, have insufficient mitochondrial capacity to buffer large calcium fluxes. Calcium buffering insufficiency is thought to be due to a reduced mitochondrial density per volume compared to non-motoneurons (Grosskreutz et al., 2007). Therefore, mitochondrial dysfunction and impaired calcium homeostasis is hypothesized to account for the selective vulnerability of motoneurons (Jaiswal and Keller, 2009; Jaiswal et al., 2009). Another contributor for the selective vulnerability of motoneruons is the requirement for axonal transport of mitochondria over very long distances, up to 1 m (Mitchell and Lee, 2009, 2012a; Lee and Mitchell, 2015). Finally, the dynamics of somatic input processing of motoneurons could explain the earlier death of fast twitch fibers in ALS (Mitchell and Lee, 2011).

Because there are so many interacting variables, it is difficult to determine which parameter(s) initiate versus simply affect the pathophysiological cyclical cascades of depressed cellular respiration, imbalances in calcium homeostasis, and intracellular elevation of oxidants in ALS. The presented data reveals that elevated oxidants appear later in the SOD1 G93A life span, closer to disease onset. However, both calcium and cellular respiration/mitochondrial mechanisms show early deficits. Much like the age-old question, "What came first, the chicken or the egg?" This meta-analysis begs the question, "What comes first improper calcium homeostasis or depressed cellular respiration?"

Scientifically justified arguments could be made for either position. Increased calcium permeability or ATP depletion from sub-par cellular respiration, or a combination of both, could initiate a dynamic instability in the motoneuron that results in the ALS phenotype. The trend lines presented in this meta-analysis reveal the presence of potential compensatory mechanisms, which attempt but ultimately fail, to re-stabilize to homeostasis. For example, between 0 and 20 days, there is a rise in HSPs and a gradual increase in calcium sensitivity. The "slight bump" in mitochondrial mechanisms/cellular respiration at pre-onset also coincides with the lowest intracellular calcium levels. Attempts to re-stabilize to homeostasis could potentially correspond to the small fluctuations apparent in **Figures 1** and **2,** although more data is needed to definitively determine their statistical significance.

Mathematical instabilities within pathophysiological feedback loops have already been identified in a dynamic meta-analysis of the SOD1 G93A mouse model (Mitchell and Lee, 2012b). If unstable pathology dynamics are the actual underlying culprit, it may not actually matter exactly which mechanism first initiated the cascade (see Future Directions). Both preclinical and clinical failures to obtain meaningful success using singlemechanism treatments illustrate the potential validity of this point. Among the many examples are: dichloroacetate (Miquel et al., 2012), which attempts at restoring the mitochondrial respiratory capacity in the astrocytes, *N*-acetyl-glucagon-like peptide-1 (Sun et al., 2013), which endogenously regulates metabolism by promoting insulin synthesis and secretion, and creatine (Groeneveld et al., 2003; Snow et al., 2003; Beal, 2011), which is known to enhance ATP synthesis. However, no single treatment to targeting cellular energetics has been effective enough to translate to an effective treatment for humans (Tadic et al., 2014).

In this meta-analysis, we reveal that, although there are some small fluctuations, cellular respiration is depressed for the entire SOD1 G93A ALS mouse lifespan. Interestingly, ALS patients, prior to the onset of their ALS, have been found to be healthier (e.g., less antecedent disease) than age, gender, and geographymatched control subjects (Mitchell et al., 2015). However, such patients could still have asymptomatic pre-ALS variations in their underlying motoneuron regulation, which make them more susceptible to instabilities. Hypervigilant regulation as been put forth as one possibility to explain how the aboveaverage pre-ALS health of patients could be correlated to a later, destabilizing motoneuron perturbation, which initiates ALS (Mitchell et al., 2015). "Hypervigilant regulation" results when underlying regulatory processes aggressively overreact to correct imbalances from homeostasis, making them 'hypervigilant' to perturbation (in control theory, referred to as a too-high feedback gain). While hypervigilant regulation would initially be overall protective, it could also result in greater later susceptibility to destabilization, especially in highly susceptible motoneurons (Mitchell et al., 2015). The temporal calcium, oxidant, and HSP fluctuations identified in this study, in combination with the oscillatory behavior of other previously identified parameters (Mitchell and Lee, 2012b) such as axonal transport (Mitchell and Lee, 2012a) and excitability (Delestree et al., 2014), are suggestive of the possible role of regulatory and homeostatic impairments as being the "cause" of ALS.

#### Future Directions

Perhaps instead of focusing on mechanistic initiation, treatments should focus on treating the underlying instability, itself (Mitchell and Lee, 2008, 2012b). Whether in engineering process or in biology, treating instability typically requires impacting multiple targets or feedback loops, which may or may not have directly initiated the destabilizing perturbation or event. Combination treatments can leverage synergistic interactions to increase treatment effect size. Multiple experimentalists have attempted combinatory treatments on the SOD1 G93A mouse model (Waibel et al., 2004; Feng et al., 2008; Del Signore et al., 2009; Del Barco et al., 2011). For example, Waibel et al. (2004) experimented with rasagiline, an anti-apoptotic with neuroprotective properties, combined with riluzole, a sodium channel blocker, to reduce excitotoxicity. The combinatory treatment did exhibit a statistically significant improvement compared to control and compared to Riluzole alone.

In addition to exploiting synergistic interactions to increase effect size, combination treatments could potentially be used to re-stabilize the system to homeostasis. Theoretical SOD1 G93A ALS models of combination treatments have shown this exciting possibility (Mitchell and Lee, 2012b). In fact, of the several thousand computationally assessed combination treatment permutations, a few percent of 2 and 3-way treatment strategies were able to mathematically stabilize the ALS pathophysiology (Mitchell and Lee, 2012b). Interestingly, energetics was one of the pathophysiological categories that most frequently appeared in synergistic or stabilizing treatment combinations. Given the early and lasting depression of ATP and respiration rates identified in the present study, it is not surprising that energetics was previously predicted to have the greatest single-category effect size (Mitchell and Lee, 2012b).

Based on the results of the present study, it would appear that therapeutics leveraging the strong interactions within the calcium-respiration-oxidation triad could be promising. As shown in **Figures 1** and **2,** prior to onset, it appears the SOD1 G93A physiology temporarily compensates toward decreasing intracellular calcium, increasing anti-oxidative HSPs, and slightly increasing respiration rate; thus, treatment to amplify these existing compensatory mechanisms would seem intuitive. However, like spinal cord injury (Mitchell and Lee, 2008), such treatments would likely have to be initiated very early in the disease process to have a meaningful effect. In fact, in the case of instability, the timing of treatment may be the most important parameter, especially given human patients will not be treated until after ALS symptoms appear. Obtaining

#### References


a finer point on the timing and statistical significance of fluctuations in intracellular calcium, ATP concentration, and free radicals, is critical to devising combination treatments that have clinically significant results. An additional essential research path is better assessment of homeostatic regulation. Modulation of regulatory pathways may prove more fruitful for re-stabilization than direct physical or chemical manipulation of cellular elements.

### Acknowledgment

Financial support provided by National Institute of Health grants NS061696 and NS081426 to CM.

of misfolded proteins involved in amyotrophic lateral sclerosis (ALS). *Hum. Mol. Genet.* 19, 3440–3456. doi: 10.1093/hmg/ddq257


amyotrophic lateral sclerosis increases the vulnerability of neuroblastoma cells to infectious injury. *BMC Infect. Dis.* 7:131. doi: 10.1186/1471-2334-7-131


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Irvin, Kim and Mitchell. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# SOD1 misplacing and mitochondrial dysfunction in amyotrophic lateral sclerosis pathogenesis

Francesco Tafuri, Dario Ronchi, Francesca Magri, Giacomo P. Comi and Stefania Corti\*

Dino Ferrari Centre, Neuroscience Section, Department of Pathophysiology and Transplantation (DEPT), University of Milan, Neurology Unit, IRCCS Foundation Ca' Granda Ospedale Maggiore Policlinico, Milan, Italy

Amyotrophic lateral sclerosis (ALS) is a fatal motor neuron disease presenting as sporadic (sALS) or familial (fALS) forms. Even if the list of the genes underlining ALS greatly expanded, defects in superoxide dismutase 1 (SOD1), encoding the copper/zinc SOD1, still remain a major cause of fALS and are likely involved also in apparently sporadic presentations. The pathogenesis of ALS is still unknown, but several lines of evidence indicate that the mitochondrial accumulation of mutant SOD1 is an important mechanism of mitochondrial dysfunction, leading to motor neuron pathology and death. The intramitochondrial localization of mutant SOD1 is debated. Mutant SOD1 might accumulate inside the intermembrane space (IMS), overriding the physiological retention regulated by the copper chaperone for superoxide dismutase (CCS). On the other hand, misfolded SOD1 might deposit onto the outer mitochondrial membrane (OMM), clumping the transport across mitochondrial membranes and engaging mitochondrialdependent cell apoptosis. The elucidation of the mechanisms ruling SOD1 localization and misplacing might shed light on peculiar ALS features such as cell selectivity and late onset. More importantly, these studies might disclose novel targets for therapeutic intervention in familial ALS as well as non-genetic forms. Finally, pharmacological or genetic manipulation aimed to prevent or counteract the intracellular shifting of mutant SOD1 could be effective for other neurodegenerative disorders featuring the toxic accumulation of misfolded proteins.

Keywords: superoxide dismutase 1, motor neuron disorder, mitochondria, mitochondrial intermembrane space, outer mitochondrial membrane, protein misfolding

### Introduction

Amyotrophic lateral sclerosis (ALS) is a lethal neurological disease caused by the selective degeneration of upper and lower motor neurons, resulting in progressive muscle denervation and paralysis (Bucchia et al., 2015). Recent studies also feature non-cell autonomous mechanisms involving contiguous and functionally related cells such as astrocytes, oligodendrocytes, microglia as well as the extracellular environment (Rizzo et al., 2014). No effective therapy is available for ALS and understanding the disease pathogenesis could help in developing effective treatments (Bucchia et al., 2015).

The majority of ALS cases are sporadic (sALS) while familial (fALS) forms account for 10% of all ALS presentations (Renton et al., 2014). Nevertheless, sALS and fALS share similar phenotypic features, being clinically indistinguishable, with fALS patients only displaying, sometimes,

#### Edited by:

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Richard Eugene Frye, Children's Hospital Boston/Harvard University, USA Henna Tyynismaa, University of Helsinki, Finland Hibiki Kawamata, Cornell University, USA

#### \*Correspondence:

Stefania Corti, Dino Ferrari Centre, Neuroscience Section, Department of Pathophysiology and Transplantation (DEPT), University of Milan, Neurology Unit, IRCCS Foundation Ca' Granda Ospedale Maggiore Policlinico, via Francesco Sforza 35, 20122 Milan, Italy stefania.corti@unimi.it

> Received: 14 June 2015 Accepted: 11 August 2015 Published: 25 August 2015

#### Citation:

Tafuri F, Ronchi D, Magri F, Comi GP and Corti S (2015) SOD1 misplacing and mitochondrial dysfunction in amyotrophic lateral sclerosis pathogenesis. Front. Cell. Neurosci. 9:336. doi: 10.3389/fncel.2015.00336 earlier onset respect to sporadic probands. To date, the genetic defect underlining ALS is identified in approximately two thirds of fALS cases, while nine of ten sporadic forms lack a molecular diagnosis (Renton et al., 2014). The application of massive parallel sequencing techniques to undiagnosed cases is expected to improve the diagnostic yield and the identification of novel genes in fALS. It is more difficult to predict if sporadic forms result from undisclosed (rare) molecular defects or are the consequence of additional factors such as environment and ageing. Indeed, the existence of multifactorial causes underlining sALS is a rapidly emerging hypothesis (Robberecht and Philips, 2013). More than 30 genes have been found mutated in ALS patients but few of them are epidemiologically relevant and currently considered for genetic testing (Renton et al., 2014). They include: SOD1, encoding for the copper/zinc superoxide dismutase 1, TAR DNA binding protein (TARDBP), fused RNA binding protein (FUS) and C9orf72, containing a pathological hexanucleotide repeat expansion which constitutes the most frequent molecular cause of sALS/fALS, towering over all other defects so far described (Renton et al., 2014).

Despite considerable efforts, the pathogenesis of ALS is still obscure. Multiple intracellular pathways have been proposed as relevant: the regulation of RNA transcription and editing, protein modification, folding and clearance, axonal transport, organelles maintenance, cell death mechanisms (Robberecht and Philips, 2013; Bucchia et al., 2015; Cozzolino et al., 2015).

The co-existence of multiple abnormalities and the plethora of results, sometimes contradictory, used to support or refuse pathogenetic hypothesis give rise to an apparently ''chaotic'' scenario, reflecting our limited comprehension of how the identified molecular defects result in motor neuron pathology and clinical onset. This confusing background is not surprising since many of the genes involved in ALS encode for proteins displaying multiple functions. The intracellular localization of the mutated proteins might also influence general protein expression and turnover in a cellspecific manner. In this regard, ALS has been considered a polygenic and multifactorial disorder, where derangement of multiple pathways might produce similar phenotypes sharing clinical, instrumental, and prognostic features (Robberecht and Philips, 2013). Since fALS and sALS also share common defects and similar alterations have been found in ALS animal models and patients' tissues, the existence of conserved mechanisms underlining motor neuron pathology has been proposed.

SOD1 mutations have been the first molecular defects described in fALS forms with dominant inheritance (Rosen et al., 1993) and most of the research studies addressing ALS pathogenesis and the development of therapeutic strategies have been performed using SOD1 transgenic models (Gurney et al., 1994). Recent discoveries in ALS genetics acknowledge the role of proteins involved in RNA metabolism and cytoskeletal organization, often enriched in central nervous tissues. Conversely, the expression of SOD1 is ubiquitous and not restricted to nor increased in spinal cord and motor neurons. Further, SOD1 levels are not developmentalregulated. SOD1 enzyme shows a single, well-defined catalytic function, dealing with the detoxification of the superoxide specie inside cells. In the hunt for the ALS culprit, the loss of SOD1 enzymatic activity has been rapidly ruled out, since mutations preserving or abolishing SOD1 activity were found to cause ALS (Cozzolino et al., 2015). These elements alone indicate that the involvement of SOD1 in an adult onset motor neuron neurodegenerative disorder is not obvious.

The seminal observation of the co-localization of SOD1 with vacuolated mitochondria in a mouse model expressing human mutant SOD1 (Jaarsma et al., 2001) advanced the hypothesis that SOD1 mutations might be toxic for mitochondria. These studies have been replicated and expanded in several experiments performed in tissues and cells obtained from ALS animal models and patients (Palomo and Manfredi, 2015). Mitochondrial impairment has been observed in major neurodegenerative disorders (Schon and Przedborski, 2011). Although motor neurons are highly reliant on oxidative phosphorylation, the relevance of mitochondria in ALS pathogenesis has been under evaluated for many years. However, signs of mitochondrial dysfunction have been observed in multiple ALS patients (Cozzolino et al., 2015) and tissues, including muscle (Corti et al., 2009). In a review of a large series of muscle biopsies, we found that respiratory chain impairment is a common feature of ALS muscles, sometimes preceding the motor neuron pathology (Crugnola et al., 2010), consistent with experimental models (Luo et al., 2013).

Differently from Parkinson's disease (PD), where several genes involved in familial forms encode for mitochondrial proteins (De Rosa et al., 2015), the mitochondrial localization of products encoded by ALS-related genes is not exclusive. We previously described an out-of frame mutation in the gene encoding for a mitochondrial subunit of cytochrome c oxidase (COX) in a patient with severe muscle COX deficiency and a clinical phenotype resembling ALS (Comi et al., 1998). Similarly, primary mitochondrial disorders might present as motor neuron phenotypes (Ronchi et al., 2012). Recently, a missense mutation in CHCHD10, encoding for an intermembrane space (IMS) protein likely involved in mitochondrial cristae remodelling, was found to segregate with disease in a large French family displaying ALS with frontotemporal dementia (Bannwarth et al., 2014). Although additional CHCHD10 mutations were detected in patients with motor neuron disorders (Cozzolino et al., 2015), mitochondrial dysfunction has been only occasionally documented (Ronchi et al., 2015).

Physiologically, a small proportion of SOD1 protein is located in the mitochondrial IMS of yeast (Sturtz et al., 2001) and mammals (Okado-Matsumoto and Fridovich, 2001; **Figure 1A**). A protective antioxidant role for mitochondrial SOD1 has been hypothesized (Okado-Matsumoto and Fridovich, 2001) although without conclusive evidences. Conversely, whether SOD1 mutants accumulate inside mitochondria and the consequence of such accumulation have been largely investigated. Although aberrant SOD1 was observed within mitochondrial matrix (Vijayvergiya et al., 2005), two main submitochondrial localizations for mutant SOD1 are generally

acknowledged: the IMS and the outer mitochondrial membrane (OMM). In this review, we focus on the pathogenetic role of mutant SOD1 elicited by its misplacing in mitochondria, depicting the pathways ruling its targeting towards IMS or OMM. We will also comment on how these events relate to ALS pathogenesis, including cell selectivity and age of onset.

### Mitochondrial Localization of Non-Mutated SOD1

Human SOD1 maps to chromosomal location 21q22.11 and encodes for a 17 kDa protein (**Figure 1B**). The encoded product, SOD1, is an enzyme which catalyses the removal of superoxide according the reaction O<sup>2</sup> <sup>−</sup> + 2H<sup>+</sup> → H2O<sup>2</sup> (superoxide dismutation). Human genome encodes for two additional superoxide dismutases: SOD2, located inside mitochondrial matrix and the extracellular SOD3. There is no apparent evolutionary relationship between SOD1 and its isoenzymes. SOD1 is generally acknowledged as a cytosolic enzyme (Vehviläinen et al., 2014), but it is now accepted that a minor fraction (less than 5%) of total SOD1 can be found in the IMS, in lower (Sturtz et al., 2001) and higher eukaryotes (Okado-Matsumoto and Fridovich, 2001).

Mitochondrial respiratory chain, assembled inside the inner mitochondrial membrane (IMM), produces high amounts of reactive oxygen species (ROS) as a bypass product of the electron transfer reactions occurring in respiratory chain complexes (mainly complex I and complex III; Murphy, 2009). The release of superoxide anion on both sides of the inner membrane requires dedicated detoxification enzymes inside mitochondrial matrix (SOD2) and IMS (SOD1). Reactions catalysed by SOD enzymes result in the production of H2O2, another dangerous ROS able to react with Cytochrome C (Fe3+) producing the highly reactive species Oxoferryl-CytC CytC (Fe4+; Vehviläinen et al., 2014). Therefore H2O<sup>2</sup> must be carefully neutralized and cells express catalase, glutathione peroxidase and peroxiredoxins to process H2O<sup>2</sup> (Palomo and Manfredi, 2015). It has been proposed that, under mitochondrial stress, exceeding SOD1 activity might paradoxically boost the production of toxic ROS inside IMS by increasing H2O<sup>2</sup> production and consequently the deleterious cytochrome-c mediated peroxidation (Goldsteins et al., 2008).

To work properly, SOD1 requires post-translational modifications including the formation of intramolecular disulfide bonds to achieve a correct folding, the binding of zinc and copper metal ions, and the exposition of a hydrophobic region required for the organization of SOD1 monomers in dimers. These post-translational changes not only are necessary for SOD1 function but also influence its subcellular localization (**Figure 2**). Despite the lack of a mitochondrial localization signal, after cytosolic translation, a small amount of SOD1 might enter mitochondria in an unfolded state, taking advantage of the OMM translocator TOM (Kawamata and Manfredi, 2008; **Figure 3A**). The copper chaperone for superoxide dismutase (CCS) promotes the establishment of disulfide bonds and catalyses the insertion of copper metal into SOD1 apoenzyme (Culotta et al., 1997). After reaching its mature form, SOD1 also remains inside IMS, with coordinated metals and disulfide bonds now preventing the leakage back to the cytosol (Kawamata and Manfredi, 2008). CCS greatly contributes to SOD1 entrapment inside IMS through the establishment of a disulfide bond between cysteines C57 and C146 and favoring the copper insertion into the apoenzyme (Kawamata and Manfredi, 2008).

CCS has been also postulated to act as a redox sensor for the subcellular distribution of SOD1. High cytosolic oxygen concentration maintains CCS in the cytosol where it folds and helps apoenzyme SOD1 to achieve a mature form. Low oxygen concentrations (hypoxia) promote the translocation of CCS across the mitochondrial membrane. Inside IMS, CCS induces the oxidative (disulfide bonds) folding of SOD1 apoezyme and inserts the copper. The rationale for this mechanism is that the expected boost of mitochondrial respiratory chain activity, induced by hypoxia, will increase

the concentration of superoxide anion inside mitochondria: the enhanced translocation of CCS, promoting SOD1 maturation, might be part of a compensatory antioxidant strategy (Kawamata and Manfredi, 2008). Therefore the intracellular distribution (and activation) of SOD1 is intimately related to the intracellular location of its chaperone CCS. CCS also lacks a canonical mitochondrial localization signal and relies for its mitochondrial import on the so-called disulfide relay system (DRS), classically used by small cysteine-rich IMS proteins (Mesecke et al., 2005). This mechanism, identified in yeast 10 years ago (Mesecke et al., 2005), is quite conserved across evolution (Fischer and Riemer, 2013). The IMS proteins MIA40 and ERV1 are key players of this folding trap mechanism. When a substrate of the DRS, like apo-CCS, enters the IMS, it undergoes a MIA40-mediated oxidation involving the shifting of disulfide bonds between specific cysteine pairs (disulfide relay). This oxidation-mediated folding retains the DRS substrate inside IMS (Herrmann and Riemer, 2010). MIA40, now reduced, is re-oxidized by ERV1/ALR, a FAD-dependent sulfhydryl oxidase that transfers electrons to cytochrome C (Herrmann and Riemer, 2010). Therefore, respiratory chain is connected with the DRS, influencing the import of several mitochondrial proteins, including assembly factors involved in the assembly of respiratory chain and the maturation of inner membrane.

Several authors have demonstrated that CCS is a substrate of DRS, differently from SOD1 (Groß et al., 2011; Klöppel et al., 2011). However, a recent observation proposed that MIA40 is dispensable for oxidation-dependent import of CCS into mitochondria: CCS might directly promote its own oxidation, as well as the oxidation of SOD1 and other, still uncharacterised, substrates (Suzuki et al., 2013). These findings reinforce the link between SOD1 and its chaperone CCS in terms of activity and subcellular distribution. Interestingly, an alternative pathway for SOD1 mitochondrial translocation has been depicted in yeast, acknowledging the role of mitochondrial inner membrane organizing system (MINOS; Varabyova et al., 2013). The MINOS machinery is a recently identified protein complex involved in the organization of the IMM (Varabyova et al., 2013). Notably, CHCHD10 may be part of this complex, given its apparent function and structural similarities with other members of the MINOS system. Although the underlining mechanism is still unclear, the mitochondrial import of SOD1 is reduced in yeast mutants lacking MINOS core components. MIA40 is also fundamental for this alternative import pathway with a different role respect to that of oxidation scaffold described in the DRS (Varabyova et al., 2013). The dysfunction of DRS pathway and MINOS machinery result in clinical phenotypes with mitochondrial impairment as observed

deposition onto OMM. (A) Mutant SOD1 enters IMS through TOM. Inside IMS mutant SOD1 can precipitate in insoluble form (1) or undergo folding via Mia40 resulting in the establishment of funtional homodimers and heterodimers (2). Mia40 transfers electrons to Cytochrome C (CytC) via Erv1 (3). TOM, translocase of outer membrane; OMM, outer mitochondrial membrane; IMS, intermembrane space; IMM, inner mitochondrial membrane; CIV, complex IV cytochrome C oxidase; MutSOD1, mutant SOD1. (B) Misfolded SOD1 localization upon mitochondria can be regulated types. High cytosolic levels of soluble MIF (macrophage migration inhibitory factor) are able to prevent SOD1 mislocalization where its absence favor SOD1 deposition onto OMM. Misfolded SOD1 interactions with OMM proteins, such as Bcl-2 and VDAC, activate pro-apoptotic mitochondrial pathway leading to cell death. This interaction can be prevented by MIF. TOM, translocase of outer membrane; OMM, outer mitochondrial membrane; IMS, intermembrane space; IMM, inner mitochondrial membrane; CIV, complex IV cytochrome C oxidase; MutSOD1, mutant SOD1.

in patients harbouring mutations in core components (Di Fonzo et al., 2009) or substrates (Bannwarth et al., 2014) of these mechanisms.

### Accumulation of Mutant SOD1 in the Mitochondrial Intermembrane Space (IMS)

More than 150 SOD1 mutations have been associated with ALS, the large part being point mutations leading to amino acid substitutions (Synofzik et al., 2012). SOD1 mutations are dominantly inherited although recessive inheritance has been also described, especially for those variants that show low penetrance.

Accumulation, misfolding, aggregation, and precipitation of proteins seem to be common hallmark of neurodegenerative disorders (Taylor et al., 2002). SOD1 mutants so far analysed show different properties on the basis of the residues affected. As an example, several SOD1 mutants display no catalytic activity while dismutase activity is totally preserved in G93A, likely the most investigated of the ALS mutations. The conservation of catalytic activity seems also to reflect the conservation of a proper folding, assayed, before the advent of conformation specific antibodies (Rotunno and Bosco, 2013), by resistance to proteinase K digestion.

Different mutations might also show different propensities to accumulate or to aggregate in high molecular weight complexes. Presence of aggregates of mutated proteins in motor neurons is increasingly considered as a hallmark of ALS. Furthermore, a relationship between insoluble mutant SOD1 aggregates and expression of signs/symptoms has been demonstrated in many different rodent models (Johnston et al., 2000; Wang et al., 2002, 2003) in line with human findings (Kato, 2008). Despite considerable study, the mechanism that confers to SOD1 aggregates cell toxicity is still undefined. It has been hypothesized that mutant SOD1 soluble species are responsible for the disease onset (Zetterström et al., 2007; Karch et al., 2009) being, indeed, more toxic than insoluble aggregated SOD1 forms (Zetterström et al., 2007; Brotherton et al., 2012; Weichert et al., 2014). The intriguing problem of SOD1 aggregation is further complicated by the fact that experimental studies have been largely performed in lines of transgenic animals (or motor neuronal cells) with strong, although heterogeneous, overexpression of mutant proteins required to display pathological phenotype. The lack of physiological SOD1 expression and the underestimated expression of wild type alleles (normally present in almost all ALS patients) challenge the results obtained in these studies. The replication of these experiments in a patient-specific context might overcome these limitations.

Other manifestations of SOD1 mutations are protein misfolding and intracellular mislocalization. Soluble misfolded mutant SOD1 was found augmented in the mitochondria of the ALS spinal cord rodents (Zetterström et al., 2007), and the specific conformation of these forms could enhance their mitochondrial entrance. Antibodies that selectively recognize misfolded/non-native SOD1 have pointed out the link between misfolded SOD1 and mitochondria in the spinal cord of ALS rodents expressing human mutant SOD1 (Vande Velde et al., 2008; Brotherton et al., 2012). SOD1 misfolding seems to be related to different post translational mechanisms, involving disulfide bridge formation (Karch et al., 2009; Kawamata and Manfredi, 2010), incorrect C6 and C111 bond (Niwa et al., 2007) and loss of metal ion binding (Banci et al., 2008). These abnormal functions lead to altered SOD1 dimerization and to a biochemically instable tertiary structure (Banci et al., 2008, 2012). Several SOD1 mutants show instability in vitro (Rodriguez et al., 2002, 2005; Lindberg et al., 2005) or are found in a partially unfolded oligomerized status in mitochondria (Deng et al., 2006; Ferri et al., 2006). Remarkably, Synofzik et al. (2012) described a family showing slowly progressive ALS phenotype due to the L117V SOD1 mutation, which is similar to wild type SOD1 respect to stability and dismutase activity. These data are in line with previous findings reporting that even minor quantities of misfolded SOD1 are sufficient to prompt ALS (Jonsson et al., 2006). Indeed, overexpressed wild type SOD1 is capable to achieve an unstable (misfolded) state, causing motor neuron disease phenotype in rodents (Graffmo et al., 2013).

ALS-linked SOD1 mutations might also affect mitochondrial localization, facilitating the crossing of the OMM. Indeed, demetalated or unfolded SOD1 enzymes due to inefficient cytosolic maturation caused by ALS mutations might increase the fraction of intramitochondrial SOD1 (Field et al., 2003). Similarly, increased mitochondrial accumulation of SOD1 (both wild type and mutant) was observed in vitro after experimental copper depletion (Arciello et al., 2011).

The accumulation of SOD1 in mitochondria of motor neurons has been proposed as a possible explanation for the selective motor neuron degeneration occurring in ALS (Bruijn et al., 2004). Indeed, several groups have reported that mutant SOD1 co-localizes with mitochondria by immunocytochemical and biochemical studies. Although not exclusive, brain, spinal cord and motor neurons seem to be particularly exposed to mutant SOD1 accumulation and sensitive to the resulting mitochondrial dysfunction (Jaarsma et al., 2001; Higgins et al., 2002; Mattiazzi et al., 2002; Sasaki et al., 2004; Ahtoniemi et al., 2008). Notably in most of the experiments, the accumulation of SOD1 inside mitochondria paralleled the disease progression.

Forcing the expression of constructs containing wild type and mutant SOD1 in different cellular compartments of a mouse neuroblastoma cell line resulted in increased cell death and mitochondrial dysfunction only when mutant species were targeted to mitochondria (Takeuchi et al., 2002). Specifically, the accumulation of SOD1 in IMS seems toxic. In cultured motor neuronal cells, obligate SOD1 expression in IMS leads to mitochondrial toxicity and cell death, giving results not dissimilar to those collected expressing untargeted SOD1 mutants (Cozzolino et al., 2009; Magrané et al., 2009). Although the mitochondrial environment induced abnormal oligomerization of wild type SOD1, mutant SOD1 tends to form insoluble aggregates inside mitochondria, leading to alterations in mitochondrial morphology, function and engagement of apoptosis. The removal of the cysteines involved in disulphide bond by genetic manipulation counteracted these changes (Cozzolino et al., 2009).

A mouse line expressing G93A SOD1 mutant targeted to IMS displayed neurological deficits although less evident compared to classical G93A mouse model (lacking mitochondrial targeting but presenting a stronger expression of the transgene; Igoudjil et al., 2011). Interestingly, the selective targeting of wild type SOD1 in IMS of the knockout SOD1 mouse prevented the development of motor neuropathy and preserved axonal maintenance. The rescue effect was associated with reduced oxidative stress inside mitochondrial compartment (Fischer et al., 2011).

The groups of Jeffrey Elliot and Giovanni Manfredi authored several articles about the aberrant accumulation of mutant SOD1 inside mitochondrial IMS. First, Elliot's team observed that G93A mice developed accelerated neurological deficits when crossed with lines overexpressing CCS in CNS tissues. The double transgenic also displayed a striking reduction of age at onset and lifespan. Protein studies revealed a significantly increased of mitochondrial SOD1 content. Consistently, the dual transgenic mice expressing wild type SOD1 did not show any abnormalities with the exception of a modest increase of mitochondrial SOD1 import. Mouse overexpressing CCS displayed a normal phenotype, rejecting the hypothesis that an overloaded DRS was at the basis of the observed mitochondrial dysfunction. Notably the genetic ablation of CCS did not prevent motor neuron phenotype and mitochondrial accumulation of mutant SOD, supporting the existence of alternative pathways for mitochondrial import of SOD1 (Son et al., 2007). The impressive phenotypic effect of CCS overexpression in presence of G93A SOD1 does not seem related to SOD1 misfolding or aggregation. The G93A substitution makes SOD1 a less adequate substrate for CCS, preserving the physical interaction with the chaperone but impairing proper oxidative folding and metallation (Proescher et al., 2008). Notably, the levels of COX activity and assembly as well as protein levels of COX-subunits were selectively reduced in spinal cord from transgenic mice, even at a pre-symptomatic stage (Son et al., 2008) likely reflecting a disturbance in COX maturation. This finding demonstrates that the abnormal mitochondrial translocation of mutant SOD1 might affect respiratory chain, hindering mitochondria from inside. However, the same authors demonstrated that the functional and biochemical defects caused by CCS overexpression depend upon the redox state of the SOD1 mutant: while G37R paralleled the findings of G93A mutation, other substitutions (G86R and L126Z) did not (Son et al., 2009). These results had been partially anticipated by in vitro studies of Ferri et al. (2006) who had observed a shifting of the redox potential of several SOD1 mutants inside mitochondria. Interestingly, the manipulation of redox state of mitochondrial environment by overexpression of antioxidant enzymes glutaredoxin 1 and 2 improved SOD1 solubility and reduced mitochondrial anomalous localization (Cozzolino et al., 2008; Ferri et al., 2010). In particular, increased activity of glutaredoxin 2 inside mitochondrial matrix prevented the IMS aggregation of insoluble mutant SOD1 (Ferri et al., 2010). The functional rescue of mitochondria is likely mediated by a general antioxidant pro-survival effect of glutaredoxin 2 not requiring a direct interaction with mutant SOD1: indeed, glutaredoxin 2 and SOD1 show a different intramitochondrial segregation.

Kawamata and Manfredi (2008) used mammalian cells to confirm that SOD1 subcellular distribution is driven by CCS in vitro. They also highlighted the importance of cysteine residues in the CCS-mediated oxidation of mutant and wild type SOD1. Finally, they found that toxic properties of mutant SOD1, including propensity to aggregate and misfolding, could override CCS-dependent mitochondrial recruitment. Once inside IMS, mutant SOD1 can preserve its soluble form and build up homo and heterodimers with wild type SOD1 or aggregate with other mutant proteins and precipitate in insoluble form. According to SOD1 mutation, one way will prevail on the other; nevertheless, these aspects are not predictive of clinical phenotype (Vehviläinen et al., 2014). The untargeted overexpression of CCS in motor neuron-like cells improved the solubility of cytosolic mutant SOD1 and increased the formation of insoluble aggregates of G93A SOD1 targeted to mitochondrial compartment (Cozzolino et al., 2009). The mitochondrial import of mutant SOD1 was enhanced by exposing NSC34 cells to inflammatory cytokines, supporting the existence of not-cell autonomous factors able to influence SOD1 localization (Ferri et al., 2008). Therefore, mutant SOD1 may escape physiological control of mitochondrial localization, performed by CCS, and remain inside mitochondria exerting toxic effects. CCS translocation into mitochondria might help to entrap mutant or wild type SOD1 enzymes inside IMS, without any influence on their transport across the OMM.

### Deposition of Mutant SOD1 Onto the Cytoplasmic Layer of Outer Mitochondrial Membrane (OMM)

The results so far presented demonstrate that both wild type and mutant SOD1 can be recovered by mitochondrial fractions of ALS animal and cell models. It is likely that the final submitochondrial location of SOD1 is the IMS and that different pathways (dependent or not on CCS) might influence this subcellular repartition. Moreover, it is conceivable that mutant SOD1 can aggregate into the IMS, as in the cytosol, contributing to mitochondrial dysfunction. Interestingly, mutant SOD1 aggregation was also observed in the mitochondrial matrix of two transgenic mice lines and the recovered SOD1 resulted less active and folded respect to cytosolic counterpart (Vijayvergiya et al., 2005). However, mitochondrial matrix has not been further investigated as a target for mutant SOD1 toxicity in other ALS models.

While both mutant and wild type enzymes have been observed inside mitochondria, the OMM has been proposed to be a more selective target for mutant SOD1 only, explaining crucial aspects of ALS pathology such as cell selectivity and symptoms onset. Two research groups have mainly contributed to this topic. Pasinelli et al. (2004) used immunoprecipitation studies to demonstrate that SOD1 associates with the OMM protein Bcl-2 in the mitochondrial fractions obtained from cells expressing wild type and mutant SOD1. These findings were also confirmed in vivo. Wild type and mutant SOD interacted with different domains of Bcl-2 protein. Indeed, the authors showed that mutant SOD1 alters the physiological SOD1/Bcl-2 binding and primes the formation of high molecular weight complexes containing SOD1 and Bcl-2. Interestingly, experiments performed on G93A mice and autoptic samples from a patient harbouring the A4V SOD1 mutation showed that Bcl-2 is selectively sequestered by mutant SOD1 in mitochondria isolated from spinal cord, but not liver (Pasinelli et al., 2004). In a back-to-back paper, Don Cleveland's team supported these findings. Mutant SOD1 was preferentially recovered in the mitochondrial fraction purified by gradient centrifugation of spinal cord mitochondria from symptomatic animal models expressing mutant (G93A) but not wild type SOD1 (Liu et al., 2004). Mutant SOD1 was absent in mitochondrial preparations obtained from other tissues. The effect of CCS on mitochondrial distribution was ruled out by genetic ablation of this gene. The preferential association of mutant SOD1 with spinal cord mitochondria was also confirmed in autoptic samples, taking advantage of a selective antibody developed to recognize the epitope created by a SOD1 truncating mutation. In a second series of experiments, the authors isolated floating mitochondria, excluding the chance of co-precipitation or co-sedimentation of interfering SOD1 located inside cytosol or other organelles (Vande Velde et al., 2008). Again, they observed that SOD1 selectively tightly associate with mitochondria from spinal cord (not the brain) in transgenic mice in an age-dependent manner. Using alkalibased extraction protocol they investigate the nature of this association. Exploiting an antibody developed to decorate only the misfolded fraction of SOD1, they conclusively observe that the mutant SOD1 associated with mitochondria is in its misfolded state, while the remaining folded protein contribute to cytosolic or mitochondrial (IMS) SOD1 content. They proposed the voltage-dependent anion channel (VDAC) channel (mitochondrial porin) as the candidate docking protein for misfolded SOD1 deposition onto OMM. VDAC was found to co-precipitate only with misfolded SOD1. Symmetrically, mutant SOD1 deposited on VDAC only when inserted in the cytosolic face of a membrane reconstructed in vitro (Israelson et al., 2010).

The experiments discussed so far supports the selective targeting of spinal cord mitochondria and the existence of an interaction between SOD1 and OMM proteins (**Figure 3B**). However, the specific target of SOD1 onto OMM (Bcl-2 vs. VDAC1) and the SOD1 specie involved in this interaction (mutant only or wild type and mutant) differed according the two groups of investigators and were object of further studies.

Bcl-2 is an integral OMM protein, while VDAC1 is a mitochondrial voltage-dependent ion channel: both are involved in the intrinsic apoptotic process. Bcl-2 regulates the release of cytochrome c by the mitochondrial channel VDAC (Shimizu et al., 1999) and inhibits proapoptotic Bax signaling. Bcl-2 has been proposed as a direct mediator of mitochondrial toxicity of mutant SOD1 since it is required to induce cytochrome c release and change in mitochondrial morphology upon exposure to mutant, but not wild type, SOD1. Notably, this effect was only observed if Bcl-2 was tethered to OMM (Pedrini et al., 2010). The authors demonstrated in vitro and in vivo that mutant SOD1 alters Bcl-2 conformation, exposing its BH3 (death) domain, without which the toxic effect was ineffective.

VDAC1 acts as a gate for ions, metabolites, nucleotides, in two different dynamic conformation states. While open it allows anions, ATP and succinate flow from mitochondria to cytoplasm; when VDAC1 is closed it acts as a small cations channel. The impairment in VDAC1 function has been also linked with mitochondrial dysfunction since the drop of ATP/ADP ratio and membrane potential drives the increase of ROS production, hence contributing to extend the amount of misfolded SOD1 (Vande Velde et al., 2008; Israelson et al., 2010). Mass spectrometry analysis revealed more widespread changes in mitochondrial protein composition and significantly reduced mitochondrial import (Li et al., 2010).

We must observe that mitochondria from different tissues have different protein repertoires. This is true for OMM (Mootha et al., 2003; Bailey et al., 2007), which is also exposed to different cytoplasm compositions (Boillée et al., 2006; Vande Velde et al., 2008; Israelson et al., 2015). Tissue type, epigenetics regulation, gene transcription, oxygen concentration and ROS production may contribute to the specification of proteins recruited onto OMM (Israelson et al., 2010). Even considering these aspects, it is difficult to think at VDAC as a selective target for mitochondrial deposition of mutant SOD1 in spinal cord motor neurons, since it is widely expressed. Moreover, protein trafficking across OMM is partially rescued by other members of the VDAC family including VDAC2 and VDAC3. The hypothesis of a direct interaction between mutant SOD1 and VDAC has been recently challenged (Tan et al., 2013). According these observations, collected in vitro and in vivo, VDAC dysfunction is the result of a toxic association between mutant SOD1 and Bcl-2. Both SOD1 and VDAC compete for Bcl-2, which acts as a bridge between the two proteins. The conformational change in Bcl-2, induced by mutant SOD1 binding, exposes the toxic BH3 domain and favors the association of modified Bcl-2 with VDAC, resulting in a pathological alteration of OMM polarity and permeability, due to VDAC closure. This stream of mismatch associations seems to be a key event in disease initiation and progression. Interestingly, the administration of small peptides designed to target SOD1 and blocking its interaction with Bcl-2 might counteract these pathological alterations (Tan et al., 2013).

Collectively, the consequence of mutant SOD1 deposition onto OMM are: (i) conformational change of Bcl-2 preventing its anti-apoptotic association with Bax; (ii) impairment of VDAC1 function engaged by direct (SOD1-mediated) or indirect (Bcl-2 mediated) abnormal interactions with mutant SOD1, leading to changes in mitochondrial potential, morphology, protein composition and transport across membranes.

The specificity of mutant SOD1 in OMM targeting seems to be dictated by factors contained in the cytosol of the considered cell rather than its mitochondria. From these considerations, two hypotheses raised: it may exist a factor in spinal motor neurons that favors mislocalization, or, alternatively, there is a factor in the cytosol of other cells preventing it (Israelson et al., 2015). In a very recent publication, Cleveland investigated different heat labile factors that can block SOD1 mitochondria accumulation, selecting MIF as the most promising candidate. MIF (22q11.23) encodes for a small (12.5 kDa) protein, displaying multimeric and monomeric forms, with various functions: folding ATPindipendent chaperone, thiol-oxidoreductase, and secreted cytokine.

MIF interacts with mutant SOD1 by preventing its binding with OMM proteins and the accumulation of misfolded SOD1. Other chaperones (Hsp70, Hsc70, aB-crystallin, cyclophilin-A, glutathione peroxidase) had no effect in blocking SOD1 misplacing onto mitochondria outer membrane (Israelson et al., 2015). The inhibitory activity of MIF is preserved even in mutant lacking thiol-oxydoreductase activity. This is an important consideration since misfolding and mislocalization are supposed to be driven by the lack of C57-C146 bond (Vehviläinen et al., 2014) and by mitochondrial binding induced by C111 (Ferri et al., 2006), respectively.

At onset, SOD1 G93A mice showed reduced MIF levels in motor neurons as an effect of rapid protein clearance (given the demonstration of conserved MIF transcription and active translation): the restoration of MIF to normal levels enhanced motor neurons survival. Since reducing misfolded SOD1 aggregates does not halt disease progression, in vivo toxicity seems to be driven by soluble misfolded SOD1, which can be reduced by the increase of MIF in a dose-dependent manner (Israelson et al., 2015).

The results presented about the toxic deposition of misfolded SOD1 onto OMM and those coming from the identification of MIF as the soluble factor responsible for the selective neutralization of misfolded SOD1 are intriguing. In particular, the misfolded accumulation of mutant SOD1 seems to be conserved in both active and inactive mutants, the latter presenting greater amount of misfolded proteins. This finding nicely fits with a general observation about the more aggressive phenotype (earlier onset, faster progression) associated with inactivating mutations (Sato et al., 2005). At the same time, even small amount of mutant misfolded proteins might produce a clinical phenotype in human while high levels of expressions of mutant constructs (different according the mutations considered) are required to show a neurologic deficit in mice.

More importantly, these studies provide novel clues about the selectivity of spinal cord involvement and disease progression. These results were confirmed and refined in independent experiments, although the same authors disclosed misfolded SOD1 in mitochondrial fractions isolated from patients' lymphoblasts, challenging the cell type selectivity (Pickles et al., 2013). The identification of MIF as a candidate for ALS therapy is an intriguing discovery but this story is at its dawn. More confirmations are required to address unresolved issues including investigations on the degradative or secretive pathways acting in motor neurons to deprive intracellular MIF levels.

### Conclusion

Several groups demonstrated positive correlation between mutant SOD1 and ALS and the main pathogenetic hypothesis is a toxic gain of function. What is the toxic function gained and why it selectively strikes spinal motor neurons is still unexplained, but certainly not unexplored. One of the most investigated pathogenetic hypothesis underlining SOD1-related ALS focuses on mitochondrial toxicity of mutant SOD1. Fifteen years after the first observation on the co-localization of mutant SOD1 with vacuolated mitochondria in G93A mice spinal cord, the issue of SOD1 localization generated contradictory results but also promising answers. The mitochondrial localization of wild type and mutant SOD1 is a result definitively acquired and the hypothesis that the mutant function associated with ALS mutations involves the aberrant processing of superoxide anions is considered unlikely. Two major models have been advanced for mutant SOD1 accumulation inside mitochondria: increased content of mutant molecule, escaping physiological regulation inside IMS and selective deposition of misfolded protein onto OMM. These two routes are not mutually exclusive and studies supporting one model have also acknowledged the other as possible. The issue of submitochondrial localization hidden more relevant questions: which are the factors involved in SOD1 targeting towards mitochondria? Is the toxicity of mutant protein linked with aggregates (cytosolic or IMS), or soluble forms (SOD1 neutralized by MIF inside cytosol)? Are the mitochondria central for downstream ALS pathogenetic events or mitochondrial dysfunction adds to other intracellular mechanisms leading to motor neuron loss? Despite its key role in metabolism, mitochondria are involved in other fundamental cellular pathways, including calcium handling. Neuroblastoma cells expressing SOD1 mutants show higher levels of intracellular calcium compared to controls and are more vulnerable to further increase of calcium concentration (Goos et al., 2007; Jaiswal et al., 2009). These findings are supported by similar observations in G93A transgenic mice (Jaiswal and Keller, 2009). The reduction in calcium uptake does not seem the consequence of disturbed mitochondrial membrane potential. Conversely, it could be related to a differential expression of transporters and regulatory proteins involved in calcium handling, an hypothesis proposed to explain the selective vulnerability of spinal cord motor neurons to calcium (Fuchs et al., 2013).

Novel in vivo and in vitro experiments are required to investigate whether these mechanisms are directly influenced by altered IMS composition, primed by mutant SOD1 accumulation, or by impaired mitochondrial transport resulting from aberrant deposition of mutant SOD1 onto OMM.

In parallel, translational studies are mandatory to challenge the therapeutic potential of the targets observed to influence SOD1 misplacing.

### Acknowledgments

JPND (Joint Programme Neurodegenerative Disease) Research grant ''DAMNDPATHS'' (2014) and ARISLA grant ''smallRNALS'' (2014) to SC are gratefully acknowledged.

### References


mutant SOD1 in familial ALS. Neurology 65, 1954–1957. doi: 10.1212/01.wnl. 0000188760.53922.05


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Tafuri, Ronchi, Magri, Comi and Corti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Muscle Expression of SOD1G93A Modulates microRNA and mRNA Transcription Pattern Associated with the Myelination Process in the Spinal Cord of Transgenic Mice

Gabriella Dobrowolny 1, 2 †, Camilla Bernardini 3 †, Martina Martini 1, 2, Mirko Baranzini <sup>3</sup> , Marta Barba<sup>3</sup> and Antonio Musarò1, <sup>2</sup> \*

<sup>1</sup> DAHFMO-Unit of Histology and Medical Embryology, Institute Pasteur-Cenci Bolognetti, IIM, Sapienza University of Rome, Rome, Italy, <sup>2</sup> Center for Life Nano Science at Sapienza, Istituto Italiano di Tecnologia, Rome, Italy, <sup>3</sup> Institute of Anatomy and Cell Biology, Università Cattolica del Sacro Cuore, Rome, Italy

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Dumitru A. Iacobas, Albert Einstein College of Medicine of Yeshiva University, USA Deborah Baro, Georgia State University, USA Eva R. Chin, University of Maryland School of Public Health, USA

\*Correspondence:

Antonio Musarò antonio.musaro@uniroma1.it

† These authors have contributed equally to this work.

Received: 09 June 2015 Accepted: 13 November 2015 Published: 01 December 2015

#### Citation:

Dobrowolny G, Bernardini C, Martini M, Baranzini M, Barba M and Musarò A (2015) Muscle Expression of SOD1G93<sup>A</sup> Modulates microRNA and mRNA Transcription Pattern Associated with the Myelination Process in the Spinal Cord of Transgenic Mice. Front. Cell. Neurosci. 9:463. doi: 10.3389/fncel.2015.00463 A crucial system severely affected in several neuromuscular diseases is the loss of effective connection between muscle and nerve, leading to a pathological non-communication between the two tissues. One of the best examples of impaired interplay between muscle and nerve is Amyotrophic Lateral Sclerosis, a neurodegenerative disease characterized by degeneration of motor neurons and muscle atrophy. Increasing evidences suggest that damage to motor neurons is enhanced by alterations in the neighboring non-neuronal cells and indicate that altered skeletal muscle might be the source of signals that impinge motor neuron activity and survival. Here we investigated whether muscle selective expression of SOD1G93A mutant gene modulates mRNAs and miRNAs expression at the level of spinal cord of MLC/SOD1G93A mice. Using a Taqman array, the Affymetrix Mouse Gene 2.0 ST approach and the MiRwalk 2.0 database, which provides information on miRNA and their predicted target genes, we revealed that muscle specific expression of SOD1G93A modulates relevant molecules of the genetic and epigenetic circuitry of myelin homeostasis in spinal cord of transgenic mice. Our study provides insights into the pathophysiological interplay between muscle and nerve and supports the hypothesis that muscle is a source of signals that can either positively or negatively affect the nervous system.

Keywords: muscle-nerve interplay, miRNA and mRNA signature, ALS, myelination process, SOD1G93A

**Abbreviations:** ALS, Amyotrophic Lateral Sclerosis; ANOVA, ANalysis Of VAriance; Cldn19, Claudin 19; DAVID, Database for Annotation Visualization Integrated Discovery; Dhh, Desert hedgehog; DIANA, Dna Intelligence ANAlysis; dmax, maximum axonal diameter; Dmax, maximum fiber diameter; Dmin, minimum fiber diameter; dmin, minimum axonal diameter; Drp2, Dystrophin related protein 2; Egr2, Early growth response 2; FC, Fold Change; FVB, Friend leukemia virus B strain; GEO database, Gene Expression Omnibus database; Gldn, Gliomedin; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; Mag, Myelin-associated glycoprotein; Matn2, Matrilin 2; mIGF-1, muscle specific isoform of the Insulin Growth Factor-1; miRNA, miR, microRNA; miRPath, miRNA pathway; MLC, Myosin Light Chain; Mpz, Myelin protein zero; Nkx2-2, NK2 homeobox 2; Ogn, Osteoglycin; OMIM, Online Mendelian Inheritance in Man; Pmp22, Peripheral myelin protein 22; Prx, Periaxin; SDS, Sodium Dodecyl Sulfate; Smtn, Smoothelin; SOD1, Superoxide dismutase 1; STRING, Search Tool for the Retrieval of Interacting Genes/Proteins; TAC, Transcriptome Analysis Console; UTR, Untranslated region.

## INTRODUCTION

Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disorder in which the functional connection between nerve and muscle is severely compromised (Musarò, 2013). ALS affects pyramidal neurons in the motor cortex and lower motor neurons originating in the brainstem and in the spinal cord. The most typical feature of this progressive lethal disease is the degeneration of motor neurons, muscle weakness and atrophy, that leads to progressive paralysis and death by respiratory failure in few years (Musarò, 2010, 2013). ALS is epidemiologically classified into two major forms: sporadic (90–95%) and familial (5–10%) form. Among the familial cases, ∼20% are caused by dominantly inherited mutations in the Cu/Zn superoxide dismutase (SOD1) protein (Rosen, 1993) a cell scavenger of superoxide.

One of the critical issue that remains to be addressed in ALS is whether motor neurons are the first and sole direct targets of ALS or if the toxic activity of mutant SOD1 is cell autonomous. Recent experimental evidence revealed that ALS is a multi-factorial and multi-systemic disease in which alterations in structural, physiological, and metabolic parameters in different cell types, such as motor neurons, glia, and muscle cells, may act synergistically to exacerbate the disease (Boillée et al., 2006; Julien, 2007; Musarò, 2013). Skeletal muscle is a source of signals that influence neuron survival, axonal growth and maintenance of synaptic connections. Notably, development in the absence of skeletal muscle results in the sequential ablation of motor neurons from the spinal cord to the brain (Kablar and Rudnicki, 1999); thus, effective connection between muscle and nerve is crucial for the survival and function of both tissues throughout life. We have previously demonstrated that muscle-specific expression of SOD1 mutation in the MLC/SOD1G93A transgenic mouse model, which selectively expresses the mutant SOD1G93A gene in the skeletal muscle under the transcriptional control of a muscle-specific promoter (MLC), induces accumulation of Reactive Oxygen Species, causes muscle atrophy with a concomitant alteration in the ultrastructure and in the functional performance of skeletal muscles, and promotes microglia activation in the spinal cord, that is a presymptomatic sign of ALS disease (Dobrowolny et al., 2008b). Moreover, it has been reported that skeletal muscle-restricted expression of the human mutant SOD1 gene causes motor neuron degeneration in old transgenic mice (Wong and Martin, 2010) and that muscle-selective alterations in mitochondrial function initiate neuromuscular junction alteration, distal axonopathy, astrocytosis, and mild motor neuron loss (Dupuis et al., 2009). We also demonstrated that forced expression of muscle specific isoform of the Insulin Growth Factor-1 (mIGF-1) exclusively in the skeletal muscle of the global SOD1G93A mice counteracted the symptoms of ALS, induced satellite cell activation, stabilized neuromuscular junctions, preserved motor peripheral nerve, and led to a reduction in astrocytosis in the SOD1G93A spinal cord (Dobrowolny et al., 2005, 2008a). All together these data indicate that nervous development and homeostasis are intimately coupled to skeletal myogenesis and muscle function. Moreover, these findings support the hypothesis that skeletal muscle is a primary target of mutant SOD1 toxicity in mice and indicate that altered skeletal muscle impairs motor neuron activity, suggesting a sequential pattern of degeneration in which muscle abnormalities precede motor neuron death. In order to add new insights into the pathogenesis of ALS and define whether skeletal muscle alteration potentially affect the nervous system we analyzed the mRNA and miRNA expression in the lumbar ventral spinal cord of MLC/SOD1G93A transgenic mouse model (Dobrowolny et al., 2008b). Interestingly, we found that muscle specific expression of SOD1G93A gene affects spinal cord miRNome and transcriptome; in particular we observed the activation of genes involved in the myelination process and a decrease in the axon diameter to total fiber diameter in the sciatic nerve, suggesting a myelinopathy in the transgenic mouse model.

### MATERIALS AND METHODS

### Mice

Animal model used: 4 month old MLC/SOD1G93A mice overexpressing the mutant SOD1 gene (SOD1G93A) under the control of the Myosin Light Chain (MLC) muscle specific promoter (Dobrowolny et al., 2008b) and 4 month old FVB (Friend leukemia virus B) (control strain). The animals were housed in a temperature-controlled (22◦C) room with a 12:12 h light-dark cycle. All animal experiments were approved by the ethics committee of the Unit of Histology and Medical Embryology-Sapienza University of Rome- and were performed in accordance with the current version of the Italian Law on the Protection of Animals.

### RNA Isolation

Total RNA was isolated from frozen lumbar spinal ventral tissue specimens using pestel homogenization, TRIzol reaction (Invitrogen, Carlsbad, CA, USA), and further on-column purification as previously described. The yield, quality and integrity of RNA were determined using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) as previously described (Bernardini et al., 2012).

### miRNA Analysis

From each sample (n = 4 per experimental group), 600 ng of total RNA was reverse transcribed using Taqman MicroRNA Reverse Transcription Kit with Megaplex Primer Pools A & B (Applied Biosystems, Foster City, CA, USA). The cDNA was analyzed with Taqman Array Mouse MicroRNA A and B Cards Set version 3 (Applied Biosystems, Foster City, CA, USA) on the Applied Biosystems 7900HT qPCR system (Applied Biosystems, Foster City, CA, USA). Most of the MicroRNAs in the B Card were undetermined, therefore their analysis was omitted. The data was normalized by global mean normalization. SDS RQ Manager version 2.4 (Applied Biosystems, Foster City, CA, USA) was used to calculate Ct values (number of cycles required for the fluorescent signal to cross the set threshold), Ct values greater than 35 were considered as non-specific or undetected, and were filtered. MicroRNAs with missing values in at least 2 of 4 replicate samples per treatment group were removed from the dataset. The remaining Ct values were normalized using the U6 reference miRNA as suggested by manufacturer's protocol. The fold change between sample groups was calculated by subtracting the relative expression of the samples and then transforming the result to anti-log2 (i.e., 2difference).

Relevant information about the experiment, is available at http://www.ncbi.nlm.nih.gov/geo/ under accession GSE71193.

#### miRNA Pathway Analysis

The putative correlation between mRNAs and miRNAs was investigated through the MiRwalk 2.0, a comprehensive database that provides information on miRNA from Human, Mouse and Rat on their predicted as well as validated binding sites on their target genes. In particular we get information on miRNA produced by 8 established miRNA prediction programs on 3 ′ UTRs of all known genes of Mouse (Dweep et al., 2011; Bernardini et al., 2012).

miRNA predicted pathways were identified using DIANA (DNA Intelligence Analysis) tools (Vlachos et al., 2012).

DIANA-miRPath is a miRNA pathway analysis web-server that utilizes predicted miRNA targets (in CDS or 3′ -UTR regions) provided by the DIANA-microT-CDS algorithm or even experimentally validated miRNA interactions derived from DIANA-TarBase v6.0. These interactions (predicted and/or validated) can be subsequently combined with sophisticated merging and meta-analysis algorithms. Essentially, this tool allows identifying all the predicted or experimentally validated miRNAs significantly targeting a selected pathway.

#### Single qPCR Validation

Single qPCR of miRNA 133a, 133b, 9, 29, 330, and 1 was performed to validate the performance of the Taqman arrays using Taqman probes from single tube Taqman microRNA assays with the diluted cDNA in a 10µl reaction volume, in triplicate for each assay, on the Applied Biosystems 7500HT system (Applied Biosystems, Foster City, CA, USA; n = 10 per experimental group).

#### Western Blot Analysis

Protein extraction from both wild type and MLC/SOD1G93A transgenic spinal cord (n = 5 for each genotype) was performed in 10 mM Tris, 150 mM Sodium Chloride, 1% NP40, 0.1% SDS, 10% Glycerol, 1% Deoxycholate, 1 mM Phenylmethylsulfonyl Fluoride, 1µg/ml Aprotinin, 1µg/ml Leupeptin, 1µg/ml Pepstatin, 1 mM Sodium Orthovanadate, and 1 mM Sodium Fluoride. Equal amounts of protein from each lysate were separated in SDS polyacrilamide gel and transferred onto a nitrocellulose membrane. Filters were blotted with antibodies against Pmp22 (Novus Biological Littleton, CO, USA. Cat. NBP1-67670), Mpz (Abcam, Cambridge, UK. Cat. Ab64685), and a-tubulin (Sigma, Saint Louis, MO, USA. Cat. T5168).

Pmp22 primary antibody was diluted 1:200 in TBST and used for blotting over night (o.n.) after 5% milk saturation. Then, filter was incubated with secondary antibody (Goat anti-Rabbit IgG HRP-conjugated) (Bethyl, Montgomery, TX, USA. Cat. A120- 201P) in 1% milk for 1 h.

Mpz primary antibody was diluted 1:500 in TBST and used for blotting o.n. after 10% milk saturation. Then, filter was incubated with secondary antibody (Goat anti-Mouse IgG HRPconjugated) (Bethyl, Montgomery, TX, USA. Cat. A90-516P) in 1% milk for 1 h.

Tubulin was diluted 1:5000 in TBST and used for blotting o.n. after 5% milk saturation. Then, filter was incubated with secondary antibody (Goat anti-Mouse IgG HRP-conjugated) (Bethyl Montgomery, TX, USA. Cat. A90-516P) in 1% milk for 1 h. Signals were acquired by ChemiDoc MP instrument (Bio-Rad, Hercules, CA, USA) and processed with Image Lab acquisition analysis software (Bio-Rad, Hercules, CA, USA).

Five independent samples for each group of animals have been used for densitometric analysis and protein level of α-tubulin was used as control for equal protein loading.

#### Microarray Analysis

Total RNA was used to create the biotin-labeled library to be hybridized on GeneChip <sup>R</sup> Mouse Gene 2.0 ST Array (Affymetrix, Santa Clara, CA, USA) covering more than 26500 RefSeq coding transcripts and more than 3500 RefSeq noncoding transcripts, following the manufacturer protocol.

Briefly, double-stranded cDNA was synthesized routinely from less than 1 microgram of total RNA primed with a poly- (dT)—T7 oligonucleotide. The cDNA was used in an in vitro transcription reaction in the presence of T7 RNA polymerase and biotin-labeled modified nucleotides during 16 h at 37◦C. Biotinylated cRNA was purified and then fragmented (35–200 nucleotides), together with hybridization controls and hybridized to the microarrays for 16 h at 45◦C. Using the GeneChip Fluidics Station 450 (Affymetrix, Santa Clara, CA, USA), the biotin-labeled cRNA was revealed by successive reactions with streptavidin R-phycoerythrin conjugate, biotinylated anti-streptavidin antibody and streptavidin R-phycoerythrin conjugate. The arrays were finally scanned in an Affymetrix GeneChip Scanner 7G Plus (Affymetrix, Santa Clara, CA, USA).

### Gene Expression Data Analysis—Differentially Expressed Gene List

The CEL files that store the results of the intensity calculations on the pixel values collected from an Affymetrix scanner and result from the hybridization, were analyzed using Transcriptome Analysis Console (TAC) from Affymetrix (Affymetrix, Santa Clara, CA, USA). Gene-level calculation was performed by Robust Multichip Average to summarize probeset signal (Irizarry et al., 2003) and normalization by quantile sketch (Bolstad et al., 2003). A data table (rma), together with the relative CEL and relevant information about the experiment, is available at http:// www.ncbi.nlm.nih.gov/geo/ under accession GSE69582.

An unpaired One-way ANOVA was then used to identify differentially expressed genes, only the genes which met our criterion (p < 0.05, fold change > 1.5) were selected in this study. The resulting gene list was then annotated according to the Gene Ontology (GO) database (www.geneontology.org). This allowed assigning a category to each gene in the list, according to three defined "ontologies" (i.e., terms representing gene product properties): cellular component, biological process and molecular function.

The expression of selected genes was quantified in real time PCR to obtain an independent validation of microarray data. Real time PCR was carried out as previously described elsewhere (Bernardini et al., 2012).

#### Functional Categorization

The gene expression list has been functionally categorized using the Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/) (Sherman et al., 2007). The algorithms implemented in this software allow identifying over-represented gene ontology (GO) terms with respect to the total number of genes assayed and annotated. To this aim, DAVID applies a modified Fisher exact test, to establish if the proportion of genes falling into an annotation category significantly differs from the background group of genes. In addition, this tool enables the fine mapping of genes within welldefined signaling and/or metabolic pathways, classified in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (www.genome.jp/kegg/). The KEGG mapping tool was employed for the functional categorization of the gene regulatory networks. For this purpose, AffyGene IDs, corresponding to the genes in the selected list, were used as queries and the whole set of genes represented on the array was used as the background group. A p < 0.05 was set.

The STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) online software (Franceschini et al., 2013) was used to search interaction relationships of the proteins encoded by differential expressed genes.

#### Histological Analysis

Sciatic nerves were dissected and quickly fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate for 2 h, washed in 0.1 M sodium cacodylate. Nerves were then post-fixed in 1% osmium in 0.1 M sodium cacodylate for 1 h, dehydrated in ethanol and embedded in epoxy resin. Samples were infiltrated with toluene and cut with a Leica ultracut R ultramicrotome (Leica Microsystems Inc. Buffalo Grove, IL) to obtain 2 um suitable for toluidine blue staining. Microscope using 40 × lenses was employed, and images were processed using Axiovision Rel 4.8 (Zeiss, Oberkochen, D) representative images of sciatic nerve thin sections are shown cropped in **Figure 5**.

The g-ratio was calculated as (dmax + dmin)/2/(Dmax + Dmin)/2 and the analysis was performed for 200 fibers. Each diameter was calculated from the average of the major axis (Dmax: maximum fiber diameter; dmax:maximum axonal diameter) and the minor axis (Dmin: minimum fiber diameter; dmin: minimum axonal diameter), (Tsutsumi et al., 2014) and was measured using Image J software (Image J 1.36, National Institutes of Health, MD, USA).

#### Statistical Analysis

Statistical analysis, if not differently specified was performed using the Mann Whitney nonparametric test and significance was established for p < 0.05.

#### RESULTS

### miRNA and mRNA Expression Profile of the Spinal Cord of the MLC/SOD1G93A Mice

First, to identify deregulated miRNAs and transcripts in the spinal cord of MLC/SOD1G93A transgenic mice we performed miRNA expression profiling and DNA microarray, using the Taqman array card and Affymetrix Mouse Gene 2.0 ST respectively (GSE71194). Among the analyzed miRNAs, 54 were significantly down-regulated (p < 0.05) in the spinal cord of MLC/SOD1G93A transgenic mice compared to wild type littermates (**Table 1**). Of note, the miRnome profiling revealed the down regulation of mir-330, mir-133, and mir-1, which are involved in denervation and reinnervation processes (Jeng et al., 2009; Tsutsumi et al., 2014). In addition mir-29 and mir-9 family member were downregulated and their deregulation was associated with different neurodegenerative diseases, including Huntington, Alzheimer, and Parkinson diseases (Saito and Saito, 2012; Tsutsumi et al., 2014) and demyelination-related diseases (Li and Yao, 2012; Tsutsumi et al., 2014). Real time PCR analysis was performed to validate the expression of these relevant miRNAs in the spinal cord of both wild type and MLC/SOD1G93A transgenic mice (**Figure 1**).

The microarray data, related to the expression of mRNA in the spinal cord of both wild type and MLC/SOD1G93A transgenic mice, were deposited in the GEO database (GSE69582) and the list of the selected significantly up-regulated and downregulated genes (p < 0.05; FC > 1.5) is shown in **Tables 2**, **3** respectively. Among these genes, nine up-regulated transcripts are involved in the myelination process such as Peripheral myelin protein 22 (Pmp22), Myelin protein zero (Mpz), Periaxin (Prx), the Early growth response 2 (Egr2) genes, Desert hedgehog (Dhh), or encode for extracellular matrix molecules such as Matrilin 2 (Matn2), Gliomedin (Gldn), Claudin 19 (Cldn19), and Smoothelin (Smtn). Other relevant modulated genes include glycoproteins, such as Osteoglycin (Ogn) and dystroglicans, such as Dystrophin related protein 2 (Drp2), associated with the myelination process. The Mpz, Pmp22, Egr2, Prx mRNAs were selected for validation by qRT-PCR analysis and results are shown in **Figure 2**. Opalin, another gene associated to myelin structures, was down-regulated in the spinal cord of MLC/SOD1G93A transgenic mice, compared to wild type littermates (**Table 3**). Opalin is expressed specifically in late stage of oligodendrocyte differentiation and has been shown to be dramatically reduced in a hypomyelination mouse model (Jiang et al., 2013; Tsutsumi et al., 2014).

In order to show changes in protein expression, as a result of altered microRNA/mRNA expression, we performed western blot analysis for Pmp22 and Mpz expression. We revealed a significant up-regulation of Pmp22 and Mpz proteins in the spinal cord of MLC/SOD1G93A compared to wild type littermates (**Figure 3**), further supporting the evidence that these factors are targets of specific microRNAs that control the denervation and reinnervation processes.

Taken together these results suggest that the muscle specific expression of the mutated isoform of the human SOD1 gene alters the expression of relevant markers of myelination in

#### TABLE 1 | List of significantly down-regulated miRNAs in MLC/SOD1G93A spinal cord.

miRNA Fold change P-value mmu-let-7b −4.97 0.04170 mmu-let-7c −4.80 0.01392 mmu-let-7d −3.85 0.02516 mmu-let-7i −3.92 0.04791 mmu-miR-100 −4.31 0.02747 mmu-miR-1 −3.69 0.01614 mmu-miR-106a −5.23 0.02034 mmu-miR-106b −4.57 0.03639 mmu-miR-10a −4.07 0.01796 mmu-miR-10b −3.75 0.03798 mmu-miR-128a −4.69 0.02756 mmu-miR-129-3p −3.55 0.03832 mmu-miR-129-5p −7.58 0.00264 mmu-miR-133a −2.75 0.03894 mmu-miR-133b −3.94 0.00340 mmu-miR-15a −8.16 0.00251 mmu-miR-15b −3.84 0.02226 mmu-miR-17 −4.43 0.04014 mmu-miR-181a −4.35 0.02206 mmu-miR-181c −7.37 0.01568 mmu-miR-188-5p −4.20 0.03755 mmu-miR-192 −3.38 0.04185 mmu-miR-194 −3.60 0.01944 mmu-miR-196b −3.24 0.04461 mmu-miR-199a-3p −4.67 0.01722 mmu-miR-19b −5.56 0.03463 mmu-miR-20b −6.42 0.02596 mmu-miR-23b −4.66 0.04223 mmu-miR-26b −5.25 0.01406 mmu-miR-29a −4.21 0.01571 mmu-miR-29b −6.33 0.01723 mmu-miR-29c −4.38 0.01636 mmu-miR-301a −4.21 0.04486 mmu-miR-30a −3.89 0.03876 mmu-miR-30e −3.37 0.03622 mmu-miR-324-5p −4.63 0.04475 mmu-miR-328 −3.75 0.04278 mmu-miR-329 −4.79 0.01774 mmu-miR-330 −8.50 0.00381 mmu-miR-340-3p −4.96 0.02453 mmu-miR-340-5p −3.83 0.03845 mmu-miR-369-3p −17.18 0.04086 mmu-miR-369-5p −3.64 0.02940 mmu-miR-491 −4.04 0.02354 mmu-miR-497 −6.93 0.00462 mmu-miR-501-3p −3.23 0.00474 mmu-miR-532-5p −4.21 0.04129 mmu-miR-544 −3.89 0.01407 mmu-miR-652 −5.73 0.02815

#### TABLE 1 | Continued



(Continued)


#### TABLE 2 | Continued


the ventral root of the lumbar spinal cord of transgenic mice.

### Functional Annotation of the Identified Genes

Functional annotation of differentially expressed genes was performed using the DAVID functional annotation tool, an integrated biological knowledge base and analytic tools aimed at systematically extracting biological meaning from large gene and protein lists (Huang da et al., 2009). Similar annotation contents were clustered into annotation group with biological meanings. The annotation results revealed that 2.6% of the gene list examined was consistent with OMIM disease categories, in particular 5 of 17 genes extracted were related to Charcot-Marie-Tooth disease. On the other hand, the gene ontology analysis revealed that 39.3% of all gene list are significantly involved in biological processes category, 36.7% are part of cell components, and 39.8% are associated with molecular function categories. The nine interesting genes mentioned above were spread out in all the categories examined. The functional annotation clustering was done using the default parameters, Mus musculus as background and classification stringency was set as medium. **Table 4** shows that the significantly modulated genes were involved in 5 functional clusters containing functional annotation for extracellular region, neuron development and differentiation, axonogenesis, and neurological system process. A further functional analysis was performed using STRING, a web base tool to explore potential protein-protein interaction,



confirming the interactions among Pmp22, Mpz, Prx, and Egr2 proteins (**Figure 4**).

### Correlation of Expression Profile between miRNAs and mRNA

miRNAs are key regulators of all important biological processes and the level at which they act on coding mRNA (transcriptional, translational, etc.) is still debated; however gene repression by microRNAs at the level of mRNA translation is the most frequently reported mechanism (Morozova et al., 2012). Therefore, we further performed a systematic analysis on the interaction between significantly modulated miRNAs and mRNAs considering only miRNAs present in our list with an inverse mRNAs expression correlation. The analysis was performed using MiRwalk tool a comprehensive database that provides information on miRNA produced by 8 established

miRNA prediction programs on 3′ UTRs of all known genes of mouse such as RNA22, miRanda, miRDB, TargetScan. Secondly, we ordered mRNAs on the basis of the number of putative targeting miRNAs and we observed that 36 genes were targeted from at least 1 to maximum 30 significantly modulated miRNAs; conversely 54 listed miRNAs targeted at least one to maximum of 11 significantly modulated genes; interestingly the most intriguing cross interaction involved genes and miRNAs of the

\*p < 0.05 compared to Wt (n = 4 for each group).

myelination process (**Table 5**). This finding clearly suggests that muscle specific overexpression of the SOD1 mutant gene induces the modulation of miRNAs and putative targeted-mRNAs involved in the

FIGURE 3 | Validation of selected factors by western blot analysis. Graphs and representative western blot for (A) Pmp 22 and (B) Mpz. White bar refers to wild type (Wt) and black bar to MLC/SOD1G93A (Tg). Values are expressed as mean ± SEM. \*p < 0.05; \*\*p < 0.005 compared to Wt (n = 5 for each group).

myelination process in the spinal cord of MLC/SOD1G93A transgenic mice.

### Mapping miRNAs to Signaling Pathways

DIANA-mirPath is a web-based computational tool to identify KEGG signaling pathways regulated by miRNAs (see Materials and Methods section). The software compares each set of miRNA targets with all-known KEGG pathway and interestingly we observed high significance for axon guidance, glutamatergic synapse, and amyotrophic lateral sclerosis (ALS), meaning that these pathways are likely to be controlled by the altered miRNAs (**Table 6**).

### Muscle Specific Expression of SOD1 Mutant Gene Induces Hypomyelination in the Mice Sciatic Nerve

Different reports indicate that ALS might be associated with motor nerve fiber (Echaniz-Laguna et al., 2006; Rajabally and Jacob, 2008; Ahdab et al., 2013). Previously, we demonstrated that the peripheral nerve was severely compromised in the global SOD1G93A mutant mice, which displayed loss of Schmidt-Lantermann incisures, a disproportionately thick myelin sheath, abundant double onion bulb structures, and increased demyelinated axons (Dobrowolny et al., 2008a). To assess whether muscle specific expression of SOD1 mutant gene induces a peripheral alteration of axon myelination we calculated the ratio of axon diameter to total fiber diameter (g-ratio) (Michailov et al., 2004). The **Figure 5A** shows the cross-sections of wild type and MLC/SOD1G93A sciatic nerves stained with toluidine blue. The histological analysis of the peripheral nerve of the MLC/SOD1G93A displayed the presence of double onion bulb structures. Moreover, the g-ratio in the transgenic sciatic TABLE 4 | Functional annotation clustering of up-regulated mRNAs.


Annotation cluster means a group of terms having similar biological functions. Enrichment score is the rank of the annotation cluster biological significance. Count column refers to number of genes involved in the same category.

nerves was significantly increased (**Figure 5B**) independently of the axonal diameter (**Figure 5C**), indicating that muscle specific expression of SOD1 mutant gene induces hypomyelination in the sciatic nerve of transgenic mice.

## DISCUSSION

The functional interplay between muscle and nerve is crucial for the survival and function of both tissues throughout life and ALS represents one of the best examples of impaired interplay between the two tissues (Musarò, 2013). Although, there have been significant advances in our understanding of the biology of ALS, no consensus has emerged as to which cells, tissues and pathways are directly affected by mutant SOD1. Motor neurons degeneration and muscle atrophy are the major pathological processes associated with ALS consistent with the role of nerve activity in muscle homeostasis and remodeling (Munson et al., 1997; Pette, 2001; Schiaffino et al., 2007). However, other cells may be involved in the pathogenesis of ALS and alteration in skeletal muscle homeostasis might represent one of the critical mediators of motor neuron degeneration. This hypothesis has been recently investigated by different groups, and we

recently demonstrated that muscle selective expression of SOD1 mutation causes pathological alterations in skeletal muscle and induces pre-symptomatic sign of ALS at the level of spinal cord (Dobrowolny et al., 2008b).

In the present study, we extended our previous work and analyzed whether muscle-specific expression of SOD1G93A is able to affect the expression of non-coding RNA and mRNA in the spinal cord of MLC/SOD1G93A transgenic mice, compared with wild type littermates. Our findings are in agreement with previous studies showing that skeletal muscle is also a source of signals that influence neuron survival, axonal growth and maintenance of synaptic connections (Funakoshi et al., 1995; Dobrowolny et al., 2005). In the absence of trophic support, muscle can have a negative impact on the nervous system, and therefore, can contribute to the alteration in the functional connection between muscle and nerve.

The major findings of our study indicate that muscle-specific expression of SOD1G93A can modulate relevant molecules of the genetic and epigenetic circuitry of the myelin homeostasis in the spinal cord of transgenic mice. In particular, we found down regulation of mir-1, mir-330, mir-29, mir-133, and mir-9 family members, whose dysregulation can have profound effects on neuronal physiology and pathology, including Huntington, Alzheimer, and Parkinson diseases (Saito and Saito, 2012). Although, it is known that deregulation of specific miRNAdependent regulatory circuitries correlates with the initiation and progression of several neurological disorders, the underlying mechanisms of these phenomena are still not fully understood.

Here we revealed the alteration of specific microRNAs and potential mRNA targets, which might be part of the short circuit that disrupts the communication between skeletal muscle and nerve and that might be part of the so called dying back

#### TABLE 5 | List of selected genes and corresponding regulated miRNAs.


Schematic representation of the inversely correlated miRNAs and predicted targets based on Mirwalk analysis.

phenomenon (Dadon-Nachum et al., 2011). In particular, among transcripts that are potential targets of the altered miRNAs, we revealed the presence of relevant markers of myelin homeostasis that are modulated by muscle restricted expression of SOD1 mutation inducing hypomyelination of the peripheral nerve. The up-regulation of Pmp22 and Mpz proteins in the spinal cord of MLC/SOD1G93A paralleled that of mRNA expression and supports the evidence that these factors are molecular targets of microRNAs, such as miR-1, miR-9, miR-133, and miR-330, that resulted differently modulated in the spinal cord of MLC/SOD1G93A mice compared to wild type littermates.

#### TABLE 6 | Signaling pathways predicted to be regulated by miRNAs.


It has been demonstrated that Myelin protein zero is upregulated in oligodendroglia following axonal injury (da Silva et al., 2015) and Periaxin is expressed by myelinating Schwann cells in association with the dystroglycan complex through Drp2 (Sherman et al., 2012). Moreover, Pmp22 exhibits a dose dependent function as both overexpression and deletion result in hereditary neuropathy (Han et al., 2013). Dhh is essential for the structural and functional integrity of the peripheral nerve (Sharghi-Namini et al., 2006) and Osteglycine, Claudin -19 and Smoothelin together with Pmp22 and Prx are associated with reduced myelination of the spinal cord in the hypogravity motor syndrome (Chelyshev et al., 2014). Moreover, mutations in Mpz, Pmp22, Egr2, and Prx are associated with Charcot–Marie–Tooth neuropathy in which the over expression of these genes leads to defective myelination processes.

Recently, some specific miRNAs, such as miR-9, miR-23, and miR-29a, were found to participate in the regulation of oligodendrocyte differentiation and myelin maintenance, as well as in the pathogenesis of demyelination-related diseases. These

miRNAs control their target mRNAs and are involved in the pathogenesis of demyelination-related diseases (Lau et al., 2008; Verrier et al., 2009; Li and Yao, 2012).

Regarding the mRNA-miRNA interplay, most of the putative interactions proposed here have not been previously described and deserve further investigation. Interestingly, the functional interaction of miR-9 with Peripheral myelin protein 22 (Pmp22) mRNA has been already demonstrated (Lau et al., 2008). In particular, miR-9 is down-regulated during oligodendrocyte differentiation and its expression level inversely correlates with the expression of its targets Pmp22. Moreover, it has been described that the inhibition of endogenous miR-29 overrides the miRNA-mediated repression of Pmp22 cultured Schwann cells during both development and post-crush injury (Verrier et al., 2009). Therefore, these data suggest that myelin gene expression is regulated by miRNAs.

Here we observed that the muscle specific expression of the SOD1 mutant gene induces the deregulation of both mir-9

#### REFERENCES


and 29 together with myelin alteration in the sciatic nerve. These data have been validated by histological analysis of the sciatic nerve that shows a reduction in the myelin sheath in the MLC/SOD1G93A mice model. In this regard it is interesting to note that the classical oligodendrocyte markers were downregulated (considering a fold change < 1.5 olig2 FC = −1.3 p = 0.02; Nkx2-2 FC = −1.36; p = 0.008; Mag FC = −1.39, p = 0.01). Therefore the reduction in oligodendrogenesis could be correlated with the hypomyelination in the MLC/SOD1G93A mouse model.

In the spinal cord of SOD1G93A transgenic mice that ubiquitously express the mutant SOD1 gene, the miRNA-9 expression is up-regulated (Zhou et al., 2013). In this model there is extensive degeneration of gray matter oligodendrocytes in the spinal cord prior to disease onset; actually new oligodendrocytes were formed but they failed to mature, resulting in progressive demyelination (Kang et al., 2013). The difference in the expression levels of miRNA-9 between the global SOD1G93A and MLC/SOD1G93A mice is likely due to SOD1G93A gene ubiquitously expressed in all tissues, including muscle, motor neurons and glia. This might generate a synergistic toxic effect, leading to a more severe phenotype.

Although, most of the miRNA and mRNA inverted correlations proposed here have not been yet validated, the elevated number of potential interactions strongly points toward a role of skeletal muscle in myelin homeostasis. Nevertheless, although these results point toward some mechanisms of action of muscle specific expression of SOD1G93A at the level of spinal cord, the elevated number of potential targets of the analyzed miRNAs make these mechanisms only mere suggestions. Additional studies will define which cell types, at the levels of spinal cord of MLC/SOD1G93A mice, specifically exhibit changes in microRNAs and mRNAs.

Overall our study provides additional insights into the effects of muscle selective expression of SOD1G93A on nerve homeostasis and reveals the potential miRNA and mRNA signature associated with the dying back phenomenon.

#### ACKNOWLEDGMENTS

We thank C. Nicoletti for technical support, B. Lozanoska-Ochser for manuscript English editing. This work was supported by ASI (grant no. 2013-088-R.0), Fondazione Roma, Telethon (GGP14066), Nando Peretti Foundation, and PRIN (grant no. 2010R8JK2X). The authors do not have financial interest in relation to this submission.

Boillée, S., Yamanaka, K., Lobsiger, C. S., Copeland, N. G., Jenkins, N. A., Kassiotis, G., et al. (2006). Onset and progression in inherited ALS determined by motor neurons and microglia. Science 312, 1389–1392. doi: 10.1126/science.1123511


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Dobrowolny, Bernardini, Martini, Baranzini, Barba and Musarò. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Ultrastructural studies of ALS mitochondria connect altered function and permeability with defects of mitophagy and mitochondriogenesis

Riccardo Ruffoli <sup>1</sup> , Alessia Bartalucci <sup>1</sup> , Alessandro Frati <sup>2</sup> and Francesco Fornai 1,2 \*

<sup>1</sup> Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Pisa, Italy, 2 I.R.C.C.S., Neuromed, Pozzilli, Italy

#### Edited by:

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Danhui Liu, Wenzhou Medical College, China Jay Nadeau, McGill University, Canada Vladimir I. Titorenko, Concordia University, Canada

#### \*Correspondence:

Francesco Fornai, Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Via Roma 55, 56126 Pisa, Italy francesco.fornai@med.unipi.it

> Received: 30 June 2015 Accepted: 17 August 2015 Published: 01 September 2015

#### Citation:

Ruffoli R, Bartalucci A, Frati A and Fornai F (2015) Ultrastructural studies of ALS mitochondria connect altered function and permeability with defects of mitophagy and mitochondriogenesis. Front. Cell. Neurosci. 9:341. doi: 10.3389/fncel.2015.00341 The key role of mitochondria in patients affected by amyotrophic lateral sclerosis (ALS) is well documented by electron microscopy studies of motor neurons within spinal cord and brainstem. Nonetheless, recent studies challenged the role of mitochondria placed within the cell body of motor neuron. In fact, it was demonstrated that, despite preservation of mitochondria placed within this compartment, there is no increase in the lifespan of transgenic mouse models of ALS. Thus, the present mini-review comments on morphological findings of mitochondrial alterations in ALS patients in connection with novel findings about mitochondrial dynamics within various compartments of motor neurons. The latter issue was recently investigated in relationship with altered calcium homeostasis and autophagy, which affect mitochondria in ALS. In fact, it was recently indicated that a pathological mitophagy, mitochondriogenesis and calcium homeostasis produce different ultrastructural effects within specific regions of motor neurons. This might explain why specific compartments of motor neurons possess different thresholds to mitochondrial damage. In particular, it appears that motor axons represent the most sensitive compartment which undergoes the earliest and most severe alterations in the course of ALS. It is now evident that altered calcium buffering is compartment-dependent, as well as mitophagy and mitochondriogenesis. On the other hand, mitochondrial homeostasis strongly relies on calcium handling, the removal of altered mitochondria through the autophagy flux (mitophagy) and the biogenesis of novel mitochondria (mitochondriogenesis). Thus, recent findings related to altered calcium storage and impaired autophagy flux in ALS may help to understand the occurrence of mitochondrial alterations as a hallmark in ALS patients. At the same time, the compartmentalization of such dysfunctions may be explained considering the compartments of calcium dynamics and autophagy flux within motor neurons.

Keywords: mitochondria, amyotrophic lateral sclerosis, autophagy, human patients, motor neuron, electron microscopy, biogenesis of mitochondria

### Introductory Statement

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive neurodegenerative disorder, which is characterized by massive motor neuron loss in the brainstem and spinal cord as well as motor cortex (Charcot, 1874; Boillée et al., 2006). The severity of this neurological disorder led to intense research efforts aimed to elucidate molecular and cellular events underlying motor neuron degeneration. In dissecting the variety of molecular mechanisms which characterize ALS several experimental approaches have been used. Multiple pathways might play a detrimental role on motor neuron survival. In fact, at mitochondrial level the occurrence of altered calcium homeostasis was described in great detail by recent studies (Fuchs et al., 2013; Barrett et al., 2014), while at cellular level the evidence of altered autophagy machinery seems to be well established (Pasquali et al., 2009). Nonetheless, a final common pathway connecting fine molecular mechanisms within mitochondria and pathological events at cellular level still needs to be clarified. Therefore, in the present short manuscript we discuss the significance of ultrastructural evidence, which was established in ALS patients for decades, in connection with altered mechanisms of calcium homeostasis and mitochondrial dynamics. Mitochondrial alterations were described in the ultrastructural pathology of ALS since early 80's by Atsumi (1981) when analyzing muscle biopsies from ALS patients. Despite their pioneer nature, these studies evidenced the earlier site of mitochondrial alterations at the level of muscle nerve endings. In fact, the routine description of motor neuron cell bodies carried out within ALS spinal cord, despite disclosing some hallmarks of ALS, rules out the potential role of ultrastructural pathology which occurs in motor nerve endings. In keeping with this, some authors emphasized mitochondrial alterations occurring within muscle nerve endings as key mechanisms of disease. Thus, Siklós et al. (1996) pointed out that, at early disease stages, ALS patients develop severe ultrastructural alterations within muscle presynaptic nerve terminals. This is known to consist of increased mitochondrial volume produced by dilution of the matrix and swelling of the organelles featuring broken cristae. These abnormalities represent a hallmark of ultrastructural pathology in ALS where giant mitochondria are often placed within big stagnant vesicular bodies, which were later identified as defective autophagy vacuoles. Remarkably, these findings in ALS patients are replicated by a number of ALS models (Sasaki and Iwata, 1996a,b, 2007; Fornai et al., 2008b; Ferrucci et al., 2010). Therefore, these models provided a useful tool to analyze the neurobiology of disease. For instance, it was established that giant mitochondria are associated with increased neuronal volume (Martin et al., 2007; Fornai et al., 2008a). Again, motor neuron cell body in ALS is filled with giant vesicles (Martin et al., 2007; Fornai et al., 2008a; Laird et al., 2008; see **Figure 1**). Not surprisingly, these giant vesicles may contain swollen and disrupted mitochondria (Fornai et al., 2008a). These vesicles often fill the whole cell body of motor neurons leading to the concept of slow necrosis (Martin et al., 2007). These vesicles stain for specific autophagy antigens indicating that autophagy pathway is often relented and/or impaired within ALS motor neurons (Fornai et al., 2008a; Laird et al., 2008). The autophagy machinery possesses a specific role in removing altered mitochondria (so-called mitophagy) which suggests that, apart from primary mitochondrial alterations, even a relented removal of aged/altered mitochondria co-exists to produce an overloading of dysfunctional mitochondria within motor neurons.

### The Characterization of Mitochondrial Alterations

Mitochondrial alterations are constantly found within motor neurons of the spinal cord, thus making it mandatory to decipher which molecular mechanism is implicated to comprehend ALS. Seminal studies by a number of research groups clearly demonstrated that mitochondrial alterations are produced by or associate with altered mitochondrial calcium homeostasis. For instance Ladewig et al. (2003) by using multiphoton microscopy and patch clamp recording demonstrated the occurrence of exaggerated calcium release and diminished calcium storage by mitochondria of motor neurons under specific stimuli. This suggests a specific vulnerability of motor neurons to develop disruption of mitochondrial calcium homeostasis upon sustained stimulation. This hypothesis was validated by Jaiswal and Keller (2009) by using a G93A mouse model of ALS. Fuchs et al. (2013) found that during the course of ALS impaired mitochondrial calcium buffering is modified. In detail, in order to compensate for a severe impairment of calcium buffering from spared mitochondria (and the loss of mitochondria) a plasma membrane calcium extrusion mechanism is upregulated at the end stage of the disease. This suggests an endogenous compensatory mechanism which might be viewed as a promising therapeutic approach to be enhanced by exogenous manipulation. Nonetheless, a recent manuscript by Parone et al. (2013) mitigated and even challenged such a concept.

## A Challenge to the Role of Mitochondria in ALS

Parone et al. (2013) demonstrated that protection of motor neuron mitochondria in the spinal cord induced by inhibiting cyclophilin D (a key regulator of calcium-mediated opening of the mitochondrial transition pore, mTP) in three varieties of superoxide dismutase 1 (SOD1) mutations, despite preserving the number of motor neurons counted in the spinal cord, did neither mitigate symptoms nor prolong survival in experimental ALS. This sharp experimental approach re-introduced the seminal role of peripheral motor denervation as a key determinant in producing palsy and lethality in ALS. These findings lend substance to very early electron microscopy studies in human patients showing that motor axon loss within muscles is advanced at early stages of disease (Atsumi, 1981). Should these data being considered as a challenge to the concept that mitochondria play a pivotal role of in ALS? This is debatable since the occurrence of peripheral degeneration of

motor axons is accompanied by severe mitochondrial pathology. Similarly, in their manuscript Parone et al. (2013) did not rule out the detrimental role of mitochondrial alterations. Then, one might consider that mitochondria in ALS motor neuron cell bodies play a sort of epiphenomenal role being not key in disease progression compared with mitochondrial alterations within motor nerve terminals. Similarly, protecting mitochondria within motor neuron cell body does not necessarily relates with protection of mitochondria within motor axons. Thus, being the axonal loss directly responsible for producing palsy and lethality, it is not surprising to observe fatal disease progression in the presence of spared motor neurons counted in the central nervous system. This point of view does not rule out the detrimental effects of mitochondrial alterations but it moves the consequence of mitochondrial damage to which motor neuron compartment is mostly affected. This confirms pioneer studies of Hart et al. (1977) and Hirano et al. (1984a,b) who found the occurrence of altered mitochondria following electron microscopy of motor neurons in patients affected by ALS. Although, it is critical to consider that ultrastructural findings in ALS patients indicate that swollen mitochondria in peripheral nerves occur early than within spinal cord motor neurons (Sasaki and Iwata, 1996a,b, 2007; Siklós et al., 1996).

### How to Reconcile the Altered Mitochondrial Calcium Homeostasis with Previous Point

A very recent manuscript by Barrett et al. (2014) discussed the apparent discrepancy between data obtained with cyclophilin D KO mice and the key role of altered mitochondrial calcium buffering observed in SOD1 mutant mice. These authors provided a series of strong points to reconcile the critical loss of calcium buffering with the lack of protection from symptoms and lethality published by Parone et al. (2013) in cyclophilin D KO mice. For instance the suppression of pathological calcium current observed in cyclophilin D KO mice might not be as effective in axonal mitochondria as that one measured in the motor neuron cell body. This hypothesis includes the chance that axonal mitochondria may possess a kind of calcium alterations which are not preventable by inhibiting expression of cyclophilin D. This includes both a higher variety of mitochondrial stressors within axons compared with cell body and the higher surface-to-volume ratio (loss of the spheroid shape) for axonal mitochondria which would render these organelles richer in density for calcium channels. Similarly, one might add to Barrett et al.'s (2014) considerations that, such a mitochondrial shape would make these organelles more exposed to a toxic microenvironment. When discussing in depth the lack of protection of cyclophilin D KO mice, Barrett et al. (2014) add a number of hypothesis about different mechanisms of neurotoxicity between axons and cell bodies of motor neurons which are plausible indeed. Apart from focusing on differential vulnerability of axon compared with cell body mitochondria, it is worth to be mentioned that a different dynamics may occur for mitochondria placed within motor axons compared with cell bodies. As reported again by Barrett et al. (2014), this difference was first described by Magrané et al. (2012). These authors, by using live imaging microscopy of photo-switchable fluorescent mitochondrial dye, demonstrated that mitochondria from G93A mice possess a slower axonal transport and decreased fusion.

### The Key Role of Mitochondrial Compartments

Altogether, these concepts lead to emphasize the role of motor neuron compartments when considering that mitochondrial alterations do represent a key event in ALS pathogenesis. Therefore, apart from the specific mechanisms it is very likely that the threshold for damage at axonal mitochondria is likely to be lower when compared with the threshold which is needed to damage mitochondria placed in the cell body of motor neurons. This would reconcile the occurrence of axonal denervation in the presence of sparing motor neuron cell bodies described by Parone et al. (2013). Thus, if one analyze the role of mitochondrial dynamics beyond the findings of Magrané et al. (2012, 2014), it is worth to be mentioned that axonal transport it is regulated by the very same class of proteins which regulate autophagy (Pasquali et al., 2014). In fact, altered mitochondrial dynamics should be viewed in a wider perspective where impaired removal of altered mitochondria (impaired mitophagy, which is a part of the autophagy machinery) plays a key role. This impairment indeed occurs in G93A mice (Fornai et al., 2008b; Pasquali et al., 2009) but it seems to extend to other ALS model and ALS related genes (Laird et al., 2008). Similarly, the impairment in mitochondrial dynamics ranges from G93A to TAR DNA binding protein 43 (TDP-43) mutant mice (Magrané et al., 2012, 2014). At the same time, apart from the formal description of a defective mitochondrial fusion (Magrané et al., 2012, 2014), one might extend the analysis to the authentic biogenesis of mitochondria which is defective again in ALS models as shown by polymerase chain reaction (PCR) of mitochondrial genes and MitoTracker green and red (Fornai et al., 2008a). As we shall see in the next paragraph, there is now abundant and very recent evidence, that autophagy of mitochondria is co-activated with mitochondria biogenesis and a defect in autophagy eventually involves a deficiency in mitochondriogenesis, whereas a stimulation of mitophagy concomitantly promotes the biogenesis of novel mitochondria. The mitochondrial compartment then plays a pivotal role in this scenario, where remote axon terminals are expected to be much more affected than neuronal cell bodies.

### Where Damaged Mitochondria Come From?

When mitochondrial alterations play a pivotal role, than it should be considered whether these may occur directly as the effect of a primary toxicity to mitochondria affecting calcium homeostasis or they can be produced by a defect of mitochondrial removal or even by a relented biogenesis of novel mitochondria. Even in these latter cases abnormal mitochondria are expected to possess altered calcium storage as shown by von Lewinski and Keller (2005). In this scenario several ALS phenotypes are likely to be included (see **Figure 2**). In fact, in the case of a mutation of the SOD1 gene, an overactive enzyme impairing mitochondrial function is produced (Higgins et al., 2002; Vehviläinen et al., 2014). In addition, in the very same strain of mice an impaired removal of mitochondria due to impaired mitophagy is documented (Pasquali et al., 2009, 2014). This may take a prominent role when specific ALS related proteins are mutated. For instance, the dynactin mutation (Münch et al., 2004) produces a defect in the autophagy flux which in turn is accompanied by stagnant autophagy vacuoles (Laird et al., 2008; Ikenaka et al., 2013). Interestingly, when alterations in the autophagy (mitophagy) machinery are described these are concomitant with defects in the biogenesis of novel mitochondria. In fact, autophagy inducers are described to increase mitochondriogenesis (Struewing et al., 2007; Fornai et al., 2008a), while a common pathway simulates both mitophagy and mitochondriogenesis (Palikaras et al., 2015a,b). Recent data show that mitophagy is tightly related to the biogenesis of novel mitochondria (Palikaras et al., 2015a,b). In detail, when a certain amount of damaged mitochondria is produced, this triggers mitophagy which mediates the removal of damaged mitochondria. This is based on SKN-1 activation, which beside promoting mitophagy, also increases mitochondrial biogenesis (Palikaras et al., 2015a,b). Thus, it is expected that a failure in the autophagy pathway comes together with a defect in the biogenesis of mitochondria. For instance, Palikaras et al. (2015b) hypothesized that suppression of mitophagy inhibits both mitochondria removal and mitochondria biogenesis, thus producing a bidirectional mechanism to increase mitochondrial alterations. Similarly, it is not surprising that autophagy inducers such as lithium or resveratrol, which are autophagy inducers, concomitantly stimulate the biogenesis of novel mitochondria (Fornai et al., 2008a; Meira-Martins et al., 2015; **Figure 1**). Thus, a sort of tightened dual feedback may adjust mitochondrial population. Not surprisingly, both lithium and resveratrol were found to improve experimental ALS and other motor

(Continued)

#### FIGURE 2 | Continued

mitochondria may potentially lead to accumulation of degenerated mitochondria, to our knowledge a specific familial ALS (fALS) phenotype due to such a defect was not described so far. Nonetheless, it is likely that, due to a dual tightened control of mitochondrial removal and biogenesis of mitochondria, a failure in the first pathway will eventually lead to a failure in the biogenesis of novel mitochondria. Thus, it is not surprising that, in all fALS phenotypes featuring a defect in the progression of autophagy, we can detect only giant, altered mitochondria in the absence of small, newly synthesized mitochondria. This confirms the eventual concomitance of mitophagy and mitochondriogenesis as indicated by Palikaras et al. (2015a,b). Degenerated mitochondria, to our knowledge a specific fALS phenotype due to such a defect was not described so far. Nonetheless, it is likely that, due to a dual tightened control of mitochondrial removal and biogenesis of mitochondria, a failure in the first pathway will eventually lead to a failure in the biogenesis of novel mitochondria. Thus, it is not surprising that, in all fALS phenotypes featuring a defect in the progression of autophagy, we can detect only giant, altered mitochondria in the absence of small, newly synthesized mitochondria. This confirms the eventual concomitance of mitophagy and mitochondriogenesis as indicated by Palikaras et al. (2015a,b).

neuron disorders (Shimada et al., 2012; Mancuso et al., 2014) and synergistic effects in ALS patients are produced by combined administration of autophagy inducers such as valproate and lithium (Boll et al., 2014). However, it is true that the sole increase in the biogenesis of mitochondria does not guarantee for neuroprotection in ALS as shown by Da Cruz et al. (2012). At the same time when autophagy is not induced (Pizzasegola et al., 2009) due to a ten-fold subtherapeutic treatment (Chiu et al., 2013), the neuroprotective effects induced by lithium on motor neurons cannot be appreciated.

### The Close Connection Between Autophagy and Mitochondria

When focusing on mitochondrial alterations in human ALS, it becomes mandatory to analyze the autophagy status since the occurrence of mitochondrial alterations is likely to be accompanied by a derangement in the autophagy machinery. Confirming this novel standpoint there is evidence in ALS patients that adds on structural mitochondrial alterations showing that a variety of autophagy markers are altered in the spinal cord of ALS patients (Sasaki, 2011). These data often led to opposite interpretation either being considered as a proof for a detrimental role of autophagy in ALS or vice versa they have been considered an evidence that a failure of the autophagy machinery occurs in ALS. In keeping with mitochondrial dynamics, it is worth to be mentioned that occurrence of big autophagy vacuoles containing mitochondria generally reflect a defect in the autophagy flux rather than a pathological over-activation of the autophagy machinery. In keeping with this, most familial ALS (fALS) are related with a defect of proteins involved in the autophagy machinery, thereby inducing a failure in the autophagy pathway. A synthetic report of these mutations is reported below along with evidence of a defect in the autophagy machinery. This summarizes and up-dates what already reported by Pasquali et al. (2014).

## A Few Examples of Specific Effects of Human ALS Genes on the Autophagy Machinery

Briefly, more than twenty years ago the SOD1 was the first gene which was associated with fALS (Deng et al., 1993; Rosen et al., 1993). Remarkably, the mutant forms of the SOD1 protein, as well as the wild-type SOD1, are degraded by the autophagy pathway, which in turn, plays a pivotal role in decreasing SOD1 toxicity (Kabuta et al., 2006). In motor neurons from fALS (SOD1) patients and transgenic SOD1 mice as well, autophagy appears to be engulfed by an excess of SOD1. In these cells, a compensatory increase in autophagy markers such as levels of LC3-II occurs (Morimoto et al., 2007; Fornai et al., 2008a), nonetheless, autophagy progression is impaired. This explains why in the presence of SOD1 G93A mutation impairment of autophagy is concomitant with an increase in autophagyrelated proteins. The gene ALS2 is responsible for an autosomal recessive fALS (Yang et al., 2001). This gene codes for the alsin protein, which sustains autophagy progression by merging endosomes with autophagosomes to produce amphisomes. In fact, alsin deficiency decreases the motility of endosomes, which accumulate as Rab5 positive giant organelles (Lai et al., 2009). Missense mutations in charged multivesicular protein 2B (CHMP2B) were recently identified in fALS patients (Parkinson et al., 2006). CHMP2B is a component of endosomal sorting complexes required for transport III (ESCRT-III), which belongs to the ESCRT proteins involved in sorting of endocytosed ubiquitinated integral membrane proteins into multivesicular bodies (MVB; Babst et al., 1998, 2002; Katzmann et al., 2001). In particular, CHMP2B enables merging of autophagosomes with either endosomes or lysosomes (Rusten and Stenmark, 2009; Manil-Ségalen et al., 2014). Thus, mutations of CHMP2B lead to impairment in autophagy progression with accumulation of LC3-II positive autophagosomes and altered cargos degradation (Filimonenko et al., 2007; Lee et al., 2007; Cox et al., 2010). The TDP-43 is mostly placed in the nucleus of healthy cells and it is involved in gene transcription and alternative splicing. Patients with TDP-43 mutations develop fALS (Kühnlein et al., 2008; Sreedharan et al., 2008; Van Deerlin et al., 2008; Yokoseki et al., 2008) and possess a misplacement of TDP-43 (from nucleus to cytoplasm) in the form of neuronal inclusions (Arai et al., 2006; Neumann et al., 2006). TDP-43 metabolism is impaired by autophagy inhibitors which produce misplacement of TDP-43, while this is reversed under the effects of autophagy activation (Wang et al., 2010). In line with this, valproate attenuates neuronal toxicity by enhancing autophagy (Wang et al., 2015), while high levels of fragments from TDP-43 engulf the autophagy machinery causing motor deficits (Caccamo et al., 2015). Some fALS patients feature mutations of sequestosome 1 (SQSTM1; Fecto et al., 2011; Rubino et al., 2012). The SQSTM1 gene codes for the protein p62, which is a major autophagy inducer. The specific role of p62 in autophagy consists in linking ubiquitinated protein aggregates to the autophagy machinery (Gal et al., 2007). Heterozygous missense mutations of the dynactin 1 (DCTN1) gene were detected in other fALS patients (Münch et al., 2004). Dynactin mutations produce an autophagy failure (Laird et al., 2008). In fact, dynactin is part of a cytoskeletal molecular complex (consisting of dyenin, dynactin and dynamitin), which is key in promoting the cytoplasmic transport of vesicles along the axon and cell body (Gill et al., 1991; Schroer and Sheetz, 1991; Waterman-Storer et al., 1997). This extends to trafficking of autophagy vesicles such as the merging of autophagosomes with lysosomes (Gill et al., 1991; Schroer and Sheetz, 1991; Waterman-Storer et al., 1997; Laird et al., 2008). In fact, autophagosome needs to be transported along microtubules to the center of the cells (centrosome), where most of the lysosomes are located (Gill et al., 1991; Schroer and Sheetz, 1991; Waterman-Storer et al., 1997). Remarkably, this fALS-producing mutation is a paradigm to connect impairment of autophagy with a compartment-dependent alteration in the flux of organelles (including mitochondria).

Similarly, mutations of optineurin, a protein involved in intracellular trafficking (Ying and Yue, 2012), were described in fALS patients (Maruyama et al., 2010). Optineurin works as an autophagy receptor containing LC3 and ubiquitinbinding domain and it plays a pivotal role in parkin-mediated mitophagy (Wild et al., 2011). Remarkably, optineurin recruits LC3 and clusters around damaged mitochondria upstream to their entrapment within autophagosomes following their parkindependent ubiquitination. Thus, it is expected that mutation of optineurin leads to accumulation of damaged mitochondria. Ubiquilin 2, a member of the ubiquilin family, which delivers substrate to autophagy, was found to produce fALS (Deng et al., 2011; Williams et al., 2012). In particular, the loss of ubiquilin inhibits conversion of LC3-I to active lapidated LC3-II, which activates autophagy (Ko et al., 2004; Rothenberg et al., 2010). Mutations of the valosin-containing protein (VCP) gene were described in fALS (Johnson et al., 2010). This gene codes for a chaperone protein involved in mitophagy through

### References


autophagosome maturation (Tanaka et al., 2010; Meyer et al., 2012; Yamanaka et al., 2012). Hexanucleotide (GGGGCC) repeat expansions in a non-coding region of chromosome 9 open reading frame 72 (C9orf72) occur in fALS (DeJesus-Hernandez et al., 2011). Very recently C9orf72 was described to be involved in the trafficking of autophagy vesicles (Farg et al., 2014).

### Conclusion

The bulk of mutations reported in the last paragraph, characterize most fALS and indicate a mechanistic connection between autophagy impairment and ALS. This evidence is based on multidisciplinary approaches encompassing in vitro protein assay and in vivo genetic manipulation. Since the autophagy machinery is key for removing altered mitochondria, it is not surprising that despite a plethora of different mutated proteins in various fALS patients, ultrastructural evidence consistently report the occurrence of a number of altered mitochondria. At the same time, the chronic reiteration of a primary injury towards mitochondria is expected to overwhelm the compensatory mitochondria turn-over. Thus, recent evidence showing impairment of the autophagy machinery in ALS is complementary with the seminal findings showing altered mitochondrial calcium homeostasis in ALS motor neurons. In fact this may occur either as a primary defect or as the consequence of altered mitochondrial turn over. Very recently, such a scenario was remarkably enriched by the evidence that impaired mitophagy necessarily triggers a failure in the biogenesis of novel mitochondria. Thus, the occurrence in ALS of a variety of defects such as: (i) fine mitochondrial dysfunctions, mostly related to calcium homeostasis; (ii) impairment of mitophagy flux; and (iii) failure of mitochondrial biogenesis (often reported as a mere fusion defect) appear more and more as different perspectives to describe similar phenomena.


degeneration and amyotrophic lateral sclerosis. Science 314, 130–133. doi: 10. 1126/science.1134108


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Ruffoli, Bartalucci, Frati and Fornai. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Comparison of dendritic calcium transients in juvenile wild type and SOD1G93A mouse lumbar motoneurons

Katharina A. Quinlan<sup>1</sup> , Jonathan B. Lamano<sup>1</sup> , Julienne Samuels <sup>1</sup> and C. J. Heckman1, 2, 3 \*

*<sup>1</sup> Department of Physiology, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA, <sup>2</sup> Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA, <sup>3</sup> Department of Physical Therapy and Human Movement Sciences, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA*

#### Edited by:

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### Reviewed by:

*Gordon William Arbuthnott, Okinawa Institute of Science and Technology, Japan Richard Anthony DeFazio, University of Michigan, USA Stefan G. Stanciu, University Politehnica of Bucharest, Romania*

#### \*Correspondence:

*C. J. Heckman, Department of Physiology, Feinberg School of Medicine, Northwestern University, 303 East Chicago Avenue, Chicago, IL 60611, USA c-heckman@northwestern.edu*

> Received: *14 January 2015* Accepted: *23 March 2015* Published: *10 April 2015*

#### Citation:

*Quinlan KA, Lamano JB, Samuels J and Heckman CJ (2015) Comparison of dendritic calcium transients in juvenile wild type and SOD1*G93A *mouse lumbar motoneurons. Front. Cell. Neurosci. 9:139. doi: 10.3389/fncel.2015.00139* Previous studies of spinal motoneurons in the SOD1 mouse model of amyotrophic lateral sclerosis have shown alterations long before disease onset, including increased dendritic branching, increased persistent Na<sup>+</sup> and Ca2<sup>+</sup> currents, and impaired axonal transport. In this study dendritic Ca2<sup>+</sup> entry was investigated using two photon excitation fluorescence microscopy and whole-cell patch-clamp of juvenile (P4-11) motoneurons. Neurons were filled with both Ca2<sup>+</sup> Green-1 and Texas Red dextrans, and line scans performed throughout. Steps were taken to account for different sources of variability, including (1) dye filling and laser penetration, (2) dendritic anatomy, and (3) the time elapsed from the start of recording. First, Ca2<sup>+</sup> Green-1 fluorescence was normalized by Texas Red; next, neurons were reconstructed so anatomy could be evaluated; finally, time was recorded. Customized software detected the largest Ca2<sup>+</sup> transients (area under the curve) from each line scan and matched it with parameters above. Overall, larger dendritic diameter and shorter path distance from the soma were significant predictors of larger transients, while time was not significant up to 2 h (data thereafter was dropped). However, Ca2<sup>+</sup> transients showed additional variability. Controlling for previous factors, significant variation was found between Ca2<sup>+</sup> signals from different processes of the same neuron in 3/7 neurons. This could reflect differential expression of Ca2<sup>+</sup> channels, local neuromodulation or other variations. Finally, Ca2<sup>+</sup> transients in SOD1G93A motoneurons were significantly smaller than in non-transgenic motoneurons. In conclusion, motoneuron processes show highly variable Ca2<sup>+</sup> transients, but these transients are smaller overall in SOD1G93A motoneurons.

Keywords: SOD1G93A mice, motor neuron, spinal cord, calcium channels, multiphoton imaging

## Introduction

The hallmark of amyotrophic lateral sclerosis (ALS) is the degeneration of both upper (corticospinal) and lower (spinal) motoneurons. Many disruptions have been found in both patients and animal models that could negatively affect motoneuron health, including altered protein degradation (Deng et al., 2011; Bendotti et al., 2012), neuroinflammation (Evans et al., 2013), and misregulated RNA processing (Polymenidou et al., 2012). Mutant superoxide dismutase-1 (SOD1) mouse models have yielded an abundance of information on properties of vulnerable motoneuron populations. Electrical and anatomical properties are altered from a very early age (Kuo et al., 2004, 2005; Bories et al., 2007; Amendola and Durand, 2008; van Zundert et al., 2008; Pambo-Pambo et al., 2009; Quinlan et al., 2011; Filipchuk and Durand, 2012; Martin et al., 2013; Leroy et al., 2014). Embryonically, mutant SOD1 spinal motoneurons have shorter dendritic branches and are hyperexcitable (Kuo et al., 2004, 2005; Martin et al., 2013), while postnatally spinal motoneurons show increased dendritic branching and their overall excitability is normalized, despite larger persistent inward Na<sup>+</sup> and Ca2<sup>+</sup> currents (NaPIC and CaPIC) (Amendola and Durand, 2008; Quinlan, 2011; Filipchuk and Durand, 2012). Whether the proliferated dendritic branches in mutant SOD1 motoneurons have similar levels of activity as dendrites of wild type motoneurons has not yet been examined.

The Ca2<sup>+</sup> currents generating the CaPIC are most likely Cav1.3 (and potentially Cav1.2) channels. Though Cav1.3 channels failed to show higher expression levels in adult SOD1 motoneurons using immunohistochemistry (Shoenfeld et al., 2014), the larger amplitude current could reflect altered channel activation or kinetics rather than altered expression levels. Any potential changes in high-voltage activated Ca2<sup>+</sup> currents (Cav2 channels) in SOD1 motoneurons remain relatively unstudied. Voltage-gated Ca2<sup>+</sup> currents measured from wild type spinal motoneurons include P/Q (Cav2.1), N (Cav2.2), R (Cav2.3), L (Cav1.2 and 1.3), and T-type (Cav3) currents (Mynlieff and Beam, 1992; Carlin et al., 2000a). In brainstem hypoglossal motoneurons, previous studies have revealed complex microdomains of wildly differing internal Ca2+. Calcium that enters through voltage-gated Ca2<sup>+</sup> channels is left unbound due to the lack of endogenous buffers, and is eventually taken up and released by the mitochondria and endoplasmic reticulum (Lips and Keller, 1999; Ladewig et al., 2003). The end result is variable internal Ca2<sup>+</sup> arising from both internal stores and external Ca2<sup>+</sup> entry in the soma and proximal dendrites (up to 80µm) of hypoglossal motoneurons (Ladewig et al., 2003). In contrast to hypoglossal motoneurons, spinal motoneurons possess even more complex dendritic arborizations, but analysis of these microdomains remains unstudied.

Therefore, in this study there were two main goals: (1) To examine normal anatomical patterns of high-voltage activated Ca2<sup>+</sup> transients elicited throughout the dendritic field of wild type motoneurons, (2) To compare the patterns of these transients in wild type motoneurons to those of SOD1G93A motoneurons.

#### Materials and Methods

#### Ethics Statement

Experiments were performed in accordance with the United States National Institutes of Health Guide for Care and Use of Laboratory Animals. Approval of the Northwestern University's Animal Care and Use Committee was obtained for all experiments performed in this study. All efforts were made to minimize animal suffering and to reduce the number of animals used.

#### Animals and Surgery

For this study, 60 juvenile mouse pups were used between postnatal day (P) 4-11. Transgenic B6SJL mice overexpressing the human SOD1G93A gene (strain 002726, Jackson Labs, Bar Harbor, ME, USA) were identified using standard PCR techniques (Rosen et al., 1993). Briefly, 20–25 mg of tissue was used for the DNA extraction with primers for amplification CAG TAA CTG AGA GTT TAC CCT TTG GT (forward) and CAC ACT AAT GCT CTG GGA AGA AAG A (reverse). Both transgenic SOD1G93A and non-transgenic littermates were deeply anesthetized with isofluorane (Henry Schein Animal Health, Dublin, OH, USA), decapitated and eviscerated. The lumbar spinal cord was removed and embedded in 2.5% w/v agar (No. A-7002, Sigma-Aldrich, St. Louis, MO, USA). The agar block was then superglued with Loctite 401 (Henkel Corporation, Rocky Hill, CN, USA) to a stainless steel slicing bath and 350µm transverse slices are made using the Leica 1000 vibratome (Leica Microsystems, Buffalo Grove, IL, USA) as described previously (Quinlan et al., 2011). During both spinal cord isolation and slicing, the spinal cord was immersed in 1–4◦C high osmolarity dissecting solution containing (mM) sucrose 234.0, KCl 2.5, CaCl<sup>2</sup> · 2H2O 0.1, MgSO<sup>4</sup> · 7H2O 4.0, HEPES 15.0, glucose 11.0, Na2PO<sup>4</sup> 1.0. The pH was adjusted to 7.35 when bubbled with 95% O2/5% CO<sup>2</sup> using 1M KOH (Fluka Analytical, Sigma-Aldrich). After cutting, the slices were incubated for >1 h at 30◦C in incubating solution containing (mM) NaCl 126.0, KCl 2.5, CaCl<sup>2</sup> · 2H2O 2.0, MgCl<sup>2</sup> · 6H2O 2.0, NaHCO<sup>3</sup> 26.0, glucose 10.0, pH 7.4 when bubbled with 95% O2/5% CO<sup>2</sup> (all reagents for solutions were purchased from Sigma-Aldrich).

#### Electrophysiology

Whole cell patch clamp was performed on motoneurons from the lumbar segments using 2–4 M glass electrodes pulled from glass capillary tubes (Item # TW150F-4, World Precision Instruments, Sarasota, FL, USA) with a Flaming-Brown P-97 (Sutter Instrument Company, Novato, CA, USA). Electrodes were positioned using a Sutter Instrument MP-285 motorized micromanipulator (Sutter Instrument Company). Whole-cell patch clamp measurements were performed at room temperature using the Multiclamp 700B amplifier (Molecular Devices, Burlingame, CA, USA) and Winfluor software (University of Strathclyde, Glasgow, Scotland). Briefly, slices were perfused with a modified Ringer's solution containing (in mM): 111 NaCl, 3.09 KCl, 25.0 NaHCO3, 1.10 KH2PO4, 1.26 MgSO4, 2.52 CaCl2, and 11.1 glucose. The solution was oxygenated with 95% O<sup>2</sup> and 5% CO<sup>2</sup> and the perfusion rate was 2.5–3.0 ml/min. Patch electrodes contained (in mM) 138 K-gluconate, 10 HEPES, 5 ATP-Mg, 0.3 GTP-Li and Texas Red dextran and Ca2<sup>+</sup> Green-1 dextran (both 150µM or both at 200µM, 3000 MW, from Invitrogen, Life Technologies, Grand Island, NY, USA). The K<sup>d</sup> of Ca green-1 varied between lots, and ranged from 406 to 737 nM.

During line scans, neurons were subjected to three depolarizing steps each at 50, 100, and 200 ms to evoke 1 action potential, 3–5 action potentials, and ∼10 action potentials per step, respectively. Stimulation protocols were developed to assess activity of high voltage gated Ca2<sup>+</sup> channels in the dendrites. Depolarizing the soma to threshold for a single spike elicited a fairly robust Ca2<sup>+</sup> signal in the soma and proximal dendrites. However, it was not clear whether single spikes would be strong enough to spread electrotonically back as far as 300 um or more when more distal processes were examined. Conversely, limiting the duration of the stimuli was necessary to minimize the exposure time (and thus any damage) to the processes by the laser. With 50, 100, and 200 ms pulses, we avoided damaging dendrites with prolonged laser exposure yet produced a depolarization robust enough to measure even in distal dendrites. As shown in **Figure 1**, Ca2<sup>+</sup> signals from the three events were averaged unless otherwise noted.

#### Two-Photon Excitation Fluorescence Microscopy

An Olympus BX-51WIF microscope fitted with an Olympus 40x/0.8NA water-dipping objective lens was used for visualizing the patched neuron. Two photon excitation fluorescence microscopy (2PEF) was performed with a galvanometer-based Coherent Chameleon Ultra II laser (Coherent, Santa Clara, CA, USA). To choose an appropriate wavelength for excitation of both fluorophores, two photon excitation spectra were run for both Texas Red and Ca2<sup>+</sup> green-1. Based on these spectra, the laser was tuned to 820 nm. The average laser power at this wavelength was about 2 W, though it is expected that some power was lost as the beam passed through the Bio-Rad Radiance modulators and the objective. Since emission of these fluorophores is well-separated, two Bio-Rad 2100 MPD photomultiplier tubes (Bio-Rad, Hercules, CA, USA): Green PMT (500–550 nm) and Red PMT (570–650 nm) were used to collect emission in the red and green channels. Unidirectional line scans were performed along the length of the dendrites with a dwell time of 10µs/pixel. Precautions were taken to reduce noise in the images, including very low ambient light levels during experiments and a heavy curtain surrounding the rig. Fortunately, in the spinal cord of young mice at the ages used in these experiments, the auto-fluorescence of tissue is very low compared to the adult cord, probably due to incomplete myelination.

#### Neuron Selection

The ventrolateral motoneuron pools could be easily visualized in the slice preparation (as seen in **Figure 1**), and electrodes were positioned above this area. Putative motoneurons were targeted based on large soma diameter (>20µM long axis). Only neurons with a resting potential <−50 mV, an action potential crossing 0 mV and a series resistance of <15 M were used. Neurons were eliminated from analysis if series resistance or resting potential varied more than 10 M or 10 mV, respectively, throughout the recording period. Cell properties are listed in **Table 1**.

#### Dendritic Morphology

Neuronal tracing and three-dimensional reconstructions from z stacks were accomplished manually through Neurolucida reconstruction software (MBF Bioscience, Williston VT) and quantified with Neurolucida Explorer. Neuronal reconstructions, z stacks, and line scan reference images were utilized to determine the path distance from the soma to the line scanned processes, along with the mean diameter of the sampled segments.

### Action Potential Evoked Ca+<sup>2</sup> Entry

All signals throughout this manuscript are measured from the area under the curve of the averaged Ca2<sup>+</sup> transient. In brief, custom software for MATLAB (Mathworks, Natick, MA, USA) was developed by our laboratory and used for precision measurement of Ca2<sup>+</sup> signal along each line scan. For each line scan, a single position along the dendrite was retained, corresponding to the pixel that generated the largest Ca2<sup>+</sup> signal for that line scan. The exact 0.2–0.3µm region of the dendrite (after applying a moving average of range 2.6µm) which gave rise to that peak signal was then assessed for (1) diameter and (2) path distance to the soma. The code is included as a Supplemental File.

#### Regions of Interest–Dendrite vs. Soma

Distinct regions of interest (ROI) along each line scan encompassing dendritic processes and somas were identified utilizing reference images and z stacks. Furthermore, an ROI devoid of processes was determined for each line scan to serve as a baseline intensity for normalizing observed Ca2<sup>+</sup> Green and Texas Red intensities.

### Quantification of Ca+<sup>2</sup> Transient

Measurement of action potential evoked Ca2<sup>+</sup> entry was achieved through analysis of the change in Ca2<sup>+</sup> Green-1 fluorescent intensity through MATLAB. Ca2<sup>+</sup> transients were analyzed following the onset of stimulus application until the end of imaging (approximately 1 s). For each scanned process, three evoked Ca2<sup>+</sup> transients were quantified from each of three protocols. For all line scans, each pixel along the length of the scan was quantified temporally in terms of Ca2<sup>+</sup> Green and Texas Red intensity normalized by the mean green and red fluorescent intensity within the baseline ROI, respectively. This method was pioneered by Bloodgood and Sabatini (2007). For each stimulus, the mean pre-stimulus baseline intensity of the Ca2<sup>+</sup> Green fluorescence was subtracted from the Ca2<sup>+</sup> Green intensity during the window from stimulus onset to imaging termination to determine the change in Ca2<sup>+</sup> Green fluorescence (1G), an indicator of the change in internal Ca2+. 1G was normalized by the pre-stimulus Texas Red fluorescence (R0) to obtain a 1G/R<sup>0</sup> curve. The area under the 1G/R<sup>0</sup> curve for each stimulus was then computed and the three values for each pixel of the line scan averaged together. As a result, a value for the area under 1G/R<sup>0</sup> was obtained for each pixel along the length of the line scan. For each line scan, the maximum value of area under 1G/R<sup>0</sup> within the dendritic or somal ROI was identified as a quantification of the Ca2<sup>+</sup> transient, as shown in **Figure 1D**. From the pixel sampled as the maximum value of area under 1G/R0, it was verified that the pixel was located within the process, then the path distance from that pixel to the soma was approximated in MATLAB based on neuronal reconstructions and line scan reference images.

#### Within-Cell Branch Analyses

For cells with scanned processes from at least three separate dendritic branches, a within-cell analyses of Ca2<sup>+</sup> transients was

FIGURE 1 | Summary of experimental methods. (A) Left panel: Photograph of a transverse spinal cord slice pictured with the patch electrode descending from the center top to the bottom right of the photo, where z-stacks of the filled cell are superimposed onto the cord. Horizontal scale bar = 0.5 mm. Right panel: Schematic showing placement of the putative motoneurons in black triangles (wild type, *n* = 13) and red triangles (SOD1, *n* = 8). (B) Left panel: Image of a patched neuron's soma and primary dendrite. The vertical dotted line indicates placement of the line scan. Scale bar = 20 um. Right panels: Green (top panel) and red (bottom panel) fluorescence is measured repeatedly over time (x axis) from the location of the line scan (dotted

performed using an arbitrarily assigned branch number as a variable in the analysis of covariance (ANCOVA) described below.

#### Statistical Analyses

All statistical analyses were performed using SPSS statistical software (IBM, Armonk, NY). Unless otherwise stated, line from left panel of B). (C) Laser is turned on only during periods of simulation, as shown in yellow trace. Numeric values for green and red fluorescent intensity are below. The cell is stimulated with depolarizing pulses (200 ms/500 pA in this example) to fire action potentials (bottom two traces), during which the green channel shows a marked increase in fluorescence corresponding to the action potentials, while the red channel shows no change. Vertical scale bar = 50 mV. (D) The averaged signal for the three events, measured as 1 Green fluorescent intensity/Red fluorescent intensity over time. All values of Ca2<sup>+</sup> transients throughout this study are calculated from the area under the curve of Green/Red fluorescent intensity, as shown in (D).

Ca2<sup>+</sup> transients were analyzed using ANCOVAs [modified ANOVAs tailored to use both categorical and quantitative variables (Fisher, 1971)]. This analysis was particularly suited to our dataset, since while testing the effect of one variable (i.e., path distance) on Ca2<sup>+</sup> transients, the effects of covariates (i.e., stimulus duration) could be accounted for. Branch


TABLE 1 | Cell properties are listed for WT (n = 13) and SOD1 (n = 8) motoneurons.

*Each parameter is listed with mean, standard deviation (St Dev), and range. Input resistance (RIN), capacitance (Cap), resting membrane potential (RMP), and series resistance (Rseries) are listed for all recorded cells. None of these parameters were significantly different between WT and SOD1 groups (p-values based on unpaired T-tests).*

diameter and path distance were used individually in the analyses since there was significant interaction between the two variables (dendritic branches with the largest diameter were always closer to the soma). Unpaired T-tests were used to test significance of electrophysiological characteristics between WT and SOD groups. T-tests were used with post-hoc Bonferroni corrected multiple comparisons tests when testing baseline fluorescence.

### Results

#### Action Potentials Initiated at Soma Elicited Robust Dendritic Ca2<sup>+</sup> Entry

In transverse spinal cord slices, eight SOD1G93A and 13 wild type visually-identified motoneurons were recorded from the lumbar region as shown in **Figure 1A**. Large neurons were targeted in the motoneuron pools, and dyes from the electrode filled the patched neurons (**Figure 1A**, first and second panels). Electrophysiological characteristics of the motoneurons are included in **Table 1**. Texas red dextran (3000 MW) was used for its strong fluorescent signal, and Ca2<sup>+</sup> transients were measured with the calcium indicator Ca2<sup>+</sup> green-1 dextran (also 3000 MW). Stimulation protocols were run simultaneously with line scans as shown in **Figures 1B,C**. As the top (green) and bottom (red) panels in **Figure 1B** show, Texas red had a much stronger signal than Ca2<sup>+</sup> green-1 at rest. But when action potentials were elicited, Ca2<sup>+</sup> green showed a robust increase in fluorescence. Customized MATLAB software was used to measure the signal from 2.6µm moving averages throughout the line scan and detect the pixel where the peak signal was centered. The anatomical characteristics of the dendrite at the location of the peak signal were used in later analysis. Three different protocols were applied to each neuron, each with three events: three depolarizing pulses of 50 ms (evoking one spike/stimulus), three pulses of 100 ms (evoking 3–5 spikes per stimulus), and three pulses of 200 ms (evoking about 10 spikes per stimulus). The three pulses of each protocol were averaged together (except where otherwise noted), and the area under the 1Green/Red curve (as shown in **Figure 1D**) was used to quantify the transient. This protocol was used in order to minimize the differences in laser penetration through different depths of the slice (by normalizing the fluorescence of the Ca2<sup>+</sup> signal to the fluorescence of the inert Texas red), and to minimize changes in the peak and decay of the Ca2<sup>+</sup> signal over time due to any potential increases in concentration of the Ca2<sup>+</sup> sensitive dye over time (Helmchen et al., 1996).

#### Baseline Fluorescence

After whole cell patch configuration was established, at least 20 min were allowed to pass for diffusion of the dyes. Since patch electrodes were large (2–4 M resistance), diffusion took place rapidly and Texas red was clearly visible throughout the processes after this time. To further examine whether time may have had an effect on strength of the Ca2<sup>+</sup> signal, ANCOVA was performed on the Ca2<sup>+</sup> signal over time. It was found that after 120 min post-break in, there was a significant drop off in the Ca2<sup>+</sup> signal, as can be observed in **Figure 2A**. This may indicate eventual run-down. Therefore, all data collected after 2 h was eliminated from further analysis. Data collected up to 120 min showed no significant effects of time, as shown in **Table 2**. The next concern was to ensure diffusion of dye was consistent between SOD1 and wild type motoneurons. Therefore, the baseline (at rest) fluorescent intensity of both Texas red and Ca2+-green-1 was measured throughout neuronal processes, as shown in **Figure 2B**. Differences in baseline [Ca2+]internal could confound the interpretation of the Ca2<sup>+</sup> green signal, but no such complication is exists for Texas red. Texas red showed a significantly lower fluorescent intensity in SOD1 motoneurons (dotted lines) than WT motoneurons (solids lines), starting at 165µm from the soma. Though axonal transport has been shown to be altered in mutant SOD1 motoneurons (Zhang et al., 1997; Warita et al., 1999; Williamson and Cleveland, 1999; Kieran et al., 2005; De Vos et al., 2007; Bilsland et al., 2010), this phenomenon is not likely related. While much is still unknown about the intracellular mechanisms of dextran-conjugated dye spread, it does not rely on microtubule-dependent active transport (Fritzsch, 1993). Rather, passive diffusion of the dye from the soma to the distal dendrites could be impaired in high-protein-expressor SOD1G93A neurons, or alternatively the dye could be more diffuse due to the larger size of SOD1 motoneurons (Amendola and Durand, 2008; Filipchuk and Durand, 2012). While this is in itself an interesting phenomenon, the problem of differential dye filling in measuring Ca2<sup>+</sup> transients was overcome by normalizing the Ca2<sup>+</sup> green-1 signal to the Texas red signal. In this way fluctuations in Ca2<sup>+</sup> green-1 are normalized by the fluctuations of Texas red, since both are the same molecular weight and should be equally mobile in the cell. Thus, all signals are presented as the change in green fluorescent intensity/red fluorescent intensity (G/R Intensity).

### Progressive Reduction in Ca2<sup>+</sup> Signal

Each stimulus protocol was composed of 3 pulses of equal duration: three 50 ms pulses, (as shown in **Figure 3A**), three 100 ms pulses, and three 200 ms depolarizing pulses (shown in **Figure 1C**). Even the Ca2<sup>+</sup> signals evoked with a single spike were quite robust (**Figure 3A**, top two panels and inset). As shown in **Figure 3A**, all Ca2<sup>+</sup> transients returned to baseline after the stimulus ended. Any neurons that showed signs of bleaching or other photodamage were removed from the analysis. Also clearly visible from the Ca2<sup>+</sup> green-1 trace (**Figure 3A**, middle panel), the first stimulus in the protocol elicited the largest response, and subsequent stimuli elicited smaller Ca2<sup>+</sup> transients. Thus, Ca2<sup>+</sup> signals from the 1st, 2nd, and 3rd pulses were plotted separately in **Figures 3B,C**. The graphs reveals two things: (1) an overall larger Ca2<sup>+</sup> signal in the soma and proximal dendrites, and (2) a significantly larger Ca2<sup>+</sup> transient for the first (blue) of the three stimuli at many points in the dendritic processes, even while action potential amplitudes remained consistent across stimuli (as in **Figure 3A** bottom panel). The drop off in Ca2<sup>+</sup> signal could indicate Ca2<sup>+</sup> channel inactivation, or activation of Ca2<sup>+</sup> dependent K channels (SK channels). This phenomenon was most pronounced with the spike trains evoked by the 200 ms pulses, and least pronounced in the 50 ms single-spike stimulus protocol (data shown is average of all protocols).

#### Predictors of Calcium Transient Size

In order to capture anatomical characteristics, several z stacks were acquired after electrophysiological and Ca2<sup>+</sup> measurements. To capture the full extent of the processes, multiple stacks were necessary. Generally, seven stacks were taken of neurons, though the exact number of stacks was dependent on the extent of the processes (range of 2 to 10 stacks). Stack sizes were 308 × 308µm (512 × 512 pixels), and the depth ranged from 75 to 204µm (step size = 1µm). All of the morphological characteristics from the dendrites were compared to Ca2<sup>+</sup> transients using ANCOVA. Ca2<sup>+</sup> transients were found to correlate significantly with three factors: (1) number of spikes evoked/length of depolarizing current pulse, (2) distance from the soma, and (3) diameter of the dendrite at that location. More action potentials (from longer current steps), larger diameter dendrites and more proximal locations all contributed to larger Ca2<sup>+</sup> signals (see **Table 2** for complete breakdown of numbers). The simplest reason that all of those factors were significant is that they all affect the passive spread of current into the dendrites. However, in addition to those three factors, 3 of 7 motoneurons (1



*Each parameter is listed with mean, standard deviation (St Dev), minimum (Min), and maximum (Max). Ca*2<sup>+</sup> *transients are listed in bins based on time, path distance from soma, dendritic diameter, and stimulus duration. Genotype values are also listed. Listed p-values are based on ANCOVA, and indicate the significance only of the listed factor (i.e., time) on Ca*2<sup>+</sup> *transients (the p-value for genotype is listed only under the genotype heading).*

of 3 SOD1G93A and 2 of 4 wild type) showed significant variation from one dendrite to the next. In other words, even when the contributions of branch diameter, distance from the soma and stimulus length were accounted for (using ANCOVAs) 3/7 motoneurons had dendrites with significantly different profiles of Ca2<sup>+</sup> entry. For example, compare the Ca2<sup>+</sup> transients from the P9 wild type motoneuron shown in **Figure 4**. As shown, points A, B, C, and D have rather different Ca2<sup>+</sup> transients. These examples are taken at equal distances from the soma, are averages of three 200 ms stimuli delivered while performing 2PEF at each location of the dendritic field. The dendritic diameter does vary somewhat between these processes, but it cannot account for the magnitude of difference between signals. Only motoneurons in which ≥3 branches had been sampled at overlapping distances were included in the analysis (n = 7). The processes with lowest and highest signals were not consistently first or last in the recording order, and data was eliminated from analysis whenever there was any sign of photodamage (including persistent increase in fluorescence after stimulation cessation; any blebbing of the process) or physiological weakening of the cell (changes in resting potential >10 mV or input resistance >10 M). Rather, the different signals from a dendrite likely reflect Ca2<sup>+</sup> microdomains as first suggested by Ladewig et al. (2003) in hypoglossal motoneurons. These microdomains could arise from altered current spread of the spike, altered presence of Ca2<sup>+</sup> channels from dendrite to dendrite, local neuromodulation of the Ca2<sup>+</sup> channels from one process to the next, or from different interactions of the external Ca2<sup>+</sup> entry with Ca2<sup>+</sup> from internal stores (also see Discussion).

#### Decreased Dendritic Ca2<sup>+</sup> Transients in SOD1G93A Motoneurons

Overall, eight SOD1G93A and 13 wild type motoneurons were recorded. Comparing Ca2<sup>+</sup> signals between these two groups, transients from SOD1G93A motoneurons were significantly smaller (p = 0.000) as shown in **Figure 5**. This decrease appears to be more pronounced distally, likely reflecting a decrease in high voltage-activated Ca2<sup>+</sup> channel activity in distal SOD1G93A dendrites (see also Discussion). **Table 2** includes data for each genotype by time, distance, dendritic diameter and stimulus duration. Potentially this could represent yet another early, homeostatic alteration of an intrinsic property in SOD1 motoneurons.

### Discussion

#### Summary

Motoneurons showed significantly more robust Ca2<sup>+</sup> transients with (1) longer depolarizing steps/increasing numbers of evoked action potentials, (2) larger dendritic diameters, and (3) shorter distances from the soma, as has been documented in other neuron types (Lips and Keller, 1999; von Lewinski and Keller, 2005; Day et al., 2008). Calcium transients with repeated stimuli

events though it returns to baseline after each event. The action potential

showed tapering off of signal, perhaps evidence of Ca <sup>2</sup><sup>+</sup> channel inactivation or activation of SK channels. Remarkably, Ca2<sup>+</sup> signals are highly variable from one dendrite to the next, with 3 of 7 motoneurons showing significantly different profiles of Ca2<sup>+</sup> entry between processes in the same neuron. These results show the complexity of local microdomains in motoneuron processes. Lastly, the Ca2<sup>+</sup> transients were significantly smaller in SOD1G93A motoneurons. This could be due to lower Ca2<sup>+</sup> channel expression/activation, or both.

each location with One Way ANOVA \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001.

### Previous Work on Motoneuron Ca2<sup>+</sup> Currents/Channels

Spinal motoneurons display a wide array of Ca2<sup>+</sup> currents. L-type Cav 1.3 channels contribute to the CaPIC, one of the currents which sets overall neuronal excitability. The PICs are also of particular interest since they are elevated in SOD1 motoneurons at this age. Cav 1.3 channels are found on both proximal and distal dendrites with a patchy distribution, probably at the sites

of synaptic contact (Westenbroek et al., 1998; Carlin et al., 2000b; Simon et al., 2003; Zhang et al., 2008a). The uneven channel distribution and modeling suggest the presence of "hotspots" along the dendritic field where Cav1.3 channels are clustered (Grande et al., 2007). Though our data does appear to show hotspots in the dendritic field, our stimulation protocol is unlikely to activate Cav1.3 type Ca2<sup>+</sup> channels. If we were activating CaPICs, we would probably not see a return to baseline in either the membrane voltage or the Ca2<sup>+</sup> signal. Instead, we do see a repolarization of the membrane and decreasing, rather than increasing Ca2<sup>+</sup> transients with each stimulus. The other L-type Ca2<sup>+</sup> channels, the Cav1.2 channels, are quite likely contributing to Ca2<sup>+</sup> signals in this study. They show intense labeling in motoneuron somata and proximal dendrites (Westenbroek et al., 1998; Simon et al., 2003), and along with the Cav1.3 channels have an increasing presence during postnatal maturation until P18 (Jiang et al., 1999). Due to the strong Ca2<sup>+</sup> signals we observed from these areas and the mean age of the neurons recorded (P8), these channels probably are contributing to the Ca2<sup>+</sup> signals here, along with the other high-voltage-activated Ca2<sup>+</sup> channels Cav2.1 (P/Q) and Cav2.2 (N) type channels. Though classically the Cav2.1 channels mediate vesicle release at axon terminals, they have also been shown to contribute substantially to whole cell Ca2<sup>+</sup> currents measured at motoneuron somata (Umemiya and Berger, 1995; Carlin et al., 2000a). Cav2.2 have been found on motoneuron somata and proximal dendrites based on immunohistochemistry, as have Cav2.3 (R) type channels (Westenbroek et al., 1998), and could also be contributing to our Ca2<sup>+</sup> signals.

### Attenuation of Ca2<sup>+</sup> Transients

As shown in **Figure 3**, successive stimuli evoked smaller Ca2<sup>+</sup> transients. Inactivation of high-voltage activated Ca2<sup>+</sup> channels has been previously described (Catterall et al., 2013). However,

another possibility is that successive activation of Ca2<sup>+</sup> channels during spiking is sufficient to open SK (Ca2+-activated K +) channels. SK channels are present on motoneurons (Safronov and Vogel, 1998), and their activation depends both on voltage and the presence of Ca2<sup>+</sup> (Li and Bennett, 2007; Abbinanti and Harris-Warrick, 2012). As Ca2<sup>+</sup> influx from the first pulse activates SK channels, the increased K<sup>+</sup> conductance would allow less voltage change in subsequent pulses. Dampening the dendritic depolarization would in turn dampen the Ca2<sup>+</sup> signal.

#### Dendrites Are Highly Variable

Landmark studies by Jaffe et al. (1992, 1994) showed Ca2<sup>+</sup> transients throughout the dendrites in hippocampal pyramidal neurons depend on dendritic invasion of the Na<sup>+</sup> spike. The strength of the dendritic Ca2<sup>+</sup> signal was shown to attenuate steeply in distal dendritic compartments that lack voltage dependent Na<sup>+</sup> channels, the degree of decline depending on intracellular resistance of the cell (Jaffe et al., 1992, 1994). In fact, spatial gradients in Ca2<sup>+</sup> signals could be achieved in models without any changes in Ca2<sup>+</sup> channel density throughout the processes (Jaffe et al., 1994). Instead, differing current spread in each compartment could explain the difference between processes. In the case of the spinal motoneurons, the anatomical contributions to varying Ca2<sup>+</sup> signals are compounded by varied Ca2<sup>+</sup> channel density as well. Cav1.3, 2.1, and 2.2 channels have been described as having a patchy distribution, or regions of higher density, along the length of motoneuron dendrites (Westenbroek et al., 2005). Moreover, hypoglossal motoneurons show local Ca2<sup>+</sup> microdomains arising from the concert of activity in voltage-gated Ca2<sup>+</sup> channels, located in varying proximity to Ca2<sup>+</sup> sequestering organelles which are themselves nonuniformly distributed in the cell (Lips and Keller, 1999; Ladewig et al., 2003). Hypoglossal motoneurons and spinal motoneurons appear to share these Ca2<sup>+</sup> microdomains, with spinal motoneurons showing significantly different Ca2<sup>+</sup> signals from one dendrite to the next in 3/7 motoneurons analyzed. Additionally, our data supports the existence of "hotspots" in high voltage gated Ca2<sup>+</sup> channels. Hotspots have been previously described in low voltage gated Ca2<sup>+</sup> channels (Bui et al., 2006; Elbasiouny et al., 2006; Grande et al., 2007; Carlin et al., 2009). Another factor that could also be contributing to the complexity of the Ca2<sup>+</sup> signals is local neuromodulation of Ca2<sup>+</sup> channels. Serotonin (5HT2R), metabotropic glutamate, and adenosine A1 receptor activation can all suppress Cav2 channel activity (Mynlieff and Beam, 1994; Bayliss et al., 1995; Del Negro and Chandler, 1998; Ladewig et al., 2004). Serotonin has even been shown to differentially modulate Ca2<sup>+</sup> signals: Ca2<sup>+</sup> transients at certain dendritic sites are strongly depressed by serotonin while other dendritic sites on the same neuron are unaffected (Diaz-Rios et al., 2007).

#### Ca2<sup>+</sup> Entry and Synaptic Inputs to SOD1 Motoneurons

Motoneurons that are vulnerable to degeneration in ALS (and mutant SOD1) pathology are the largest, fast fatigable motoneurons which lack Ca2<sup>+</sup> buffering proteins (Lips and Keller, 1998, 1999; Palecek et al., 1999b; Pun et al., 2006; Hegedus et al., 2007), in fact they have Ca2<sup>+</sup> buffering capacities that are 5 to 6 times lower than found in disease-resistant motoneurons (Lips and Keller, 1998; Palecek et al., 1999a). The mitochondria are largely responsible for uptake of unbound internal Ca2+. Near disease end stage in hypoglossal motoneurons, Ca2<sup>+</sup> handling is completely remodeled such that there is decreased dependence on the mitochondria for Ca2<sup>+</sup> uptake and increased plasma membrane extrusion (Fuchs et al., 2013). Altered Ca2<sup>+</sup> sequestering and an exaggerated response of the electrical gradient of the inner membrane to stimulation-induced Ca2<sup>+</sup> influx has also been demonstrated (Damiano et al., 2006; Bilsland et al., 2008; Jaiswal et al., 2009; Nguyen et al., 2009; Li et al., 2010). These studies may indicate a lifelong battle for Ca2<sup>+</sup> homeostasis in motoneurons. In the juvenile age studied here, current results suggest that high voltage-activated Ca2<sup>+</sup> channels may be less active (or less prevalent) in SOD1 motoneurons, while in contrast, the Cav1.3 channels responsible for the PIC appear to be more active (Quinlan et al., 2011). Interestingly, cultured wild type spinal motoneurons exposed to SOD1G93A astrocyte conditioned media showed somewhat similar results (Fritz et al., 2013). These motoneurons developed electrical similarities to SOD1G93A motoneurons (including larger PICs). They also showed reduced amplitude dendritic Ca2<sup>+</sup> transients evoked by spontaneous synaptic events (though the frequency of these spontaneous events increased). Since this study used the membrane permeable Ca2<sup>+</sup> indicator Fura-2 AM, dye diffusion would not be a factor here (Fritz et al., 2013). The contribution of the synaptic machinery to the activation of dendritic Ca2<sup>+</sup> channels should not be overlooked. Herron and Miles (2012) describe enlarged C-boutons on motoneurons of male, but not female SOD1 mice. Bories et al. (2007) reported smaller amplitude depolarizations in membrane voltage of spinal motoneurons evoked with dorsal root stimulation, while in hypoglossal motoneurons, van Zundert et al. (2008) observed an increased frequency of spontaneous AMPAergic and GABAergic events with no change in current amplitude. Further study is critically needed to determine the exact contributions of motoneuron Ca2<sup>+</sup> channels and altered synaptic inputs in motoneuron vulnerability. This work is increasingly possible since great strides have been made in identifying neurons which synapse directly onto spinal motoneurons (Butt and Kiehn, 2003; Miles et al., 2007; Quinlan and Kiehn, 2007; Crone et al., 2008; Zhang et al., 2008b; Kasumacic et al., 2010; Bui et al., 2013; Talpalar et al., 2013). Indeed, a recent study indicates alterations in synaptic inputs and the cell signaling pathway mTOR could be a very promising target for treatment of ALS (Saxena et al., 2013).

#### Dendrites of SOD1 Motoneurons

Our results show for the first time that high expressor SOD1 motoneurons may have an impedance to dye filling. It would be tempting to assume dendritic transport is altered similarly to axonal transport (Zhang et al., 1997; Warita et al., 1999; Williamson and Cleveland, 1999; Kieran et al., 2005; De Vos et al., 2007; Bilsland et al., 2010). However, transport of dextrans occurs independent of microtubule-dependent active transport (Fritzsch, 1993). Since protein expression levels in the SOD1 model used here are orders of magnitude higher than in nontransgenic mice, perhaps it has reached a point that impedes diffusion of dye. Alternatively, the dendritic arborization is expanded in SOD1 spinal motoneurons (Amendola and Durand, 2008; Filipchuk and Durand, 2012), so with potentially more dendrites to fill, it could be that the dye was just less effective at filling an expanded area. Despite the limitations in dye filling, many steps were undertaken to normalize Ca2<sup>+</sup> signals between regions that were differently filled with dye, or experienced differences in laser penetration (processes that were deeper in the slice than others). Specifically, use of 3000 MW Texas red to normalize transients detected by 3000 MW Ca2<sup>+</sup> green-1 can control for these variables.

#### Conclusions

Intracellular Ca2<sup>+</sup> has broad impact on neurons, affecting the immediate electrophysiology, the short term cell signaling pathways, and in the long term, neurodegenerative pathways. In the present study we show that dendritic Ca2<sup>+</sup> entry is highly variable even in wild type motoneurons. This likely allows for the fine tuning of synaptic inputs and intrinsic cell currents through neuromodulatory processes and variable expression of dendritic Ca2<sup>+</sup> channels. In the SOD1G93A motoneurons, the complexity of Ca2<sup>+</sup> entry is certainly not diminished. Dendritic Ca2<sup>+</sup> entry can be significantly different from one branch to the next, and the overall Ca2<sup>+</sup> signal in SOD1 motoneurons is smaller.

### References


### Author Contributions

KQ designed, performed and analyzed the experiments, performed statistical analysis, interpreted the data and wrote the manuscript. JL designed the customized software to match transients to their anatomical parameters, performed the data analysis and statistical analysis. JS assisted with data analysis. CH developed the initial conception of these experiments, interpreted the data, and contributed to the manuscript.

#### Acknowledgments

Support was provided by NIH NINDS NS 077863 and NS 089313 (CJH) with additional funding from ALS Association 1626; NIH NINDS F32 NS063535 (KQ); Undergraduate Research Grant from NU Office of the Provost (JL). Authors made use of equipment supported by NS054850.

### Supplementary Material

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fncel.2015. 00139/abstract


presymptomatic ALS mice. J. Neurophysiol. 91, 571–575. doi: 10.1152/jn. 00665.2003


mice with a G93A mutant SOD1 gene. Brain Res. 819, 120–131. doi: 10.1016/S0006-8993(98)01351-1


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Quinlan, Lamano, Samuels and Heckman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Oxidative stress and mitochondrial damage: importance in non-SOD1 ALS

#### **Maria Teresa Carrì 1,2\*, Cristiana Valle2,3 , Francesca Bozzo1,2 and Mauro Cozzolino<sup>4</sup>\***

<sup>1</sup> Department of Biology, Università di Roma Tor Vergata, Rome, Italy

<sup>2</sup> Fondazione Santa Lucia, IRCCS, Rome, Italy

3 Institute of Cell Biology and Neurobiology, IBCN, National Research Council, CNR, Rome, Italy

4 Institute of Translational Pharmacology, National Research Council, CNR, Molecular Mechanisms of Neurodegenerative Diseases, Rome, Italy

#### **Edited by:**

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### **Reviewed by:**

Rasna Sabharwal, University of Iowa, USA Ellen F. Barrett, University of Miami Miller School of Medicine, USA Adrian Israelson, Ben-Gurion University of the Negev, Israel

#### **\*Correspondence:**

Maria Teresa Carrì, Department of Biology, Università di Roma Tor Vergata, Via della Ricerca Scientifica 1, 00133, Rome, Italy e-mail: carri@bio.uniroma2.it; Mauro Cozzolino, Institute of Translational Pharmacology, National Research Council, CNR, Molecular Mechanisms of Neurodegenerative Diseases, Via del Fosso del Cavaliere 100, 00133, Rome, Italy e-mail: mauro.cozzolino@ift.cnr.it

#### **INTRODUCTION**

Amyotrophic Lateral Sclerosis (ALS) is a fatal motor neuron disease occurring less than once in 10000 people and thus classified as a rare disease. ALS is not purely a motor neuron disease and neuroinflammation and muscle degeneration also contribute to the non-cell autonomous death of motor neurons. As for other neurodegenerative diseases, ALS is sporadic in most cases, with a clear genetic origin in about 10% of patients, and is clinically heterogeneous in age and site of disease onset as well as rate of disease progression.

Different mechanisms such as protein aggregation, oxidative stress (OS), mitochondrial damage (MD), excitotoxicity and RNA dysmetabolism seem to contribute to ALS. Evidence on the relevance of these mechanisms comes in large part from models based on the expression of ALS-linked mutant superoxide dismutase 1 (SOD1). However, mutant SOD1 causes less than 1% of ALS cases and the list of mutant proteins found in ALS patients has grown noticeably in the last decade (Marangi and Traynor, 2014), to include a number of proteins related to cell functions as diverse as RNA metabolism (TDP-43, FUS/TLS, Senataxin,

**Abbreviations:** ER, endoplasmic reticulum; ROS, reactive oxygen species; RNS, reactive nitrogen species; PDI, protein disulphide isomerase.

It is well known that mitochondrial damage (MD) is both the major contributor to oxidative stress (OS) (the condition arising from unbalance between production and removal of reactive oxygen species) and one of the major consequences of OS, because of the high dependance of mitochondrial function on redox-sensitive targets such as intact membranes. Conditions in which neuronal cells are not able to cope with MD and OS seem to lead or contribute to several neurodegenerative diseases including Amyotrophic Lateral Sclerosis (ALS), at least in the most studied superoxide dismutase 1 (SOD1)-linked genetic variant. As summarized in this review, new evidence indicates that MD and OS play a role also in non-SOD1 ALS and thus they may represent a target for therapy despite previous failures in clinical trials.

**Keywords: Amyotrophic Lateral Sclerosis, ALS, mitochondria, oxidative stress, neurodegeneration**

Ataxin2, HNRNPA2/B1, ELP3, HNRNPA1), vesicle trafficking (Alsin, FIG4, OPTN, VABP, CHMP2B) and proteasomal function (UBQLN2, VCP). Noticeably, almost half of the familial cases are linked to a non-coding sequence, i.e., to an expanded hexanucleotide repeat in the C9orf72 gene.

Because mutations in SOD1, an ubiquitous superoxide scavenging enzyme, were the first identified cause of familial ALS in 1993, OS has been an obvious candidate to explain the pathogenesis of this disease. MD was also demonstrated in patients and in mutant SOD1 models soon after (reviewed in Cozzolino et al., 2012) and the link between the two, OS and MD, appeared most probable, if not obvious.

We will now briefly review recent evidence that OS and MD are found also in non-SOD1 linked ALS.

#### **OXIDATIVE STRESS IN ALS**

OS biomarkers have been repeatedly found in sporadic ALS patients, which may indicate that abnormal OS is relevant to the pathogenesis of this disease (D'Amico et al., 2013). A number of environmental factors may contribute to the generation of noxious free radicals, both Reactive Oxygen Species (ROS) and Reactive Nitrogen Species (RNS). For instance, transition metal-mediated OS has been proposed to contribute to ALS pathogenesis more than 10 years ago (reviewed in Carrí et al., 2003) and the role of metals is still discussed (Lovejoy and Guillemin, 2014). While no single pro-oxidant factor has emerged as crucially linked to sporadic ALS, recent genetic evidence reinforces the concept that OS may play a main role in familial ALS including in non-SOD1 ALS. For instance, expression of mutant valosin-containing protein (VCP, a member of the type II AAA+ ATPase family with a number of cellular functions including mitochondrial quality control, autophagy, vesicle transport and fusion, 26 S proteasome function, and assembly of peroxisomes) found in ALS patients makes human neuroblastoma SH-SY5Y cells susceptible to OS induced by treatment with L-buthionine sulfoximine, an inhibitor of the synthesis of glutathione (GSH; Hirano et al., 2014). GSH is a free-radical scavenger tripeptide which is among the main regulators of the intracellular redox state. Its levels are lower in the motor cortex of ALS patients than in healthy volunteers *in vivo* (Weiduschat et al., 2014) and decreasing GSH accelerates neurological deficit and mitochondrial pathology in the mutant SOD1 ALS mice model (Vargas et al., 2011). Furthermore, GSH depletion in cultured neurons induces the formation of cytoplasmic TDP-43 inclusions which are found in sporadic ALS patients (Iguchi et al., 2012).

Alterations in markers of OS other than GSH are found as a consequence of the expression of ALS-linked mutations. For instance, expression of mutant TDP-43 in a motor neuronlike cell line induces OS, MD and nuclear accumulation of nuclear factor E2-related factor 2 (Nrf2) that is an indicator and modulator of OS (Duan et al., 2010); in a yeast model, TDP-43-expressing cells display increased markers of OS, apoptosis, and necrosis and these cytotoxic effects are potentiated upon expression of disease-associated variants (Braun et al., 2011) and *Drosophila* flies engineered for motoneuron-directed expression of TDP-43 have increased levels of protein carbonylation and Glutathione S transferase D1, both known markers of OS (Zhan et al., 2015). Overall these data indicate that OS is important also for TDP-43-triggered cell death, although the mechanism is still debated.

OS may not only increase the net production of ROS and RNS, but also affect protein conformation and structure, leading to the accumulation of the abnormal protein inclusions that are extensively described in ALS mouse models and patientderived tissue. Cysteine-mediated aggregation of mutant SOD1 has been widely studied and wild type SOD1 also hyperaggregates when oxidized (Guareschi et al., 2012). Treatments that deplete the cellular pool of GSH exacerbate mutSOD1s insolubility, whereas an overload of intracellular GSH or overexpression of glutaredoxins 1 and 2 significantly rescues mutSOD1s solubility in the cytosol and in mitochondria, respectively (Ferri et al., 2010). Interestingly, recent evidence suggests that also wild-type and mutant TDP-43 aggregation is caused by incorrect disulphide bonds involving Cys residues in one of its RNA recognition motifs, and that aggregation is promoted by OS (Cohen et al., 2012; Shodai et al., 2013).

Aggregation of TDP-43 (Parker et al., 2012) and FUS (Gerbino et al., 2013; Vance et al., 2013) proceeds, at least in part, through the stress granules (SGs) pathway. SGs are highly dynamic structures that are formed upon OS and contain RNA-binding proteins, transcription factors, RNA helicases and nucleases that work as sorting granules for mRNAs undergoing degradation, storage or translation (Baron et al., 2013 and references therein). SGs are found as a consequence of FUS mutation and mutated FUS is more rapidly directed to SGs after OS than wild type FUS (Bosco et al., 2010). Furthermore, TDP-43 is recruited to SGs in conditions of OS (Colombrita et al., 2009). These SGs may simply sequester a subset of mRNAs thus inducing cell dysfunction or serve as nucleation site for larger protein aggregates as the ones found in ALS patients.

Possibly representing a cellular response to this kind of aggregation and as a reaction against ER stress and unfolded protein response (UPR), protein disulfide isomerase (PDI) expression levels are upregulated in spinal cord tissue of ALS patients where PDI co-localizes with SOD1, TDP-43, and FUS (Atkin et al., 2008; Honjo et al., 2011; Farg et al., 2012; Walker et al., 2013). PDI co-localizes also with VAPB inclusions in a *Drosophila melanogaster* model of ALS (Tsuda et al., 2008). However, PDI itself contains an active site thiol group that is necessary for its activity and it is found oxidized through Snitrosylation in ALS (Walker et al., 2010), which may lead to a cycle of PDI upregulation and inactivation that perpetuates redox dysregulation and protein aggregation.

Interestingly, even mutant C9orf72 repeats may be related to OS. Indeed, similarly to what happens when expressing mutant SOD1, motor neurons differentiated from patients' iPSC (induced Pluripotent Stem Cells) and expressing expanded C9orf72 repeat display a significant induction of catalase, which is indicative of OS, and a significant change in levels of the mitochondrial transporter MTX3 (Kiskinis et al., 2014).

Mitochondria are a known target of OS because of their high dependance on membrane integrity and because they possess their own DNA and RNA that may be damaged by oxidation as well.

#### **MITOCHONDRIAL DAMAGE IN ALS**

Keeping in mind that mitochondria are not only a target of OS, but also the main site of production of ROS, it is obvious that OS may result from impairment of mitochondrial function.

A number of studies have reported altered mitochondrial respiratory complex activity in ALS tissues including postmortem brain and spinal cord tissue, patient lymphocytes, and in the SOD1 transgenic mouse model of ALS (for a review Cozzolino and Carrì, 2012; Tan et al., 2014). MD may follow the accumulation of aggregated SOD1 associated to mitochondria (Ferri et al., 2006) and result from the interaction of misfolded SOD1 with mitochondrial partners (Israelson et al., 2010; Li et al., 2010 and references therein). Evidence is accumulating that MD takes place also as a consequence of the expression of other ALS-related proteins, some of which are directly linked to the function of these organelles. For instance, mutations in VCP account for about 2% of familial ALS and R155H/+ *VCP* knock-in mice show signs of motor damage due to muscle denervation and degeneration accompanied by extensive accumulation of abnormal mitochondria in the intermyofibrillar space; in this model slow motor neuron loss is accompanied by TDP-43 accumulation and extensive aggregation of mitochondria (Yin et al., 2012; Nalbandian et al., 2013). In a 2013 study in VCP-deficient cells and in fibroblasts with *VCP* mutations, Bartolome et al. demonstrated that VCP deficiency causes significant mitochondrial uncoupling, leading to decreased mitochondrial membrane potential, increased mitochondrial oxygen consumption and reduction of cellular ATP production (Bartolome et al., 2013).

ALS may derive from mitochondrial DNA instability, as suggested in a study on a family with a late-onset phenotype including motor neuron disease and cognitive decline resembling frontotemporal dementia. Those patients carried a missense mutation in the CHCHD10 gene that encodes a mitochondrial protein located in the intermembrane space and enriched at cristae junctions. Muscle fibers from those patients are raggedred and cytochrome c oxidase-negative with respiratory chain deficiency and abnormal assembly of complex V. Their fibroblasts have respiratory chain deficiency, mitochondrial ultrastructural alterations and fragmentation of the mitochondrial network (Bannwarth et al., 2014).

Neuronal mitochondrial abnormalities occur also in models of familial ALS associated with proteins that are (at least apparently) not related with these organelles. MD is observed together with oxidative damage and induction of mitophagy in the motor neuron-like NSC34 cell line expressing wild type or mutant TDP-43 (Duan et al., 2010; Hong et al., 2012). Mice expressing wild-type human TDP-43 show a pathology resembling ALS and motor neurons from these mice display cytoplasmic TDP-43 positive inclusions composed of mitochondria aggregates (Xu et al., 2010), that may arise from defective intracellular trafficking and result in the reduction of mitochondria at nerve terminals of neuromuscular junctions (Shan et al., 2010). However, in mice expressing the mutant TDP-43(A315T) protein, morphological abnormalities appear after the onset of transport defects (Magrané et al., 2014). In another study, MD in heterozygous TDP-43(A315TKi) animals is indicated by the formation of abnormal neuronal mitochondrial cristae and decrease in expression of Parkin and the fatty acid transporter CD36 along with an increase in fatty acids, HDL cholesterol, and glucose in the blood (Stribl et al., 2014). TDP-43 also perturbs ERmitochondria interactions and this is associated with disruption in cellular Ca2<sup>+</sup> homeostasis (Stoica et al., 2014). Disturbance of mitochondrial Ca2<sup>+</sup> transport in ALS have been widely described in SOD1 models *in vitro* and *in vivo* by the group of B. Keller (Jaiswal and Keller, 2009; Jaiswal et al., 2009) and by others; this aspect has been recently reviewed in detail (Barrett et al., 2014; Tadic et al., 2014) and thus we will not discuss it in this paper.

Not much evidence of functional MD in FUS-linked ALS is reported. However, mitochondria are disorganized in postmortem neurons from juvenile ALS patients (Huang et al., 2010) and shorter in cultured motor neurons expressing cytoplasmic FUS mutants (Tradewell et al., 2012). Furthermore, overexpression of ALS-linked human mutant FUS leads to Golgi fragmentation and mitochondria aggregation in rats (Huang et al., 2012).

How the overexpression or mutations of two proteins, TDP-43 and FUS, which are clearly involved in physiological mRNA metabolism, may result into MD in ALS is not clear at all.

One possibility is that misregulation of some step of RNA maturation leads to altered expression of proteins involved in mitochondria homeostasis. This may be the case for FUS, which transcriptionally regulates the expression of OS protection genes through the interaction with peroxisome proliferatoractivated receptor coactivator 1-α (PGC-1α; Sánchez-Ramos et al., 2011), that is a known regulator of mitochondrial biogenesis and function. In this context, MD may arise indirectly from modulation of Sirtuins (class III histone deacetylases). In fact PGC-1α is downregulated in ALS patients and it is known that silencing of PGC1α reduces the expression of SIRT3, which is the main mitochondrial deacetylase with a number of substrates (including SOD2), whose main function is to counteract ROS production and detoxification (Bause and Haigis, 2013), while overexpression of SIRT3 stimulates the expression of PGC1α causing a further decrease of ROS in a positive feedback loop (Kong et al., 2010). Interestingly, both SIRT3 and PGC-1α protect against mitochondrial fragmentation and neuronal cell death in rat spinal cord motor neurons overexpressing ALS-linked mutant SOD1-G93A (Song et al., 2013). Furthermore, nuclear SIRT1 also induces the expression of OS response genes and promotes mitochondrial biogenesis by activating PGC-1α (Hall et al., 2013). Nonetheless, SIRT1 overexpression does not protect neurons against toxicity induced by mutant G93A-SOD1 (Valle et al., 2014 and reference therein).

Another possibility is that, similarly to mutant SOD1, TDP-43 and FUS have a pathological tendency to associate with mitochondria and to aggregate. Indeed, both wild type and a truncated form of TDP-43 were found localized in the mitochondria in NSC34 cells (Hong et al., 2012) and TDP-43 aggregates around mitochondria in a yeast model where altered mitochondrial respiratory activity is observed (Braun et al., 2011). Furthermore, FUS interacts with several proteins involved in mitochondrial metabolism and significantly reduces ATP levels in cells with FUS accumulation (Wang et al., 2015).

#### **CONCLUSIONS**

Previous failures in translating antioxidant and mitochondriaprotective strategies that were effective in the mutant SOD1 mouse model into positive clinical trials has cast a doubt on the real relevance of OS and MD in ALS and suggested that the SOD1 model represents only a fraction of a more heterogeneous population. However, new evidence has accumulated in recent years supporting the view that MD and OS play a role also in non-SOD1 ALS (**Figure 1**). This evidence is still somewhat sparse and often comes from studies in cell or invertebrate models that may represent only a few aspects of the human pathology. Nonetheless, while it has become clear that ALS is a complex disease with various presentation including an overlap with fronto-temporal dementia (Talbot, 2014), OS and MD may still represent a common denominator of motor neuron degeneration in ALS, and therefore a valuable target

that encourages transfer of new knowledge from research into preclinical testing.

#### **ACKNOWLEDGMENTS**

MTC is supported by ARiSLA (Project OligoALS) and by the Bruno and Ilse Frick Foundation for ALS research 2012 Award; MC is supported by Ministero della Salute (Project RF-2010- 2309849) and by AriSLA (Project FUSMALS).

We wish to apologize to the many authors whose original contributions have not been cited owing to space restrictions.

#### **REFERENCES**


Bause, A. S., and Haigis, M. C. (2013). SIRT3 regulation of mitochondrial oxidative stress. *Exp. Gerontol.* 48, 634–639. doi: 10.1016/j.exger.2012.08.007


protein composition and decreases protein import. *Proc. Natl. Acad. Sci. U S A* 107, 21146–21151. doi: 10.1073/pnas.1014862107


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 23 December 2014; accepted: 27 January 2015; published online: 17 February 2015*.

*Citation: Carrì MT, Valle C, Bozzo F and Cozzolino M (2015) Oxidative stress and mitochondrial damage: importance in non-SOD1 ALS. Front. Cell. Neurosci. 9:41. doi: 10.3389/fncel.2015.00041*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 Carrì, Valle, Bozzo and Cozzolino. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

# Reactive oxygen species trigger motoneuron death in non-cell-autonomous models of ALS through activation of c-Abl signaling

*Fabiola Rojas1†, David Gonzalez1†, Nicole Cortes1, Estibaliz Ampuero1, Diego E. Hernández2, Elsa Fritz1, Sebastián Abarzua1, Alexis Martinez2, Alvaro A. Elorza1,3, Alejandra Alvarez2, Felipe Court2 and Brigitte van Zundert1\**

#### *Edited by:*

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### *Reviewed by:*

*Tibor Kristian, University of Maryland School of Medicine, USA Sonia Cortassa, Johns Hopkins University, USA Nuno Raimundo, University Medical Center Gottingen, Germany*

#### *\*Correspondence:*

*Brigitte van Zundert, Center for Biomedical Research, Faculty of Biological Sciences and Faculty of Medicine, Universidad Andres Bello, Avenida Republica 217, Santiago, Chile bvanzundert@unab.cl*

> *†These authors have contributed equally to this work.*

> > *Received: 01 February 2015 Accepted: 11 May 2015 Published: 09 June 2015*

#### *Citation:*

*Rojas F, Gonzalez D, Cortes N, Ampuero E, Hernández DE, Fritz E, Abarzua S, Martinez A, Elorza AA, Alvarez A, Court F and van Zundert B (2015) Reactive oxygen species trigger motoneuron death in non-cell-autonomous models of ALS through activation of c-Abl signaling. Front. Cell. Neurosci. 9:203. doi: 10.3389/fncel.2015.00203* *<sup>1</sup> Center for Biomedical Research, Faculty of Biological Sciences and Faculty of Medicine, Universidad Andres Bello, Santiago, Chile, <sup>2</sup> Faculty of Biological Sciences, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>3</sup> Millennium Institute of Immunology and Immunotherapy, Santiago, Chile*

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease in which pathogenesis and death of motor neurons are triggered by non-cell-autonomous mechanisms. We showed earlier that exposing primary rat spinal cord cultures to conditioned media derived from primary mouse astrocyte conditioned media (ACM) that express human SOD1G93A (ACM-hSOD1G93A) quickly enhances Na<sup>v</sup> channelmediated excitability and calcium influx, generates intracellular reactive oxygen species (ROS), and leads to death of motoneurons within days. Here we examined the role of mitochondrial structure and physiology and of the activation of c-Abl, a tyrosine kinase that induces apoptosis. We show that ACM-hSOD1G93A, but not ACM-hSOD1WT, increases c-Abl activity in motoneurons, interneurons and glial cells, starting at 60 min; the c-Abl inhibitor STI571 (imatinib) prevents this ACM-hSOD1G93A-mediated motoneuron death. Interestingly, similar results were obtained with ACM derived from astrocytes expressing SOD1G86R or TDP43A315T. We further find that co-application of ACM-SOD1G93A with blockers of Na<sup>v</sup> channels (spermidine, mexiletine, or riluzole) or anti-oxidants (Trolox, esculetin, or tiron) effectively prevent c-Abl activation and motoneuron death. In addition, ACM-SOD1G93A induces alterations in the morphology of neuronal mitochondria that are related with their membrane depolarization. Finally, we find that blocking the opening of the mitochondrial permeability transition pore with cyclosporine A, or inhibiting mitochondrial calcium uptake with Ru360, reduces ROS production and c-Abl activation. Together, our data point to a sequence of events in which a toxic factor(s) released by ALS-expressing astrocytes rapidly induces hyperexcitability, which in turn increases calcium influx and affects mitochondrial structure and physiology. ROS production, mediated at least in part through mitochondrial alterations, trigger c-Abl signaling and lead to motoneuron death.

Keywords: ALS, non-cell-autonomous, motor neuron, mitochondria, reactive oxygen species (ROS), c-Abl

### Introduction

Amyotrophic lateral sclerosis (ALS) is a fatal paralytic disorder caused by the progressive degeneration of upper and lower motoneurons during adulthood, and results in death by respiratory failure, usually within 3–5 years of diagnosis. The majority of ALS cases are sporadic (sALS), but ∼10% are familial (fALS) and are generated by mutations in at least 24 identified ALS-associated gene loci, including those for superoxide dismutase (SOD1) and transactive response DNAbinding protein 43 (TDP-43), as well as by hexanucleotide expansion in C9orf72 (Bento-Abreu et al., 2010; DeJesus-Hernandez et al., 2011; Ferraiuolo et al., 2011; Renton et al., 2011, 2014; Wegorzewska and Baloh, 2011; Sreedharan and Brown, 2013). Much of our understanding of ALS, however, is based on a subgroup of fALS patients who carry mutations in the gene that encodes SOD1; the exact mechanisms how SOD1 becomes toxic have not been elucidated (Cleveland and Rothstein, 2001; Pasinelli and Brown, 2006). *In vivo* and *in vitro* studies with transgenic ALS animal models (especially with the SOD1G93A mice) yielded the identification of pathogenic changes in affected motoneurons: these alterations include mitochondrial dysfunction, hyper-excitability, glutamate excitotoxicity, nitroxidative stress from reactive oxygen species (ROS) or reactive nitrogen species (RNS; collectively leading to nitroxidative stress), protein aggregation and misfolding, proteasome impairment, cytoskeletal disruption, activation of cell death signals, and dysregulation of transcription and RNA processing (Cleveland and Rothstein, 2001; Bruijn et al., 2004; Pasinelli and Brown, 2006; Ferraiuolo et al., 2011; Cozzolino et al., 2012; van Zundert et al., 2012).

An increasing number of *in vitro* studies report that astrocytes that express mutant SOD1 selectively kill motoneurons through non-cell-autonomous toxicity (Vargas et al., 2006; Di Giorgio et al., 2007; Nagai et al., 2007; Cassina et al., 2008; Marchetto et al., 2008; Haidet-Phillips et al., 2011; Castillo et al., 2013; Fritz et al., 2013; Meyer et al., 2014; Re et al., 2014; Rojas et al., 2014). Studies of interactions between neurons and astrocytes suggest that similar pathogenic changes occur in human ALS patients and in transgenic ALS models, including those that are based on mitochondrial dysfunction, hyperexcitability, and nitroxidative stress, underscoring the value of these model systems (Cassina et al., 2008; Marchetto et al., 2008; Fritz et al., 2013; Rojas et al., 2014). Despite substantial progress in the identification of pathogenic changes, as well as of the cell types that contribute to them, no cure exists for this profoundly debilitating disease, and the mechanisms that underlie motoneurons death in ALS remain largely unknown; in fact, we do not know even whether the neuronal abnormalities are a primary or secondary event, or whether they result from a compensatory mechanism (van Zundert et al., 2012). In part this is because classical approaches for studying neuron-glial interactions use a co-culture system wherein neurons are grown on a feeder layer of astrocytes, thus masking the temporal interplay between original and secondary pathogenic events.

To circumvent this, we use astrocyte conditioned media (ACM), secreted by primary astrocytes derived from transgenic ALS mouse models (including ACM-SOD1G93A), and expose primary wild-type (WT) rat spinal cord cultures (4 DIV) to this ACM for varying times (mins, hours, days). Use of this *in vitro* system, along with electrophysiological recordings, calcium imaging, immunostaining and pharmacology, led us to determine that applying ACM-hSOD1G93A, but not ACM-hSOD1WT, to primary WT spinal cord cultures rapidly increases the neurons' evoked action potentials (eAPs; starting at 15–20 min after ACM exposure); this is followed by calcium influx and the generation of intracellular nitroxidative stress (ROS/RNS; starting at 30 min), thereby leading to specific and robust motoneuron death within days (Fritz et al., 2013; Rojas et al., 2014).

Here we wanted to establish the mechanisms whereby ROS/RNS mediates pathogenesis and death of motoneurons, and focused on the interplay between oxidative stress, mitochondrial structure and physiology, and c-Abl activation; these processes are linked to ALS pathology and influenced by ROS. Although the link between oxidative stress and impaired mitochondrial function has been established in diverse ALS model systems (von Lewinski and Keller, 2005; Grosskreutz et al., 2010; Carrì and Cozzolino, 2011; Drechsel et al., 2012; Tan et al., 2014), use of conventional co-culture systems has yielded little about the causal relationship between the stress and the dysfunction. Previous studies also have implicated active c-Abl in a variety of neurodegenerative diseases, including in ALS (Katsumata et al., 2012), Alzheimer's disease (Alvarez et al., 2004; Cancino et al., 2008; Estrada et al., 2011; Gonzalez-Zuñiga et al., 2014), and Parkinson's disease (Ko et al., 2010; Imam et al., 2011, 2013).

In addition to its classic function in leukemia pathogenesis, the c-Abl no-receptor tyrosine kinase plays a role in neuronal development and is required for the proper functioning of differentiated neurons (Moresco and Koleske, 2003; Hernández et al., 2004; Bradley and Koleske, 2009). Activated c-Abl has important roles in neuronal cytoskeleton remodeling, promotes dendritogenesis, and regulates adhesion, migration and growth cone path-finding (Koleske et al., 1998; Lu et al., 2002; Rhee et al., 2002; Woodring et al., 2002; Moresco et al., 2003); these processes, as well as cell death, are dependent on the activation c-Abl by of phosphorylation of its tyrosine 245 (Tyr245) and tyrosine 412 (Tyr412; Zhu and Wang, 2004; Gonfloni et al., 2012). In the hippocampus, c-Abl is localized in both the preand postsynaptic regions and regulates synaptic structure and function (Moresco et al., 2003; Perez de Arce et al., 2010; Vargas et al., 2014). The contribution of c-Abl signaling activation to neuronal apoptosis has also been reported. For example, it has been shown that c-Abl regulates the choice between cell survival in an arrested state and apoptosis (Wang, 2005), controlling the function and stabilization of p73 in response to genotoxic stress (Tsai and Yuan, 2003). In addition, c-Abl and Cdk5 cooperatively regulate the maximal activation of p53, which results in neuronal apoptosis in response to oxidative stress by hydrogen peroxide (Levav-Cohen et al., 2005; Lee et al., 2008). c-Abl is activated by a wide range of stimuli, including inflammation, DNA damage, amyloid beta, and oxidative stress (van Etten, 1999; Klein et al., 2011; Schlatterer et al., 2011b).

In the present study, we used the ACM-SOD1G93A *in vitro* model system, together with immunostaining, real-time imaging with fluorescent markers for nitroxidative stress and mitochondrial membrane depolarization, electron microscopy, as well as pharmacological treatments, to demonstrate that c-Abl activation and mitochondrial swelling and membrane depolarization play key roles in the pathogenesis and death of motoneurons induced by toxic factor(s) released from SOD1G93A-expressing astrocytes. Our findings suggest that mitochondria are an important, but not an exclusive, source of ROS/RNS production which activates the c-Abl signaling pathway. And finally, we use diverse compounds that reduce Nav channel activity and extracellular calcium levels, to unveil that hyper-excitability and calcium influx into the cytoplasm occur upstream of ROS/RNS production, alterations on the mitochondrial structure and membrane potentiation, and c-Abl activation.

### Materials and Methods

### Animals

Care and use of rodents was in accordance with the US National Institute of Health guidelines, and was approved by the Institutional Animal Care and Use Committee of Andres Bello University. Hemizygous transgenic mice carrying mutant human SOD1G93A (high copy number; B6SJL; Cat. No. 002726) and WT human SOD1WT (B6SJL; Cat. No. 002297) were originally obtained from Jackson Laboratories (Bar Harbor, ME, USA). Non-transgenic littermates and transgenic mice over-expressing the gene for human SOD1WT were used as controls. Transgenes were identified by polymerase chain reaction (Wegorzewska et al., 2009; Castillo et al., 2013; Fritz et al., 2013). The SOD1G93A mice, but not the hSOD1WT mice, develop signs of neuromuscular deficits (tremor of the legs and loss of extension reflex of the hind paws) starting at 3 months of age and have an average lifespan of 19–21 weeks (Gurney et al., 1994). Mice carrying SOD1G86R (Ripps et al., 1995) or TDP43A315T (Wegorzewska et al., 2009) develop similar loss of motor function between 3 and 4 months and do not survive to the age of 4 months.

### Conditioned Media Preparation

Astrocyte-onditioned media was prepared as described (Nagai et al., 2007; Castillo et al., 2013; Fritz et al., 2013; Rojas et al., 2014). Briefly, cultures of astrocytes were prepared from P1-2 WT mice and from ALS transgenic mice. Cultures were maintained in DMEM (Hyclone, Cat. No. SH30081.02) containing 10% FBS (Hyclone, Cat. No. SH30071.03; lot ATC31648) and 1% penicillin–streptomycin (Gibco, Cat. No. 15070-063) at 37◦C 5% CO2. Cultures reached confluence after 2–3 weeks and contained >95% GFAP+ astrocytes. Residual microglia were removed by shaking cultures in an orbital shaker (200 r.p.m. in the incubator) overnight (7 h), at which point media was replaced by spinal culture media (see below). After 7 days, ACM was collected, centrifuged (500 *g* for 10 min) and stored at −80◦C; before use, it was supplemented with 4.5 mg ml−<sup>1</sup> D-glucose (final concentration) and penicillin/streptomycin, and filtered. A chick

While in Nagai et al. (2007) the ACM-hSOD1G93A is added to the cultures undiluted, for all our experiments presented here and previously the ACM-hSOD1G93A as well as the ACM-hSOD1WT was diluted 8–10 fold. The exact dilution was determined for each new batch of ACM by comparing the motoneuron toxicity of the ACM from transgenic animals carrying the ALScausing mutants (ACM-hSOD1G93A, ACM-SOD1G86R or ACM-TDP43A315T) to that of ACM generated from mice carrying the WT human SOD1 gene (ACM-hSOD1WT) or from nontransgenic littermates; at the selected dilutions the conditioned media derived from the astrocytes expressing the ALS-causing genes robustly killed motoneurons, whereas the control media did not affect motoneuron survival. The ACM was applied to ventral spinal cord cultures derived from rats because better quality motoneurons are obtained from rats than from mice; a number of studies have shown that such mixed species cocultures (from rat, mice, human) do not appear to induce any side effects (e.g., Pehar et al., 2004; Di Giorgio et al., 2007; Nagai et al., 2007; Castillo et al., 2013; Fritz et al., 2013; Re et al., 2014; Rojas et al., 2014).

#### Primary Spinal Cord Neuronal Cultures

Pregnant Sprague-Dawley rats were deeply anesthetized with CO2, and primary spinal cultures were prepared from E14 pups (Sepulveda et al., 2010; Fritz et al., 2013; Rojas et al., 2014). Briefly, whole spinal cords were excised and placed into ice-cold HBSS (Gibco, Cat. No. 14185-052) containing 50 μg/ml penicillin/streptomycin (Gibco, Cat. No. 15070-063). The dorsal part of the spinal cord was removed using a small razor blade, and the ventral cord was minced and enzymatically treated by incubating in pre-warmed PBS 1x containing 0.25% trypsin (Gibco, Cat. No. 15090-046) for 20 min at 37◦C. Cells were transferred to a 15 ml tube containing neuronal growth media containing 70% MEM (Gibco, Cat. No. 11090- 073), 25% Neurobasal media (Gibco, Cat. No. 21103-049), 1% N2 supplement (Gibco, Cat. No. 17502-048), 1% L-glutamine (Gibco, Cat. No. 25030-081), 1% penicillin–streptomycin (Gibco, Cat. No. 15070-063), 2% horse serum (Hyclone, Cat. No. SH30074.03; lot AQH24495) and 100 mM sodium pyruvate (Gibco, Cat. No. 11360-070); they were precipitated, transferred to a new 15-ml-tube containing 2 ml of growth media, resuspended by mechanical agitation through fire-polished glass Pasteur pipettes of different tip diameters, and counted; 4,8 × 10<sup>6</sup> cells were plated on freshly prepared poly-L-lysine-coated 24-well plates (1 mg/ml; 30.000–70.000 MW; Sigma, Cat. No. P2636). Cells were cultured for 7 days at 37◦C under 5% CO2, and supplemented with 45 μg/ml chick leg extract (Sepulveda et al., 2010); the media was refreshed every 3 days.

#### Cell Survival Analysis

To measure survival of motoneurons and interneurons, cultures were immunolabeled and counted as previously described (Sepulveda et al., 2010; Castillo et al., 2013; Fritz et al., 2013; Rojas et al., 2014). Briefly, primary spinal cultures were fixed at 7 DIV with 4% paraformaldehyde, and immunostained with a rabbit polyclonal antibody against MAP2 (1:400; Cat. No. sc-20172; Santa Cruz Biotechnology) to label all neurons (interneurons plus motoneurons) and with a mouse monoclonal SMI-32 antibody (1:600, Cat. No. SMI-32R; Sternberger Monoclonals) to reveal the presence of unphosphorylated neurofilament-H, which is expressed specifically in motoneurons in spinal cord cultures (Urushitani et al., 2006; Nagai et al., 2007); previously we found that our WT primary spinal cultures typically contain at least 8–10% motoneurons until 12 DIV (Sepulveda et al., 2010). Fluorescent neurons were visualized with epifluorescent illumination on an Olympus IX81 microscope or on a Nikon C1 confocal microscope on which stacks of 0.50-μm optical sections were acquired through entire neurons. Labeling patterns were documented with a 20x objective and a Q-Imaging Micropublisher 3.3 Real-Time Viewing camera; MAP2- and SMI-32-positive neurons were counted off-line within 20 randomly chosen fields, and the percentage of SMI-32 positive motoneurons within the total number of MAP2-positive cells was calculated. Each condition was replicated in at least three independent cultures and in duplicate.

#### Pharmacological Treatments in Culture

Mexiletine (Tocris, Cat. No. 2596) was dissolved in water to 100 mM and used at final concentration of 25 nM. Riluzole (Sigma, Cat. No. R116) was dissolved in distilled water (plus 10% Tween20) at 100 μM, and added to cultures to final concentration of 100 nM. Spermidine (Sigma, Cat. No. S2626) was dissolved in water at 100 mg/ml and added to cultures to a final concentration of 10 μM. Trolox (Sigma, Cat. No. 238813) was dissolved in distilled water at 100 mM and added to cultures to final concentration of 1 μM. Esculetin (Sigma, Cat. No. 17795) was dissolved in dimethyl sulfoxide (DMSO), and added to cultures to final concentration 25 μM. Tiron (Sigma, Cat. No. D7389) was dissolved in distilled water at 100 mM and added to cultures to final concentration of 25 μM. Ru360 (Calbiochem Cat. No 557440) was dissolved in water at 0.5 mg/mL and added to cultures to final concentration 5 μM. Cyclosporin A (CsA), (Sigma Cat. No. 30024) was dissolved in DMSO at 50 mM, and added to cultures to final concentration 10 μM. EGTA (Sigma Cat. No. E3839) was dissolved at 100 mM in NaOH 1M, and added to cultures to a final concentration of 200 μM. All stock solutions were stored at −20◦C.

#### c-ABL Immunofluorescence Labeling

For identification c-Abl phosphorylation in specific cell types, primary spinal cultures were fixed at 7 DIV with 4% paraformaldehyde, and immunostained with a rabbit polyclonal antibody against MAP2 and SMI-32 antibody, as indicated above in the section "cell survival analysis." A mouse monoclonal antibody anti-glial fibrillary acidic protein (GFAP; 1:600, Sigma, Cat. No. G393) was also used. For detecting active c-Abl, a mouse monoclonal antibody recognizing phosphorylated Tyr-412 (1:1000 for slices; 1:600 for cultures *in vitro*; 1:2000 for western blot; Sigma; Cat. No. C5240) was used. For immunostainings in tissue, hSOD1G93A (>P120) and hSOD1G86R (P95) transgenic mice were sacrificed, perfused with 4% PFA and sectioned in crytostat obtaining 40 μm slices. All antibody bindings were visualized with the appropriate Alexa fluorescent secondary antibodies (1:500; Life Technologies). Our WT primary spinal cord cultures typically contain at least 6–10% motoneurons until 12 DIV (Sepulveda et al., 2010). Immunolabeled neurons were documented on an upright Olympus Fluoview 1000 confocal microscope (60x oil objective) or a Nikon Eclipse Ti-U microscope equipped with a SPOT PursuitTM USB Camera CCD (14-bit), Epi-fl Illuminator, mercury lamp and Sutter Smart-Shutter with a lambda SC controller. For quantification of cell number, cultures were photographed using a 20x objective; MAP2- and SMI-32-positive neurons were counted offline within 20 randomly chosen fields, and the percentage of SMI-32 positive motoneurons within the total number of MAP2-positive cells was calculated. Each condition was replicated in at least three independent cultures, and in duplicate. For phosphoc-Abl quantification in cultures, fluorescence intensity was quantified using ImageJ software (NIH, Bethesda, MD, USA). Briefly, cell body was marked manually to set a region of interest. The mean fluorescence was quantified for each cell and the background was subtracted choosing a region without cells. The fluorescence corresponding to control cells were normalized at 1.

#### Mitochondrial Membrane Potential Measurements with TMRM

Changes in mitochondrial membrane potential (mt-) were determined with the potentiometric dye tetramethyl rhodamine methyl ester, TMRM (Molecular Probes, Cat. No. T-668). At 4 DIV primary ventral spinal cord cultures were washed twice with Hanks solution (Invitrogen, Cat. No. 4025134), loaded in an incubator for 30 min at 37◦C, 5% CO2 and in the dark with 50 nM of the TMRM dye diluted in Hanks solution. Immediately after, the fluorescent signal was acquired at two time points: before (time 0 min) and 30 min following ACMhSOD1G93A addition. The excitation and emission wavelengths of TMRM fluorescent dye are 550 and 575 nm, respectively. The pictures were taken using Olympus IX81 (Olympus) microscope equipped with a digital camera Orca-R2 (Hamatsu) at the 100X magnification. Scale bar 20 μm. At least three independent fields were acquired for each condition and at least 10 cells were used for quantification of the fluorescence signal. Cells were marked by drawing a region of interest around the cell body, and mean fluorescence intensity was calculated for each cell after subtraction of the background signal using the image analysis module in ImageJ software.

### Nitroxidative Stress Measurements with CM-H2DCF-DA

The intracellular levels of ROS/RNS were measured with CM-H2DCFDA (Invitrogen, Cat. No. C6827) as previously described (Rojas et al., 2014). H2DCF-DA is not a specific probe for a specific oxidant and has been used to monitor certain ROS/RNS (see Discussion). The CM-H2DCF-DA stock solution (1 mM) was prepared in DMSO and was diluted in the culture medium to a final concentration of 1 μM just before addition to the cells. After application of the diverse ACMs to the spinal cord cultures for different time (minutes-hours-days), cells were washed Rojas et al. Role for c-Abl in ALS

(PBS 1x) to remove the ACMs and exposed to CM-H2DCF-DA for 30 min at 37◦C in dark, to label both motoneurons and interneurons. To facilitate the CM-H2DCF-DA membrane penetration, 0.004% Pluronic acid F-127 (Invitrogen, Cat. No. P-3000MP) was added to the culture medium to facilitate dye entry, eliminate possible hydrolysis of dyes by external esterases and maintain better cell integrity (Appaix et al., 2012). After the incubation time, the CM-H2DCF-DA-cointaing culture medium was removed and cultures were washed twice with PBS 1X and suspended in culture medium (500 μl final volume). Next, cells were immediately imaged using an upright Nikon Eclipse Ti-U microscope equipped with a SPOT PursuitTM USB Camera CCD (14-bit), Epi-fl Illuminator, mercury lamp and Sutter Smart-Shutter with a lambda SC controller. Cells were photographed using a 20x objective. As CM-H2DCF-DA is a non-fluorescent dye it passively diffuses into cells and is hydrolyzed intracellularly to the DCFH carboxylate anion that is trapped inside; oxidation of DCFH results in the formation of the fluorescent product DCF, with excitation and emission wavelengths λex/λem = 492– 495/517–527 nm. The exposure time was kept below 4 s in order to avoid photo-oxidation of the ROS/RNS sensitive dye and for all given treatments fields were exposed for exactly the same amount of time. At least three independent fields were acquired for each condition and at least 10 cells per field were used for quantification of the fluorescence signal. Cells were masked by drawing a region of interest around the cell body, and mean fluorescence intensity was calculated for each cell after subtraction of the background signal using the image analysis module in ImageJ software. All cells, independent of ttheir relative intensity unit (RIU), were included in the analysis. Cultures were also incubated with H2O2 (200 μM for 20 min) to serve as a positive control.

#### Electron Microscopy

For electron microscopy analysis, ventral spinal cord cultures were processed as previously described (Villegas et al., 2014). Briefly, cells were fixed at 4◦C o.n. with ultrastructure solution (2.5% glutaraldehide; 0.1 M picric acid; 0.05 M cacodylate buffer; pH 7.4). Cultures were wash in the same buffer, immersed in 1% OsO4 for 1 h followed by a 2 h in block incubation with 2% uranyl acetate. Cultures were then dehydrated using a graded series of ethanol, propylene oxide and infiltrated with Epon (Ted Pella Inc.). Ultrathin sections were contrasted with 1% uranyl acetate and lead citrate. Grids were observed with a Philips Tecnai 12 electron microscope operated at 80 kV. Negative films were developed and scanned.

#### Western Blot

The spinal cord and brain from post-symptomatic hSOD1G93A and control WT mice were lysed in RIPA buffer (50 mM Tris-HCL pH 7,4; 1% NP-40; 0,5% Na-deoxycholate, 0,1% SDS, 150 mM NaCl and 2 mM EDTA) with protease inhibitor cocktail (Roche, Complete min tables Cat. No. 11836153001) and phosphatase inhibitors [10 mM sodium orthovanadate and 50 mM sodium fluoride (New England Biolabs, Ipswich, MA, USA)]. Protein concentrations were determined using micro BCA Protein Assay Kit (Thermo Scientific, Rockford, IL, USA).

Protein samples (30 μg per lane loaded) were separated by 10% SDS-polyacrilamide gel electrophoresis (PAGE) and then transferred to a nitrocellulose membrane (Thermo Scientific, Rockford, IL, USA). Primary antibodies were used at following dilutions: unphosphorylated anti-c-Abl, 1:1000 (Santa Cruz, USA Cat. No. sc-56887); anti-n-cadherin 1:1000 (Santa Cruz, USA, Cat. No. sc-271386). For detecting active c-Abl, rabbit polyclonal antibody recognizing phosphorylated Tyr-412 (1:2000; Sigma; Cat No C5240) were used. Secondary antibodies were used at 1:2000 and detection was performed using ECL Western Blotting Substrate (Thermo Scientific, Rockford, IL, USA). Anti-phosphoc-Abl was diluted in 5% fat-free milk in TBS plus 0.5% Tween-20 (Winkler, Chile), otherwise 5% fat-free milk in PBS plus 0.5% Tween-20 was used.

#### Data Analysis

ANOVA, followed by *post hoc* Tukey tests, was used to detect significant changes. Student's *t*-tests were used to compare the response of two cell populations to individual treatments. Unless otherwise stated, error bars represent the mean <sup>±</sup> SEM; <sup>∗</sup>*<sup>p</sup>* <sup>≤</sup> 0.05, ∗∗*<sup>p</sup>* <sup>≤</sup> 0.01, ∗∗∗*<sup>p</sup>* <sup>≤</sup> 0.001 vs. control; #*<sup>p</sup>* <sup>&</sup>lt; 0.05, ##*<sup>p</sup>* <sup>&</sup>lt; 0.01, ###*p* < 0.001 compared to ACMs.

#### Results

#### ACM-hSOD1G93A Rapidly Increases Phosphorylation of c-Abl in Motoneurons, Interneurons and Astrocytes

We first examined whether c-Abl is phosphorylated in the central nervous system of symptomatic mice that express hSOD1G93A. Immunostaining assays as well as western blots of whole lysate extracts from the spinal cord and brain were used to determine the expression of native c-Abl proteins and of c-Abl that is phosphorylated on tyrosine 412 (Tyr412; a site that enhances c-Abl catalytic activity; Hantschel and Superti-Furga, 2004). In agreement with a previous study (Katsumata et al., 2012), we found a robust increase in the phosphorylation of c-Abl in the brain and spinal cord of symptomatic hSOD1G93A mice (P120), compared to non-transgenic littermates; however, unlike this earlier study, our samples did not show an increase in the levels of expression of native c-Abl (Supplementary Figure S1). Interestingly, we also found similar increases in c-Abl phosphorylation levels in the spinal cord and motor cortex of symptomatic SOD1G86R mice (P95; Supplementary Figure S2).

To investigate whether soluble toxic factor(s) released by astrocytes that carry the hSOD1G93A mutation can trigger the activation of c-Abl, we exposed 4 DIV ventral spinal cord cultures to ACM-hSOD1G93A, fixed the cells at different post treatment times (0–120 min), and immunostained with an antibody against phosphorylated c-Abl (Tyr412; **Figure 1A**). To assess which cell type in our model system displays phosphorylated c-Abl, we used an antibody against microtubule-associated protein 2 (MAP2; recognizing interneurons and motoneurons), the SMI-32 antibody (recognizing motoneurons in ventral spinal cord cultures), and an antibody against the glial fibrillary acidic protein (GFAP; expressed in astrocytes). Analysis of the immunostaining

#### FIGURE 1 | Continued

Exposure of primary spinal cord cultures to ACM-SOD1G93A induces increases in c-Abl phosphorylation. (A) Flow diagram of experiment. Primary wild-type (WT) rat spinal cord cultures (4 DIV) were exposed to ACM derived from transgenic mice overexpressing SOD1G93A (ACM-SOD1G93A) for 0–120 min. Next cultures were washed, fixed and immunostained with an antibody recognizing phosphorylated c-Abl (Tyr-412) and cell markers to define cell type. (B) Spinal cord cultures untreated (0 min; upper images) or treated with ACM-SOD1G93A for 90 min (lower images) were triple-labeled with anti-microtubule-associated protein 2 (MAP2) antibody (white) to visualize neurons (arrow show interneuron that is MAP2+/SMI32−), with the SMI-32 antibody (red) to identify motoneurons (arrowhead), and with the phospho-c-Abl antibody (green) to show active c-Abl. Cultures were also stained with DAPI to visualize their nucleus. Inset shows selected motoneuron in the merge. Scale bar, 25 μm. (C) Spinal cord cultures untreated (0 min; upper images) or treated for 90 min with ACM-SOD1G93A (lower images) were triple-labeled with anti-MAP2 antibody (white) to visualize neurons (arrows), with the GFAP antibody (red) to identify astrocytes (arrowhead), and with the phospho-c-Abl antibody (green) to show active c-Abl. Inset shows selected astrocyte in the merge. Scale bar, 25 μm. (D) Graphs showing the c-Abl-P fluorescent intensity (relative intensity unity; RIU) at 0, 30, 60, 90, and 120 min after application of ACM-SOD1G93A (red line): immunestaining was used to identify c-Abl-P within a particular cell type, including in motoneurons (D1; MAP2+/SMI32+), interneurons (D2; MAP2+/SMI32−), and glial cells (D3; MAP2−/GFAP+). Results obtained ACM-SOD1WT are also included (gray lines). In all experiment, H2O2 (200 μM for 20 min) served as positive control. Note that c-Abl-P fluorescence peaked in all three cell types after 90 min of incubation with ACM-SOD1G93A. (E) Graphs showing the percentage of motoneurons (E1), interneurons (E2), and glial cells (E3) positive for c-Abl-P after 90 min (at peak) of exposure to ACM-SOD1G93A. Values represent mean <sup>±</sup> SEM from at least three independent experiments performed in duplicate, analyzed by ANOVA (D) or *t*-test (E). ∗*p* < 0.05, ∗∗*p* < 0.01, ∗∗∗*p* < 0.001 vs. control.

revealed that in ventral spinal cord cultures at 4 DIV, ∼4% of the cells are motoneurons (MAP2+/SMI32+/GFAP−), ∼37% are interneurons (MAP2+/SMI32−/GFAP−), and ∼59% are astrocytes (MAP2+/SMI32−/GFAP−; data not shown). Using this staining strategy to identify cell types, we show that acute application of ACM-hSOD1G93A to 4 DIV spinal cord cultures induces a robust phosphorylation of c-Abl in motoneurons (arrowheads and inset in **Figure 1B**) and interneurons (arrows in **Figure 1B**) and astrocytes (arrowheads and insets in **Figure 1C**). Quantification of fluorescence shows that intensity speaks for all three cell types at 90 min of ACM-hSOD1G93A exposure (red lines in graphs **Figures 1D1−3**). Strong c-Abl phosphorylation was also produced in spinal cord cultures by short-term application of H2O2 (200 μM for 20 min; green squares in **Figures 1D1−3**). Additional analysis of the number of cells that were immunopositive for phosphorylated c-Abl revealed that following 90 min of exposure to ACM-hSOD1G93A all motoneurons (100%; **Figure 1E1**), plus the vast majority of interneurons (>80%; **Figure 1E2**) and astrocytes (>90%; **Figure 1E3**) expressed the activated form of c-Abl. Analysis of the subcellular distribution of phosphorylated c-Abl further revealed that ACM-hSOD1G93A leads to a homogenous distribution of this activated tyrosine kinase in all three cell types (**Figures 1C–E1**; insets). Of note is that increases in activated c-Abl also occurs in the cell's nucleus where it can perform its pro-apoptotic function and repress neuronal genes expression (Zhu and Wang, 2004; Yoshida, 2008; Gonzalez-Zuñiga et al., 2014). Interestingly, we also detected similar increases in c-Abl phosphorylation levels in motoneurons that were subjected to ACM derived from astrocytes expressing SOD1G86R or TDP43A315T (Supplementary Figure S3). In contrast, control ACM obtained from astrocytes that are harvested from transgenic mice carrying the nonpathological WT human SOD1 gene (ACM-hSOD1WT) did not induce c-Abl phosphorylation in any of the three cell types (**Figures 1D1−3**; gray lines).

#### Treatment of Cultures with c-Abl Inhibitor STI571 Prevents Motoneuron Cell Death Induced by ACM-hSOD1G93A

To determine whether the increased phosphorylation of c-Abl induced by ACM-hSOD1G93A contributes to motoneuron death, 4 DIV spinal cord cultures were co-incubated with ACM-hSOD1G93A plus the c-Abl kinase inhibitor STI571, and motoneuron survival (MAP2+/SMI32+) was assessed at 7 DIV (**Figure 2A**). We first used immunostaining (as in **Figure 1**) to assess the effectiveness of STI571 in preventing c-Abl phosphorylation in the three spinal cord cell types. STI571, also called imatinib, when applied at micromolar levels to primary neuronal cultures effectively inhibits the phosphorylation of c-Abl following various stimuli, including oxidative stress (Alvarez et al., 2004, 2008; Cancino et al., 2011; Klein et al., 2011). We found that 4 DIV cultures treated for 90 min with both ACM-hSOD1G93A and 2 μM STI571 displayed a significant reduction in the intensity of immunoreactivity for phosphorylated c-Abl (**Figures 2B1−3**) and in the percentage of cells that were positively stained (Supplementary Figures S4B1−3); these changes were observed for motoneurons (**Figure 2B1**), interneurons (**Figure 2B2**), and astrocytes (**Figure 2B3**). Following chronic treatment of the cultures with ACM-hSOD1G93A plus STI571, and analyzed 3 days later (at 7 DIV), we found that the reduction in intensity (**Figures 2C1−3**) and in the percentage of cells positive for phosphorylated c-Abl (Supplementary Figures S4C**1−3**) was maintained in all three cell types. Importantly, application of STI571 was able to prevent motoneuron cell death induced by SOD1G93A (**Figure 2D**). Similarly, STI571 prevented also motoneuron cell death induced by SOD1G86R or TDP43A315T (Supplementary Figures S3C–E). These results document that activation of c-Abl in two completely unrelated ALS models leads to the death of motoneurons. In contrast to the ACMs obtained from ALS astrocytes, chronic application of STI571 with control ACM-hSOD1WT did not alter c-Abl intensity in motoneurons (**Figure 2C4**) or number of surviving motoneurons (**Figure 2E**). Collectively, these results indicate that the favorable effects of STI571 are not a result of a non-specific beneficial influence of this compound, but rather that the inhibitor drug counterbalances the toxic effects of ACM-hSOD1G93A. Because ACM-hSOD1WT did not induce a significant change in the phosphorylation of c-Abl, in motoneuron cell survival, or in ROS/RNS production (as also shown previously; Fritz et al., 2013; Rojas et al., 2014), results from use of the control ACM are omitted from most figures.

ACM-SOD1G93A was applied to 4 DIV spinal cord cultures acutely (for 90 min when c-Abl-P peaks; see Figure 1) or chronically (3 days) either alone or in the presence of c-Abl kinase inhibitor STI571 (1 μM). Phosphorylation of c-Abl was measured at 4 and 7 DIV. Cell survival was measured at 7 DIV. (B) Graphs showing fluorescence intensities (RIU) for c-Abl-P at 4 DIV when treated acutely (90 min) with ACM-SOD1G93A alone or ACM-SOD1G93A plus STI571; motoneurons (B1), interneurons (B2) and glial cells (B3) were identified by immunostaining (as in Figure 1). (C1**−**3) Same as in (B), but c-Abl-P is

(3 days) with ACM-SOD1WT alone or with STI571. (D,E) Graphs showing the relative percentage of motoneurons that survived at 7 DIV, after being treated with STI571 and ACM-SOD1G93A (D) or ACM-SOD1WT (E). Values represent mean ± SEM from at least three independent experiments performed in duplicate, analyzed by One-Way ANOVA followed by a Tukey *post hoc* test. ∗*p* < 0.05, ∗∗*p* < 0.01, ∗∗∗*p* < 0.001 relative to control conditions. #*p* < 0.05, ##*p* < 0.01 compared to survival with the ALS-causing ACM to at 7 DIV without STI571.

#### Nitroxidative Stress Induced by ACM-hSOD1G93A Leads to Activation of c-Abl

Next we investigated possible mechanisms underlying the induction of c-Abl activation by ACM-hSOD1G93A. Because c-Abl is activated by oxidative stress and induces cell death (van Etten, 1999; Klein et al., 2011; Schlatterer et al., 2011b), we focused on the role of ROS/RNS. Indeed, we find that H2O2– induced oxidative stress caused a strong phosphorylation of c-Abl in motoneurons, interneurons and astrocytes (see green squares in **Figure 1D1−3**). Conversely, we wanted to determine whether co-application of antioxidants with the toxic ACM-hSOD1G93A can prevent c-Abl activation and motoneuron cell death. Keeping in mind that antioxidants can be used as a therapeutic strategy for ALS, we tested several different types of compounds with antioxidant and/or free scavenger capacities, that also have the biochemical properties to effectively penetrate the blood– brain-barrier: these include Trolox (a H2O soluble vitamin E analog), esculetin, and 4,5-dihydroxy-1,3-benzenedisulfonic (tiron). Trolox and esculetin were selected because they were identified in a large screening assay as having strong antioxidant activity, and as protecting the viability of mutant SOD1G93Aexpressing cell lines under stress conditions (Barber and Shaw, 2010). In a previous study, we tested multiple doses (ranging from 100 to 100 μM), and found that 1 μM Trolox and 25 μM esculetin reduced intracellular levels of ROS/RNS, and largely prevented primary motoneuron cell death induced by ACMhSOD1G93A (**Figure 3D**), without affecting the survival of control motoneurons (Rojas et al., 2014; and shown in **Figure 3E**). We also tested tiron, because it permeabilizes the mitochondrial membrane and is more effective in reducing oxidative stress compared to the mitochondria-targeted antioxidant MitoQ (Oyewole et al., 2014). We assessed the antioxidant capacity and toxicity of multiple doses (ranging from 1 μM to 1 mM) of this antioxidant (not shown), and found that 25 μM tiron strongly reduces H2O2–induced oxidative stress (Supplementary Figure S5) in the absence of cell death of control motoneurons (**Figure 3E**).

Notably, all three antioxidants effectively abrogated the phosphorylation of c-Abl in motoneurons, interneurons and astrocytes that is induced by acute (90 min exposure and tested at 4 DIV; **Figures 3B1−3**) or chronic (3 days exposure and tested at 7 DIV; **Figures 3C1−3**) treatment with ACM-hSOD1G93A (**Figures 3B,C** show c-Abl-P intensities, while the percentage of cells positive for c-Abl-P is shown in Supplementary Figure S6). Importantly, motoneuron death was also largely prevented following chronic co-application of ACM-hSOD1G93A and Trolox, esculetin, or tiron (**Figure 3D**), whereas no increase in the number of motoneurons was observed under control conditions with use of any of these antioxidants (**Figure 3E**). Together, our data indicate that counterbalancing the production of ACM-hSOD1G93A-induced ROS/RNS protects motoneurons from death by diminishing, at least in part, c-Abl activation.

#### ACM-hSOD1G93A Leads to Mitochondrial Swelling and Membrane Depolarization

Because dysfunctional mitochondria constitute a major source for elevating ROS production, and because mitochondrial damage is a common feature in ALS patients and in mouse models of ALS (Manfredi and Xu, 2005; Jaiswal et al., 2009; Grosskreutz et al., 2010; Tan et al., 2014), we wanted to know whether toxic factors released by hSOD1G93A-expressing astrocytes induce mitochondrial defects in the neurons of our culture system. We first analyzed the morphology and membrane potential of mitochondria in 4–5 DIV spinal cord neurons under control conditions. To fluorescently label mitochondria, we used Tetramethylrhodamine Methyl Ester, Perchlorate (TMRM; **Figure 4**); live time-lapse fluorescence microcopy imaging was accompanied by phase-contrast images to select motoneurons on the basis of their large soma (>20 μm) and extent at least five primary dendrites. The TMRM fluorescent labeling methods indicate that under basal conditions, mitochondria in interneurons and motoneurons of 4–5 DIV spinal cord cultures are mainly organized into a tubular network (especially in the soma), with a few isolated round mitochondria also visible (**Figure 4B**). To get direct evidence for this, we visualized the ultra-structure of mitochondria using transmission electron microscopy (**Figure 5**); it was revealed that neurons in control 4–5 DIV spinal cord cultures display prominent mitochondria, most with an elongated shape and fewer with a globular morphology (see Materials and Methods for identifying neuronal vs. glial mitochondria).

We also studied the membrane potential and morphology of mitochondria of 4–5 DIV spinal cord neurons that were exposed to ACM-hSOD1G93A for different times (minutes to hours). TMRM staining showed that already short-term exposure (30 min) of spinal cord neurons to ACM-hSOD1G93A induces a decrease in the fluorescent labeling in the majority (but not all) of the neurons (**Figure 4B**). Quantification reveals that ∼60% of the neurons displayed a robust reduction in the levels of TMRM fluorescence (see **Figures 4C,D** for scatter plots and graph with average data, respectively); it is not clear why the TMRM signal in another population of neurons (∼40%) remained unchanged. Because the trapping of the TMRM dye inside the mitochondria matrix depends on the mitochondrial membrane potential (m), the loss of staining could be a consequence of mitochondrial membrane depolarization, which can be induced by activation of uncoupling proteins, or by the opening of the mitochondrial permeability transition pore (mPTP); alternatively, it could be the result of mitochondria degeneration. The results of two independent sets of experiments support a depolarization event: (i) use of mitoDsRed2 (a marker of the mitochondrial matrix used to assess mitochondrial morphology and that is insensitive to m) resulted in no obvious reduction in the number or intensity of labeled mitochondria in cells treated with ACM-hSOD1G93A for 0–90 min (not shown), indicating that mitochondria were not lost; and (ii) ultra-structural analysis did not reveal any reduction in the number of mitochondria in neurons that had been exposed to ACM-hSOD1G93A for 4 h (**Figure 5**). Electron microscopy images, however, did show that exposure to ACM-hSOD1G93A induces mitochondrial alterations within neurons with many mitochondria displaying a subtle, but significant, more swollen morphology (**Figures 5A,B**); this

#### FIGURE 3 | Continued

or tiron (25 μM) Phosphorylation of c-Abl was measured at 4 and 7 DIV. Cell survival was measured at 7 DIV. (B) Graphs showing fluorescence intensities (RIU) for c-Abl-P at 4 DIV when treated acutely (90 min) with ACM-SOD1G93A alone or ACM-SOD1G93A plus the antioxidants; motoneurons (B1), interneurons (B2) and glial cells (B3) were identified by immunostaining (as in Figure 1). (C) Same as in (B), but c-Abl-P is measured at 7 DIV when treated chronically (3 days) with ACM-SOD1G93A alone or in the presence of the antioxidants. (D,E) Graphs showing the relative percentage of motoneurons that survived at 7 DIV, after being treated with antioxidants and ACM-SOD1G93A (D) or ACM-SOD1WT (E). Values represent mean <sup>±</sup> SEM from at least 3 independent experiments performed in duplicate, analyzed by One-Way ANOVA followed by a Tukey *post hoc* test. ∗*p* < 0.05, ∗∗*p* < 0.01, ∗∗∗*p* < 0.001 relative to control conditions. #*p* < 0.05, ##*p* < 0.01, ###*p* < 0.001 compared to survival with the ALS-causing ACM to at 7 DIV without antioxidants.

kind of morphology is usually associated with the opening of the mPTP (Ruiz et al., 2014). These observations collectively indicate that ACM-hSOD1G93A induces rapid morphological and physiological alterations in mitochondria. Because increased action potential firing is one the first pathological events detected in our *in vitro* model system (Fritz et al., 2013; Rojas et al., 2014), we tested whether reducing hyper-excitability with the Nav channel blocker spermidine (10 μM) is capable of preventing the morphological alterations of neuronal mitochondria induced by ACM-hSOD1G93A. Interestingly, electron microscopy showed that co-application of ACMhSOD1G93A and spermidine for 4 h abrogates alterations in the morphology of neuronal mitochondria (**Figures 5A,B**), without changing the number of mitochondria (**Figure 5C**). We also found that chronic exposure of ACM-hSOD1G93A plus spermidine also prevents mitochondrial swelling (Supplementary Figure S7).

#### Order of Pathological Events, and Interplay between ROS and Mitochondria

Our data strongly indicate that upon ACM-hSOD1G93A exposure the neuron's hyper-excitability is the first alteration to occur (at 15–30 min), whereas activation of c-Abl is the last event (at 60–90 min); by contrast, calcium influx, ROS generation and mitochondrial swelling and membrane depolarization are all intermediate events, detectable starting at 30 min after ACM-hSOD1G93A exposure. To determine the exact order of these pathological events and mechanisms whereby c-Abl activation mediates pathogenesis and death of motoneurons, we studied the interplay between mitochondrial structure and membrane physiology, oxidative stress, and c-Abl activation. For this, ROS/RNS production (**Figure 6**) and c-Abl activation (**Figure 7**) were measured in cultures that were exposed to ACM-hSOD1G93A alone or in the presence of multiple pharmacological compounds. We used blockers of Nav channels (25 nM mexiletine, 100 nM riluzole and 10 μM spermidine; as in Fritz et al., 2013; Rojas et al., 2014); chelation of extracellular calcium (200 μM EGTA) to prevent calcium influx; and antioxidants (1 <sup>μ</sup>M Trolox; as in **Figure 3**). To intent to prevent mitochondrial alterations, we used Ru360, a selective inhibitor of calcium uptake in mitochondria (Matlib et al., 1998), as well as cyclosporin A (CsA) to inhibit the mPTP (Crompton et al., 1988, 1998; Broekemeier et al., 1989). For ROS/RNS production, we measured intracellular DCF fluorescence in spinal cord cultures at 60 min of exposure to ACM-hSOD1G93A (**Figure 6A**), at the peak of DCF fluorescence levels as we have shown previously (Rojas et al., 2014). In agreement with our previous work (and repeated here for clarity), we found that co-incubation of ACMhSOD1G93A with diverse the Nav channel blockers mexiletine, riluzole or spermidine effectively inhibits the increase in DCF fluorescence levels induced by the toxic ACM (**Figure 6B**). In addition, we found that the calcium chelator EGTA resulted in lower intracellular DCF fluorescence levels to degrees similar to that observed in untreated cultures (**Figure 6C**). Co-application of ACM-hSOD1G93A with mitochondrial protectors also strongly reduced DCF fluorescence; cyclosporin A being more effective than Ru360 (**Figure 6D**). These results indicate that ROS/RNS is at least in part released by mitochondria when neurons are exposed to ACM-hSOD1G93A. We used several control, including ACM-hSOD1G93A plus Trolox which shows, as expected, that ROS/RNS production was completely prevented (**Figure 6E**). We also found that incubation of spinal cord cultures with ACMhSOD1WT alone or together with EGTA, cyclosporin A or Ru360 did not affect DCF fluorescence levels (Supplementary Figures S8A–C); the finding that co-incubation of ACM-hSOD1WT with the Nav channel blockers mexiletine, riluzole or spermidine also does not alter DCF fluorescence levels was shown previously (Rojas et al., 2014).

To uncover the upstream events that contribute to c-Abl activation in motoneurons and are mediated by ACMhSOD1G93A, we co-applied ACM from astrocytes expressing hSOD1G93A plus the different compounds (**Figure 7A**); phosphorylation of c-Abl was measured at 90 min (i.e., at the peak of c-Abl phosphorylation as shown in **Figure 1D**1). In **Figure 3** we already showed that application of the antioxidants Trolox, esculetin or tiron prevent the increase in phosphorylation of c-Abl. We also tested the effects of Nav channel blockers on ACM-induced c-Abl activation, and find that co-application of ACM-hSOD1G93A plus mexiletine, spermidine or riluzole to spinal cord cultures partly prevents c-Abl phosphorylation in motoneurons (**Figure 7B**). Finally, we find that while the calcium chelator EGTA slightly reduced ROS/RNS production (**Figure 7C**), the mitochondrial protectors Ru360 or cyclosporin A (**Figure 7D**) strongly prevent the c-Abl phosphorylation induced by ACM-hSOD1G93A. Incubation of spinal cord cultures with ACM-hSOD1WT alone or together with any of the above indicated compounds did not alter c-Abl phosphorylation levels (Supplementary Figures S8D–H). Collectively, our data indicate that application of ACM-hSOD1G93A leads to the Nav-channelmediated hyper-excitability and calcium influx that trigger mitochondrial swelling and membrane depolarization; the impaired mitochondria contribute, at least in part, to ROS/RNS production that leads to activation of a lethal c-Abl signaling cascade.

FIGURE 4 | Mitochondrial potential membrane change after ACM-hSOD1G93A. (A) Flow diagram of experiment. Four DIV spinal cord cultures were incubated with TMRM (50 nM) during 30 min and then exposed to ACM-hSOD1G93A for 30 min. Pictures were acquired using a combination of real-time fluorescence and phase-contrast imaging. (B) Representative pictures of an untreated neuron (0 min; upper images) or treated for 30 min with ACM-SOD1G93A (lower images). Fluorescence images, brightfield images and their merges are also shown. Scale bar in main pictures 10 and 2 μm in insets. (C) Scatter plot showing the percentage of TMRM fluorescence (arbitrary units) in every individual neuron

### Discussion

#### Deciphering the Sequence of Pathological Events in Motoneurons with Use of the ACM-hSOD1G93A Model System

Amyotrophic lateral sclerosis is, at least in part, a non-cellautonomous disease in which astrocytes that express ALScausing genes contribute to disease pathogenesis (Clement et al., measured under the conditions indicated. Note that the majority (63%) of the ACM-hSOD1G93A-treated neurons show a decrease in TMRM fluorescence relative to the control condition, but that another neuronal population (37%) does not. For clarity, we show all neurons (response + no response), those that show reduced TMRM fluorescence after ACM-hSOD1G93A exposure (response) and those that did not displayed TMRM alterations (no response). (D) Graph showing the same data as in C. Values represent mean ± SEM from at least three independent experiments performed in duplicate, analyzed by One-Way ANOVA followed by a Tukey *post hoc* test. ∗∗∗*p* < 0.001 relative to control conditions.

2003; Ilieva et al., 2009). Compelling evidence documents that primary astrocytes that express mutant SOD1-expressing and are derived from mice (Nagai et al., 2007; Di Giorgio et al., 2007; Castillo et al., 2013; Fritz et al., 2013; Rojas et al., 2014), rats (Vargas et al., 2006; Cassina et al., 2008), and humans (Marchetto et al., 2008) selectively kill motoneurons, but spare interneurons. Remarkably, astrocytes that are differentiated from neuronal progenitor cells (NPCs) derived either from post-mortem spinal

cord tissue or skin biopsies from FALS (SOD1 and C9orf72) and sALS patients, also display non-cell-autonomous toxicity, and selectively kill motoneurons in a co-culture model system (Haidet-Phillips et al., 2011; Meyer et al., 2014; Re et al., 2014). Several pathological phenotypes observed in ALS patients and animal models are recapitulated in these *in vitro* neuronastrocyte co-cultures, including those that are based on hyperexcitability (Fritz et al., 2013; Rojas et al., 2014), mitochondrial dysfunction (Cassina et al., 2008), and nitroxidative stress (Cassina et al., 2008; Marchetto et al., 2008; Rojas et al., 2014). However, the sequence of pathological events involved in these changes has not been systemically studied. We believe that unraveling the sequence has been difficult because the co-cultures used to study neuron-astrocyte interactions employ neurons that grow on a feeder layer of astrocytes, hereby precluding assessment of the temporal interplay between pathogenic events (Vargas et al., 2006; Di Giorgio et al., 2007; Nagai et al., 2007; Cassina et al., 2008; Marchetto et al., 2008; Haidet-Phillips et al.,

2011; Meyer et al., 2014; Re et al., 2014). Nagai et al. (2007) establishes a unique *in vitro* ALS model system, and show that exposure of spinal cultures to conditioned media obtained from astrocytes that express mutated human SOD1 kills primary motoneurons, as well as motoneurons derived from embryonic mouse stem cells. We used (with some adaptions; see Materials and Methods) this *in vitro* model system (Nagai et al., 2007), along with pharmacological interventions and time-dependent measurements at the single-cell level, to systematically unravel the types and temporal order of pathological events induced in motoneurons by ALS-ACM. In addition, the highly diluted ACM-SOD1G93A that we use (see Materials and Methods) enables us to show that motoneuron death is triggered by specific soluble toxic mediator(s), and not by the lack of survival factors. Making use of electrophysiological recordings, calcium imaging and fluorescent probes to detect ROS/RNS, and immunostaining to identify surviving interneurons and motoneurons, we had previously documented that exposure of primary WT spinal cord

cultures to ACM-hSOD1G93A leads to a rapid increase in the excitability of neurons (detected already at 15–20 min) and to influx of calcium, which in turn generates intracellular ROS/RNS followed by specific and robust motoneuron death (∼50%) within a matter of days (Fritz et al., 2013; Rojas et al., 2014). In the current study, we have studied the mechanisms that underlie astrocyte-mediated toxicity of motoneurons in more detail, with use of a variety of methods, including ROS/RNS detection with the DCF fluorescent probe; immunostaining to detect c-Abl phosphorylation and the ratio of interneurons:motoneurons; and two different fluorescent probes (TMRM and mitoDsRed2) plus electron microscopy to determine mitochondrial function and structure, respectively. We provide evidence here that toxic factors released by hSOD1G93A–expressing astrocytes first increase Nav-channel-mediated excitability in neurons, which in turn increases calcium influx, and triggers functional and

Frontiers in Cellular Neuroscience | www.frontiersin.org June 2015 | Volume 9 | Article 203 |

structural changes in mitochondria. Our data indicate that ROS/RNS, generated at least in part through mitochondrial alterations, activate c-Abl signaling. Although mitochondria are a main source of ROS production for most cells, it has been well established that ROS can be produced at multiple sites in mammalian cells, including by NADPH oxidase (Nox), amino acid oxidases, cytochrome P450 enzymes, cycloxygenases, lipoxygenases and xanthine oxidase (Turrens, 2003). Of interest is also that ROS produced by these different sources does not always activate pathophysiological cellular processes, but can also function as a beneficial signaling messenger (Massaad and Klann, 2011). In ALS pathology, toxic ROS can be produced by dysfunctional mitochondria (Tan et al., 2014; and see below), as well as through Nox enzymes (Wu et al., 2006; Marden et al., 2007; Harraz et al., 2008). Additional studies, which are out of the scope of this present study, should be performed to identify which other source(s), in addition to mitochondria, could be involved in the production of ROS in the ACM-hSOD1G93A model system. Our findings collectively lead us to conclude that the key hallmarks of ALS, including ROS/RNS production and mitochondrial swelling and membrane depolarization, are recapitulated in our *in vitro* neuron-astrocyte model. We predict that our focus on elucidating the mechanisms of non-cellautonomous motoneuron pathology and death, and on the contribution of Nav-channels, ROS/RNS, mitochondria and the c-Abl pathway, will yield advances for the use and/or development of therapeutic interventions for this devastating disease.

#### Dysfuntional Mitochndria as a Source of ROS/RNS

Mitochondria participate in energy metabolism, intracellular calcium homeostasis, production of intracellular ROS, and regulation of apoptosis (Nicholls, 2009; Rasola and Bernardi, 2011; Maryanovich and Gross, 2013). Morphological and functional abnormalities in mitochondria are a common feature of ALS, and are detected in biopsied and post-mortem tissue from symptomatic sALS and fALS patients, as well as in mutant SOD1 mouse models and the cell cultures derived from these mice (reviewed in Manfredi and Xu, 2005; Bento-Abreu et al., 2010; Grosskreutz et al., 2010; Tan et al., 2014). Degenerated, vacuolized and swollen mitochondria are detected in transgenic mice that express mutations in SOD1, around the time when clinical symptoms appear (Dal Canto and Gurney, 1995; Wong et al., 1995; Kong and Xu, 1998). Moreover, ultra-structural analysis reveals that mitochondrial swelling is detected in spinal cord motoneurons of hSOD1G93A transgenic mice at P14, 2–3 months *prior to* the time that motoneurons degeneration is visible, and before clinical symptoms are manifest (Bendotti et al., 2001); these findings underscore early mitochondrial abnormalities in this mouse model and suggest that mitochondrial dysfunction might contribute to motoneuron pathology. Electron microscopy also shows that in our *in vitro* ALS model system mitochondrial swelling is an early event in motoneuron pathogenesis, and is induced by soluble mediator(s) secreted by hSOD1G93Aexpressing astrocytes. Swelling of mitochondria may be caused by activation of mPTP, a voltage-dependent, high conductance mega channel whose opening is critically regulated and triggered by various processes, including matrix calcium accumulation, ROS, and adenine nucleotide depletion (Miller, 1998; Kroemer and Reed, 2000; Rasola and Bernardi, 2011; Bernardi and Di Lisa, 2015).

Our findings with cyclosporine A and Ru360, along with results from use of mitochondrial fluorescent probes, suggest that hSOD1G93A astrocyte-triggered abnormalities are caused, at least in part, by the accumulation of calcium in mitochondria; the calcium uptake could induce the opening of mPTP in motoneurons which is followed by a rapid loss of m, uncontrolled matrix calcium and oxidative species release, matrix swelling and eventually rupture of the outer membrane, and release of apoptogenic factors (Rasola and Bernardi, 2011). Thus the opening of mPTP may induce both apoptotic and necrotic death signal

Because mitochondria predominantly generate superoxide anion (O2 •−) as a by-product of the respiratory chain functioning (Murphy, 2009), which is not detectable by the DCF probe used in this study (Myhre et al., 2003), our findings indicate that other reactive oxygen and nitrogen species contribute to the motoneuron pathology observed in our model system. On the other hand, DCF does detect hydrogen peroxide (H2O2; in combination with cellular peroxidases), hydroxyl radicals (•OH), and peroxynitrite (ONOO−; Estévez et al., 1999; Myhre et al., 2003; Gomes et al., 2005; Martin et al., 2007; Kalyanaraman et al., 2012). Of particular interest is ONOO− which is formed from •NO and O2 •− and is strongly implicated in several models of ALS, but is predominantly detected by the amounts of 3-nitrotyrosine formation in proteins (Abe et al., 1997; Bruijn et al., 1997; Beckman et al., 2001; Radi, 2004; Wu et al., 2006). Reactive and SOD1G93A-expressing astrocytes are a major source of •NO, which then can target neighboring cells and affects the survival of WT motoneurons (Vargas et al., 2006; Cassina et al., 2008). Based on these reports, we hypothesize that conditioned media derived from hSOD1G93Aexpressing astrocytes includes •NO and/or leads to the generation of •NO in the WT astrocytes that are present within spinal cord cultures upon acute exposure of the toxic ACM; this •NO in turn permeates into neurons, where it interacts with the intracellular O2 •− produced by mitochondria, and forms ONOO−.

#### Activation of c-ABL in ALS

Active c-Abl is implicated in a variety of neurodegenerative diseases, including Alzheimer's (Alvarez et al., 2004; Cancino et al., 2008; Estrada et al., 2011; Gonzalez-Zuñiga et al., 2014) and Parkinson's diseases (Imam et al., 2011). Recently, Katsumata et al. (2012) reported a significant increase in c-Abl expression in post-mortem spinal cord tissues from sALS patients, and in the spinal cord of symptomatic hSOD1G93A mice; they also demonstrated that oral administration of the c-Abl inhibitor dasatinib to animals aged P56–P120 attenuates motoneuron loss, alleviates motor dysfunction, and significantly increases lifespan (Katsumata et al., 2012). In the current study, we also document that phosphorylation of c-Abl is significantly increased in the brain and spinal cord of symptomatic hSOD1G93A mice; furthermore, we show that activated c-Abl is detectable in symptomatic hSOD1G86R mice. Importantly, exposure of WT spinal cord cultures to conditioned media derived from astrocytes expressing SOD1 (SOD1G93A or SOD1G86R) TDP43 (TDP43A315T) mutants, activates the c-Abl tyrosine kinase in WT motoneurons. Our results show that activation of c-Abl in two completely unrelated ALS models leads to the death of motoneurons via non-cell autonomous processes. We used our *in vitro* ACM-hSOD1G93A model system to investigate the potential causes of c-Abl activation in ALS: in agreement with earlier studies, which report that oxidative stress can induce c-Abl activation (van Etten, 1999; Klein et al., 2011; Schlatterer et al., 2011a), we found that ROS (H2O2) increases c-Abl phosphorylation, and that co-application of antioxidants plus ACM-hSOD1G93A strongly reduce activation of this kinase, in a manner similar to that of the c-Abl inhibitor STI571. Moreover, we also show that reducing c-Abl activation by antioxidants as wells as by STI571 protects astrocyte-mediated Rojas et al. Role for c-Abl in ALS

toxicity in motoneurons. Experiments with use of mitochondrial protectors (cyclosporine A and Ru360) point to a key role for mitochondrial-produced ROS in the activation of c-Abl. Based on these findings, we hypothesize that inhibition of ROS production (by protecting mitochondrial dysfunction or by application of antioxidants) in ALS animals might largely prevents motoneuron pathology and significantly extends the lifespan of these mice.

#### Hyper-Excitability is a Primary Pathological Event in Astrocyte-Triggered Motoneuron Degeneration

Our current and earlier findings with use of the ACM *in vitro* model (Fritz et al., 2013; Rojas et al., 2014) strongly implicate hyper-excitability as a critical pathological event in astrocytetriggered motoneuron degeneration, and show that it occurs upstream of mitochondrial impairment, ROS production, and c-Abl activation. The use of electrophysiological recordings in motoneurons in acute slice preparations of neonatal SOD1G93A mice (P4–P10), reveals that increased excitability (van Zundert et al., 2008) precedes the mitochondrial damage that is detected at 2 weeks of postnatal life (Bendotti et al., 2001). Additional studies also show that electrophysiological abnormalities in motoneurons are the earliest physiological alterations detected in diverse ALS rodent models (Kuo et al., 2005; Bories et al., 2007; Pambo-Pambo et al., 2009; Quinlan et al., 2011; van Zundert et al., 2012). Hyper-excitability has been reported in fALS and sALS patients (van Zundert et al., 2012), and also in induced pluripotent stem cell-derived motoneurons (i-motoneurons) generated from ALS patients that harbor mutations in SOD1, C9orf72, and FUS/TLS (Wainger et al., 2014). Generation of action potentials, and hence of membrane excitability, is regulated by the number and functioning of Nav and Kv channels on neuronal cells. Currently, we have little insight into the mechanism responsible for hyper-excitability, but we do know that it is related to an increase in persistent Nav channel currents (van Zundert et al., 2008, 2012; Fritz et al., 2013; Pieri et al., 2013), and to a reduction in the amplitude of the delayed-rectifier K channel current (Wainger et al., 2014). The possibility that a factor(s) released by ALS astrocytes regulates the function and/or expression of these ion channels is intriguing, and its identification is of great importance. Independent of the mechanisms, pharmacological assays *in vitro* show that decreasing the activity of Nav channels of motoneurons by co-application of ACM-hSOD1G93A plus the Nav channel blockers mexiletine, riluzole or spermidine (Fritz et al., 2013; Rojas et al., 2014; current study), or by increasing Kv channel activity via treatment of ALS i-motoneurons with retigabine (Wainger et al., 2014), inhibits hyper-excitability and improves motoneuron survival. These compounds may also be promising candidate drugs for attenuating pathogenesis and

### References

Abe, K., Pan, L. H., Watanabe, M., Konno, H., Kato, T., and Itoyama, Y. (1997). Upregulation of protein-tyrosine nitration in the anterior horn cells of amyotrophic lateral sclerosis. *Neurol. Res.* 19, 124–128.

delaying the onset of disease-specific symptoms of fALS and sALS patients; neurophysiological studies in both types of patients have uncovered neuronal hyper-excitability (van Zundert et al., 2012). In addition, because these patients are similar to each other in terms of other pathological events (e.g., mitochondrial impairment and production of ROS), clinical symptoms, and the benefits gained from treatment with riluzole (a non-specific Nav channel blocker; Bellingham, 2011), we hypothesize that motoneuron hyper-excitability represents a general feature of sALS and fALS patients, and that motoneuron death is triggered by activation of a common fatal pathogenic signaling pathway (van Zundert et al., 2012). Further support for this hypothesis will come from results of phase II trial that have been initiated by the ALS Therapy Alliance (ATA) and the Northeast ALS (NEALS) Consortium with use of mexiletine to treat sALS patients. Our findings also underscore the need for further tests in diverse ALS animal models and in ALS patients with the compounds that successfully prevent motoneuron damage in ALS neuronastrocyte *in vitro* systems: within this context, the anti-oxidants Trolox and esculetin are especially promising because of their strong antioxidant activity and their ability to cross the bloodbrain-barrier (Barber and Shaw, 2010).

## Author Contributions

FR and SA performed MN survival experiments. NC executed DCF measurement. DG, FR, EA, AM. performed c-Abl immunostainings *in vitro* and *in vivo*. EF and FC performed EM studies. NC, DH, EA executed TMRM and DsRed2 assays. DG and AM performed western blot assays. All analyzed the data and wrote the manuscript.

## Acknowledgments

We thank Luis Melo for technical support and Dimarco/Gene X-Press for their continued contribution to facilitate our science. This work was supported by DRI USA 2013-0030, ALS Therapy Alliance-CVS Pharmacy, Anillo-RING ACT1114, Fondecyt 1140301 and Nucleus DI-603-14/N to BvZ. Cochilco-Fondecyt 1100995, Millennium Institute of Immunology and Immunotherapy P09/016-F, and Nucleus UNAB DI-209-12/N to AE.

## Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2015.00203/abstract

Alvarez, A. R., Klein, A., Castro, J., Cancino, G. I., Amigo, J., Mosqueira, M., et al. (2008). Imatinib therapy blocks cerebellar apoptosis and improves neurological symptoms in a mouse model of Niemann-Pick type C disease. *FASEB J.* 22, 3617–3627. doi: 10.1096/fj.07- 102715


Hantschel, O., and Superti-Furga, G. (2004). Regulation of the c-Abl and Bcr-Abl tyrosine kinases. *Nat. Rev. Mol. Cell Biol.* 5, 33–44. doi: 10.1038/nrm1280


neurons derived from human embryonic stem cells. *Cell Stem Cell* 3, 649–657. doi: 10.1016/j.stem.2008.10.001


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Rojas, Gonzalez, Cortes, Ampuero, Hernández, Fritz, Abarzua, Martinez, Elorza, Alvarez, Court and van Zundert. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Homocysteine aggravates ROS-induced depression of transmitter release from motor nerve terminals: potential mechanism of peripheral impairment in motor neuron diseases associated with hyperhomocysteinemia

Ellya Bukharaeva1,2,3 , Anastasia Shakirzyanova1,2 , Venera Khuzakhmetova1,3 , Guzel Sitdikova<sup>2</sup> and Rashid Giniatullin1,2 \*

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Dietmar Fuchs, Innsbruck Medical University, Austria Hai Huang, Tulane University, USA Richard Anthony DeFazio, University of Michigan, USA

\*Correspondence:

Rashid Giniatullin, A. I. Virtanen Institute for Molecular Sciences, Department of Neurobiology, University of Eastern Finland, P.O. Box 1627/Neulaniementie 2, 70211 Kuopio, Finland rashid.giniatullin@uef.fi

Received: 09 July 2015 Accepted: 18 September 2015 Published: 06 October 2015

#### Citation:

Bukharaeva E, Shakirzyanova A, Khuzakhmetova V, Sitdikova G and Giniatullin R (2015) Homocysteine aggravates ROS-induced depression of transmitter release from motor nerve terminals: potential mechanism of peripheral impairment in motor neuron diseases associated with hyperhomocysteinemia. Front. Cell. Neurosci. 9:391. doi: 10.3389/fncel.2015.00391

<sup>1</sup> Department of Physiology, Kazan Federal University, Kazan, Russia, <sup>2</sup> A. I. Virtanen Institute for Molecular Sciences, Department of Neurobiology, University of Eastern Finland, Kuopio, Finland, <sup>3</sup> Kazan Institute of Biochemistry and Biophysics, Kazan, Russia Edited by:

> Homocysteine (HCY) is a pro-inflammatory sulphur-containing redox active endogenous amino acid, which concentration increases in neurodegenerative disorders including amyotrophic lateral sclerosis (ALS). A widely held view suggests that HCY could contribute to neurodegeneration via promotion of oxidative stress. However, the action of HCY on motor nerve terminals has not been investigated so far. We previously reported that oxidative stress inhibited synaptic transmission at the neuromuscular junction, targeting primarily the motor nerve terminals. In the current study, we investigated the effect of HCY on oxidative stress-induced impairment of transmitter release at the mouse diaphragm muscle. The mild oxidant H2O<sup>2</sup> decreased the intensity of spontaneous quantum release from nerve terminals (measured as the frequency of miniature endplate potentials, MEPPs) without changes in the amplitude of MEPPs, indicating a presynaptic effect. Pre-treatment with HCY for 2 h only slightly affected both amplitude and frequency of MEPPs but increased the inhibitory potency of H2O<sup>2</sup> almost two fold. As HCY can activate certain subtypes of glutamate N-methyl Daspartate (NMDA) receptors we tested the role of NMDA receptors in the sensitizing action of HCY. Remarkably, the selective blocker of NMDA receptors, AP-5 completely removed the sensitizing effect of HCY on the H2O2-induced presynaptic depressant effect. Thus, at the mammalian neuromuscular junction HCY largely increases the inhibitory effect of oxidative stress on transmitter release, via NMDA receptors activation. This combined effect of HCY and local oxidative stress can specifically contribute to the damage of presynaptic terminals in neurodegenerative motoneuron diseases, including ALS.

> Keywords: amyotrophic lateral sclerosis, neuromuscular junction, homocysteine, oxidative stress, NMDA receptors

## Introduction

Homocysteine (HCY, 2-amino-4-sulfanylbutanoic acid) is a sulphur-containing endogenous amino acid, which is produced in the methylation cycle of protein metabolism and involved in maintaining the cells redox balance. An elevated plasma level of HCY (termed hyperhomocysteinemia, hHCY) markedly decreases cell viability (Kolling et al., 2013). Extremely high HCY levels (up to 200 µM) has been found in some patients with disrupted HCY metabolism, and is believed to be at the root of certain vascular disorders including stroke and coronary occlusions (Kang et al., 1992; Stanger et al., 2009). According to a number of studies, abnormally elevated levels of HCY in plasma and cerebrospinal fluids correlate with amyotrophic lateral sclerosis (ALS), a motor neuronal disease that causes muscle degeneration (Zoccolella et al., 2008, 2010; Levin et al., 2010; Valentino et al., 2010; Veeranki and Tyagi, 2013). Inherited deficiency of the enzyme methylenetetrahydrofolate reductase (MTHFR), transforming HCY into methionine through remethylation, is associated with severe muscular hypotonia (Huemer et al., 2015). There is conflicting evidence about the role of the frequent MTHFR C677T polymorphisms as risk factors for ALS (Kühnlein et al., 2011; Ricci et al., 2012; Sazci et al., 2012). The variant MTHFR C677T, known to be associated with increased levels of HCY (Kang et al., 1988; Li et al., 2003) is more frequent among ALS patients (Kühnlein et al., 2011), however an Italian population study found no association (Ricci et al., 2012). Apart from a direct link between MTHFR and ALS, there are other possible mechanisms how the increasing level of HCY due to aging or improper diet (Zhang et al., 2008; Song et al., 2013) can speed up the development of disease in patients with ALS.

Recently, a link of the other Ca2+-dependent enzyme transglutaminase 2 with several neurodegenerative diseases including ALS has been proposed in a study showing HCYinduced activation of THP-1 monocytes associated with oxidative stress (Gurrò et al., 2015).

The role of HCY in ALS is supported by several studies showing the beneficial effects of folate or vitamin B12/B6 treatments (Zhang et al., 2008; Song et al., 2013), which are well known most efficient approaches to lower the level of endogenous HCY. In line, a large-scale human trial on the action of methylcobalamin in ALS patients is currently ongoing (Ikeda et al., 2015).

According to the ''dying-back'' hypothesis of ALS (Moloney et al., 2014; Pollari et al., 2014) the dysfunction of nerve terminals precedes pathological neurodegenerative changes in a motoneuron. One of the commonly reported mechanisms of HCY action is the enhanced production of reactive oxygen species (ROS; Loureiro et al., 2010; Lee et al., 2013; Veeranki and Tyagi, 2013). In the neuromuscular junction oxidative stress induced by hydrogen peroxide (H2O2) inhibits synaptic transmission at the presynaptic site (Giniatullin and Giniatullin, 2003; Giniatullin et al., 2006). The role of oxidative stress in degeneration of the neuromuscular junction in ALS was reviewed by us recently in the frame of this research topic (Pollari et al., 2014). However, despite the recent reports that hHCY is associated with impaired muscle functions (Swart et al., 2013; de Jager, 2014) the action of HCY on neuromuscular transmission and its role in the oxidative damage of motor nerve terminals has not been investigated so far.

Thus, the first aim of the current study was to test the ability of HCY to promote the depressant action of ROS on synaptic transmission using isolated neuromuscular junctions.

Several independent observations are pointing towards a potential correlation between enhanced glutamate receptor signaling and ALS progression. In the nervous system, extracellular HCY is able to stimulate the glutamate N-methyl D-aspartate (NMDA) receptor (Lipton et al., 1997; Poddar and Paul, 2013; Abushik et al., 2014). Homocysteinic acid, an analog of HCY with potent agonist activity on NMDA receptors triggers calcium accumulation in motor neurons (Adalbert et al., 2002). Increased glutamate levels were found in blood plasma of patients with ALS (Cecchi et al., 2014). A presynaptic glutamate release inhibitor Riluzole applied to ALS patients as a medical treatment extends life by 3–5 months (Miller et al., 2012) suggesting a role of glutamate receptors in the progression of this disease.

At the mammalian neuromuscular junction both NR1 and NR2A subunits of NMDA receptor have been found (Mays et al., 2009). In rats, the inhibition of NMDA receptors suppressed contractions of the diaphragm muscle, without changing acetylcholine (ACh) signaling (Koyuncuo˘glu et al., 1998). Hyperactivation of NMDA receptors is often associated with oxidative stress (Lee et al., 2013), calcium overload and ROS generation (Ratan and Baraban, 1995). Thus, the second aim of this study was to test the involvement of muscle NMDA receptors in synaptic action of HCY and ROS. For this aim, the effect of HCY and H2O<sup>2</sup> on transmitter release was investigated at the mouse neuromuscular junctions pre-treated with the selective blocker of NMDA receptors.

### Materials and Methods

#### Preparation and Solutions

Experiments were performed on isolated mice phrenic nerve—diaphragm preparations kept at room (22–23◦C) temperature. Mice (BALB/c strain) of both sexes of 20–25 g body weight were euthanized in accordance with regulations of the European Community Council Directive (September 22, 2010; 2010/63/EEC). Animal experiments were approved by the Ethical Committee of the University of Eastern Finland. Preparations were placed in a small 3.5-ml volume chamber and continuously superfused with gassed (95% O2/5% CO2) physiological solution containing (in mM): NaCl 120; KCl 5; CaCl<sup>2</sup> 2; NaHCO324; NaHPO<sup>4</sup> 1; MgCl<sup>2</sup> 1; glucose 11, at pH 7.3–7.4. All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) and were dissolved directly before the use. We used D, L-HCY with expected ∼50% of the natural active L-form of HCY (Lipton et al., 1997). Drugs were applied via bath perfusion (rate ∼ 2 ml/min).

#### Electrophysiology

Miniature end-plate potentials (MEPPs) recorded by microelectrodes (resistance 5–8 MΩ; filled with 3 mM KCl) as shown in details in Naumenko et al. (2011). In brief, MEPPs were acquired using the data acquisition board NI PCI6221 (National Instruments, Austin, TX, USA), visualized with WinEDR V3.2.7 software (Strathclyde University, UK), digitally stored at 50 kHz. Off-line amplitudes and interevent intervals of MEPPs were analyzed with the WinEDR V3.2.7 and ClampFit V10.2.0.14 (Molecular Devices, Sunnyvale, CA, USA) software. Subsequent statistical analysis was made using the Origin graphic software (Origin Lab Corp, Northampton, MA, USA). MEPPs were collected during 3 min from each of 8–10 synapses at the same muscle preparation to calculate mean values of amplitudes and frequencies in control solution and after 30–35 min application of 300 µM H2O2. Muscle fibers with a resting membrane potential less negative than −60 mV were rejected from further analysis.

#### Statistical Analysis

In each experimental series H2O<sup>2</sup> effect was counted by comparing the mean MEPP frequencies and amplitudes before and after ROS application (30–35 min). Mean values before H2O<sup>2</sup> were taken as 100%. The data are presented as the mean ± SEM (n = number of synapses with indication of number of animals used). Statistical significance was assessed by using Student's t-test (for parametric data) or Mann-Whitney test (for nonparametric data). Differences were considered significant when p < 0.05.

### Results

We first tested the action of acute oxidative stress on transmitter release in control conditions. MEPPs representing spontaneous release of ACh quanta from nerve terminals occurred at a mean frequency of 0.60 ± 0.05 s−<sup>1</sup> and an amplitude of 0.52 ± 0.02 mV (n = 46 synapses/6 mice). These quantal events are classical readouts of the functional state of the motor synapse when the frequency of MEPPs reflects the presynaptic function whereas the amplitude of MEPPs characterizes the functional state of the postsynaptic membrane (Fatt and Katz, 1952). To model acute oxidative stress we used the mild oxidant H2O<sup>2</sup> in concentration of 300 µM. Based on our previous experience (Giniatullin et al., 2006) we considered this concentration efficient and safe. H2O<sup>2</sup> applied via bath perfusion significantly reduced the frequency of MEPPs (**Figures 1A,C,D**). After 30–35 min exposure to H2O<sup>2</sup> the mean frequency of MEPPs decreased by 32.3 ± 8.6% (n = 47 synapses/6 mice, p < 0.05 by Mann-Whitney test) regarding control untreated preparation. No changes were observed in MEPP amplitudes (47 synapses/6 mice, p > 0.05; **Figures 1A–D**) indicating a pure presynaptic effect of H2O2.

The temporal characteristics of MEPP were not changed by H2O2. Thus, the rise-time in control was 0.20 ± 0.01 ms (n = 47 synapses/6 mice) and the decay time constant of MEPP was 3.18 ± 0.28 ms (n = 44 synapses/5 mice). After H2O<sup>2</sup> treatment these parameters remained unchanged (95 ± 7% and 94 ± 8% from control values, respectively, n = 44 synapses/5 mice, p > 0.05).

Next, we tested the action of HCY on spontaneous ACh release and whether this exposure changed the inhibitory effect of H2O2. As we intended to model naturally long-lasting hHCY in the short time window we selected the relatively large concentration of HCY of 500 µM which nevertheless corresponds the severe hHCY in humans (Kang et al., 1992; Stanger et al., 2009). Notably, as we used the racemic D, L-form of HCY, the effective concentration of the naturally occurring L-form was actually only half of the total (Lipton et al., 1997). Nevertheless, after incubation of the neuromuscular preparations for 2 h in solution containing 500 µM HCY, the mean frequency of MEPPs was the same as in untreated preparations (+5.8 ± 9.0% change comparing with untreated samples; n = 56 synapses/6 mice). No changes were observed in the amplitudes of MEPPs (**Figures 1A–D**) and in the time-course of MEPPs (**Figure 1B**) after HCY exposure (rise-time = 94 ± 5% and the decay time constant = 95 ± 7%, p > 0.05, n = 49 synapses/ 6 mice).

However, after HCY pre-treatment (500 µM, 2 h) the inhibitory potency of H2O<sup>2</sup> was largely increased. MEPPs frequency decreased by 60.0 ± 3.6% from the level in the presence of HCY before H2O<sup>2</sup> application (n = 55 synapses/7 mice, p < 0.05; **Figures 1A–D**). Thus, the depressant effect of H2O<sup>2</sup> on spontaneous ACh release was significantly (p < 0.05 by Mann-Whitney test) more (almost two times) efficient after muscles exposure to HCY compared to the action of H2O<sup>2</sup> alone. No changes were observed in the amplitudes of MEPPs (**Figures 1A–D**). Thus, HCY largely amplified the impairment of motor nerve terminals induced by acute oxidative stress.

Next, we explored potential mechanisms underlying the sensitizing effect of HCY. As we (Abushik et al., 2014) and others (Lipton et al., 1997; Poddar and Paul, 2013) showed that HCY can act via activation of NMDA receptors we next tested the role of these receptors in the sensitizing action of HCY. Several previous reports indicated the presence of NMDA receptors at the neuromuscular junction (Grozdanovic and Gossrau, 1998; Mays et al., 2009; Malomouzh et al., 2011; Walder et al., 2013). In these experiments we applied H2O<sup>2</sup> after 20 min pre-incubation with the selective NMDA receptor blocker AP-5 (50 µM, Abushik et al., 2014), following by incubation of muscles with a mixture of 50 µM AP-5 and 500 µM HCY for 2 h. Remarkably, H2O<sup>2</sup> (300 µM) applied after muscle incubation in the mixture of AP-5 plus HCY did not cause the depressant action on spontaneous ACh release from nerve endings (**Figures 2A,B**). Thus, MEPPs frequency in 30–35 min of H2O<sup>2</sup> exposure was even by 8.8 ± 11.1% higher than control level after exposure to AP-5 + HCY (n = 53 synapses/6 mice, p > 0.05). However, a similar insignificant trend (10 ± 7% change, n = 48 synapses/5 mice, P = 0.85) was also observed when AP-5 was applied to the muscle alone (**Figure 2B**). MEPP amplitudes were not significantly changed by H2O<sup>2</sup> applied after pre-exposure to AP-5 + HCY (**Figures 2A,C**). Likewise, MEPPs kinetics was not significantly changed (rise-time = 96 ± 6% and the decay time constant = 94 ± 8% of control values in AP-5 + HCY, n = 53 synapses/6 mice, p > 0.05).

The results of these experiments indicated NMDA receptormediated mechanism of sensitizing action of HCY.

The main finding of our study is that the redox active HCY increases the vulnerability of the mammalian neuromuscular junction to the inhibitory action of ROS via a glutamatergic mechanism. This finding uncovers a novel plausible mechanism for peripheral synaptic impairments, which may be relevant to motor neuron diseases such as ALS, as these disorders can be associated with hHCY and with oxidative stress. By modeling these conditions, our study suggests a new explanation for why motoneurons (and their peripheral parts) are more vulnerable to ALS associated with hHCY than other type of neurons.

There is strong evidence of the development of oxidative stress and of the detrimental role of ROS in ALS (reviewed in Pollari et al., 2014). At the neuromuscular junction, ROS are primarily generated at postsynaptic sites during intense muscle activity (Jackson, 2008) and can be induced via presynaptic mitochondria activity (David and Barrett, 2003; see also **Figure 3**). Dysfunction of mitochondria and Ca2<sup>+</sup> dysregulation contribute to the muscle denervation and motor neuron death that occur in mouse models of ALS (Jaiswal, 2013; Barrett et al., 2014).

The presynaptic transmitter releasing machinery is the most redox-sensitive part of the motor synapse (Giniatullin and Giniatullin, 2003; Giniatullin et al., 2006) and such a mechanism probably contributes to the muscle fatigue during the development of ALS (Pollari et al., 2014).

Despite the recent report that hHCY is associated with impaired muscle functions (Swart et al., 2013; de Jager, 2014) the action of HCY has not yet been studied directly at the neuromuscular junction. However, the role of HCY in ALS where the impairment of the neuromuscular transmission is one of the early and leading symptoms of the disease is consistently discussed. Thus, many studies reported the high level of HCY in ALS (Zoccolella et al., 2008, 2010; Levin et al., 2010; Valentino et al., 2010; Veeranki and Tyagi, 2013). In particular, high levels of HCY in both plasma and cerebrospinal fluids were found to correlate with ALS muscle degeneration (Zoccolella et al., 2008; Valentino et al., 2010). However, conflicting data should also be mentioned. Thus, while some authors found a link between MTHFR C677T polymorphisms as risk factors for ALS (Kühnlein et al., 2011; Sazci et al., 2012) others (Ricci et al., 2012) did not detect a direct association between the two.

Although a causal link remains disputable, the simultaneous presence of hHCY and ALS associated oxidative stress in nerve terminals could initiate the mechanism of presynaptic depression observed in the current study. According to this view there is a ''two keys mechanism'' when combined HCY and ROS synergistically impair the presynaptic compartment of the neuromuscular junction (**Figure 3**). The attractiveness of this hypothesis is that it integrates all of the three potential partners of the neuromuscular junction—muscle fibers, Schwann cells and nerve terminal which all could be sources of ROS (**Figure 3**).

HCY induced exacerbation of ALS is supported by clinical evidence of the favorable effect of HCY-lowering folate and B12 supplements (Zhang et al., 2008; Song et al., 2013; Ikeda et al., 2015). One could suggest that the measurement

of endogenous level of HCY in ALS patients should be monitored to determine the best type of treatments (and perhaps prophylaxis). Increase in HCY levels by low plasma folate or improper nutrition (Kang et al., 1992) might be an additional overlapping condition aggravating the development of ALS. Therefore, even though unable to cure ALS alone, the folic acid therapy could be an important component of a more complex disease therapy. Another important implication of our results is that patients with ALS should avoid nitrous oxide for anaesthesia as the latter could induce an acute burst of HCY (Asghar and Ali, 2012; Morris et al., 2015).

Interestingly, increased total HCY levels can be detected in a variety of neurological diseases such as stroke (Kang et al., 1992), Alzheimer's disease and other types of dementia (Isobe et al., 2005; de Jager, 2014), psychiatric disorders or infections such sepsis (Ploder et al., 2010) without reported signs of ALS symptomatology. This raises a question: why would ALS symptomatology develop only in this specific group of patients? To answer to this question we propose a working hypothesis based on the main finding of the current study: the interaction between ROS and HCY, that whereas hHCY is not specific to ALS (being observed in other abovementioned diseases) the coincidence of hHCY and oxidative damage locally in the peripheral part of the motoneuron is specific to ALS. In addition, the level of HCY is almost 10 times higher in blood than in the CSF (Valentino et al., 2010) suggesting a higher probability of HCY-mediated toxicity in peripheral tissues such as nerve endings in the skeletal muscle. Thus, our hypothesis can provide a novel explanation of why peripheral endings of the motoneuron are more vulnerable to ALS than central neurons affected by other brain diseases.

In the current study, in order to model the detrimental effects of HCY in a short time frame we used high concentrations of the D-L form of HCY. In fact, the actual concentration of the active L-form of HCY (Lipton et al., 1997) in our experiments is at least two times lower (∼250 µM) which is close to what is observed in severe hHCY (Kang et al., 1992; Stanger et al., 2009). Notably, even at such high concentrations HCY did not significantly affect neuromuscular transmission parameters per se. The accuracy of measurement of the fraction of total and free HCY in pathological states is still a matter of debate (Adam et al., 2013). It is also worth noting that, in addition to HCY, a related endogenous compound homocystine is present in the blood (Zhang et al., 2014), and its molecular targets deserve future studies.

The main finding of the current study was that the exposure to HCY largely promoted the presynaptic inhibitory action of H2O2. Moreover, in our study we provide a mechanistical explanation for this phenomenon identifying the key role of glutamate NMDA receptors in the sensitizing effect of HCY. Indeed, the selective blocker of NMDA receptors AP-5 completely prevented the additive effect of HCY on the inhibition of transmitter release by H2O2. Our finding is consistent with previous studies showing that HCY induces excitotoxic effects mainly in cells expressing NMDA receptors (Boldyrev et al., 2013). Activation of NMDA receptors by HCY results in an increase of cytoplasmic Ca2<sup>+</sup> and ROS accumulation (Loureiro et al., 2010; Leal and Gomes, 2015) which all are likely happening in skeletal muscle during hHCY (**Figure 3**).

Glutamate ionotropic receptors are located at vertebrate neuromuscular junctions (Grozdanovic and Gossrau, 1998). Knockouts of NMDA receptor subunits in mice induces phenotypes which are similar to those typically associated with neuromuscular disorders (Single et al., 2000; Meng et al., 2003). The presence of AP-5 sensitive NMDA receptors in the muscle has been shown in several previous studies involving immunolabeling and electrophysiological recordings (Mays et al., 2009; Malomouzh et al., 2011; Walder et al., 2013). However, in contrast to the high diversity of glutamate receptor subunits observed in the CNS, only a limited number of glutamate receptor subunit types is expressed in the neuromuscular junction (Mays et al., 2009). As we did not see changes in the amplitude of MEPPs, we suggest only limited and probably diffuse expression of postsynaptic NMDA receptors in skeletal muscle.

Downstream NMDA receptor signaling pathway could involve the highly diffusible messenger NO (**Figure 3**). It is known that in neurons, glutamate, acting on post-synaptic NMDA receptors can induce NO release as nitric oxide synthase is a part of the NMDA/NOS complex (Courtney et al., 2014). Consistent with this, HCY neurotoxicity could be due to the release of NO (Stamler et al., 1993) which is a potent inhibitor of ACh release from motor nerve endings (Giniatullin et al., 2005; Valiullina and Sitdikova, 2012). Early studies indicate that the activation of muscle NMDA receptors by bath-applied glutamate increases synthesis of NO in muscle fibers and NO acts in a retrograde manner on motor nerve terminals to inhibit nonquantal release of ACh (Malomouzh et al., 2003).

NO can also cooperate with superoxide to produce the strong oxidant, peroxynitrite, which is considered as a candidate inducer of cell death in ALS (Urushitani and Shimohama, 2001). Whereas the involvement of NO in HCY induced sensitizing effects require further investigations, especially taking into account potential species-dependent action of NMDA/NO mediated signaling which could explain the fact that the retrograde presynaptic inhibition by ROS pre-conditioned by HCY was prevented by the NMDA antagonist AP-5.

Schwann cells surrounding nerve terminals could be additional contributors to the impaired presynaptic function (Pollari et al., 2014) via release of gliotransmitters (Araque et al., 2014) including D-serine (Wu et al., 2005; see also **Figure 3**). Glial cells D-serine like glycine can potentiate the neurotoxicity of HCY via the glycine binging site of NNDA receptor (Wolosker, 2007; McCully, 2009; see also **Figure 3B**). In line with this, D-serine can contribute to the pathophysiology of familial ALS associated with the R199W mutation in D-amino acid oxidase (Paul and de Belleroche, 2015).

In summary, we show here that motor nerve terminals are highly vulnerable to the combined action of HCY and oxidative stress, both of which can be present in ALS and likely contribute together to the presynaptic impairment of the neuromuscular function. Thus, we propose a model, which may be relevant for ALS and raises a number of important questions with significant translational impact, which could be investigated further in animal ALS models and in future clinical studies.

#### References


#### Funding

This project was supported by the Russian Scientific Fund (Grant No 14-15-00618).

#### Acknowledgments

The authors are grateful to Anton Hermann, Gundars Goldstein, Mustafa Atalay and Genevieve Bart for useful suggestions.


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Bukharaeva, Shakirzyanova, Khuzakhmetova, Sitdikova and Giniatullin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# NSC-34 Motor Neuron-Like Cells Are Unsuitable as Experimental Model for Glutamate-Mediated Excitotoxicity

Blandine Madji Hounoum<sup>1</sup> , Patrick Vourc'h1,2 , Romain Felix <sup>3</sup> , Philippe Corcia1,4 , Franck Patin1,2 , Maxime Guéguinou<sup>3</sup> , Marie Potier-Cartereau<sup>3</sup> , Christophe Vandier <sup>3</sup> , Cédric Raoul <sup>5</sup> , Christian R. Andres 1,2 , Sylvie Mavel <sup>1</sup> and Hélène Blasco1,2 \*

1 Institut National de la Santé et de la Recherche Médicale (INSERM U930) "Imagerie et Cerveau", CHRU de Tours, Université François-Rabelais, Tours, France, <sup>2</sup> Laboratoire de Biochimie et de Biologie Moléculaire, Hôpital Bretonneau, CHRU de Tours, Tours, France, <sup>3</sup> Institut National de la Santé et de la Recherche Médicale (INSERM U1069) "Nutrition, Growth and Cancer", Université François-Rabelais de Tours, Tours, France, <sup>4</sup> Centre SLA, Service de Neurologie, CHRU de Tours, Tours, France, <sup>5</sup> The Neuroscience Institute Montpellier, Institut National de la Santé et de la Recherche Médicale (INSERM UMR1051), Saint Eloi Hospital, Montpellier, France

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Joe Tauskela, National Research Council Canada, Canada Tatsuro Mutoh, Fujita Health University School of Medicine, Japan Andrea Nistri, Scuola Internazionale Superiore di Studi Avanzati (SISSA), Italy

> \*Correspondence: Hélène Blasco helene.blasco@univ-tours.fr

Received: 16 February 2016 Accepted: 25 April 2016 Published: 09 May 2016

#### Citation:

Madji Hounoum B, Vourc'h P, Felix R, Corcia P, Patin F, Guéguinou M, Potier-Cartereau M, Vandier C, Raoul C, Andres CR, Mavel S and Blasco H (2016) NSC-34 Motor Neuron-Like Cells Are Unsuitable as Experimental Model for Glutamate-Mediated Excitotoxicity. Front. Cell. Neurosci. 10:118. doi: 10.3389/fncel.2016.00118 Glutamate-induced excitotoxicity is a major contributor to motor neuron degeneration in the pathogenesis of amyotrophic lateral sclerosis (ALS). The spinal cord × Neuroblastoma hybrid cell line (NSC-34) is often used as a bona fide cellular model to investigate the physiopathological mechanisms of ALS. However, the physiological response of NSC-34 to glutamate remains insufficiently described. In this study, we evaluated the relevance of differentiated NSC-34 (NSC-34D) as an in vitro model for glutamate excitotoxicity studies. NSC-34<sup>D</sup> showed morphological and physiological properties of motor neuron-like cells and expressed glutamate receptor subunits GluA1–4, GluN1 and GluN2A/D. Despite these diverse characteristics, no specific effect of glutamate was observed on cultured NSC-34<sup>D</sup> survival and morphology, in contrast to what has been described in primary culture of motor neurons (MN). Moreover, a small non sustained increase in the concentration of intracellular calcium was observed in NSC-34<sup>D</sup> after exposure to glutamate compared to primary MN. Our findings, together with the inability to obtain cultures containing only differentiated cells, suggest that the motor neuron-like NSC-34 cell line is not a suitable in vitro model to study glutamate-induced excitotoxicity. We suggest that the use of primary cultures of MN is more suitable than NSC-34 cell line to explore the pathogenesis of glutamate-mediated excitotoxicity at the cellular level in ALS and other motor neuron diseases.

#### Keywords: ALS, Ca2+ influx, differentiation, NSC34, glutamate receptors, NMDA

**Abbreviations:** α-MEM, Alpha-modified Eagle's medium; ALS, amyotrophic lateral sclerosis; AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid; Ca2+, Calcium ion; DAPI, 4,6-diamino-2-phenylindole; DMEM, Dulbecco's modified Eagle's medium; FCS, fetal calf serum; KA, kainite; MEM-NEAA, modified Eagle's medium non-essential amino acids; NSC, Neuroblastoma × Spinal Cord; NSC-34D, differentiated NSC-34; MN, motor neurons; NTFs, neurotrophic factors; PBS, phosphate buffered saline; P/S, penicillin/streptomycin; RA, all-trans retinoic acid.

### INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is one of the most common neurodegenerative diseases in adults, caused by the selective death of motor neurons (MN). Studies of the physiology of ALS support the involvement of genetic factors and microenvironmental factors with mechanisms such as glutamate excitotoxicity and oxidative stress, but these mechanisms await elucidation. Glutamate excitotoxicity is a major contributor to dysfunction and death of MN in the pathogenesis of ALS (Heath and Shaw, 2002; Van Den Bosch et al., 2006; Spalloni et al., 2013; Blasco et al., 2014). Among the many drugs targeting this pathogenic mechanism tested in clinical trials (Blasco et al., 2014), one treatment (riluzole) is used in common practice to slow the progression of the disease by blocking glutamatergic neurotransmission in the central nervous system. However, the mechanisms of toxicity caused by glutamate and their links with other physiopathological pathways remain poorly understood.

Overstimulation of glutamate receptors facilitates the entry and consequently the excess of calcium (Ca2+) in cell compartments, leading to a cascade of destructive events by calcium-dependent enzymatic pathways and mitochondrial dysfunction with the generation of free radicals (Van Den Bosch et al., 2000; Blasco et al., 2014). Glutamate receptors are divided into two families: the ligand-gated cation channels (ionotropic; Lodge, 2009) and G protein-coupled receptors (metabotropic; Niswender and Conn, 2010). Ionotropic receptors are further divided into three categories according to their non-natural preferred agonists, i.e., AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid), NMDA (N-methyl-D-aspartate) and kainate (AK; Lodge, 2009). AMPA and NMDA receptors are largely responsible for calcium flux across neuronal cell membranes (Hollmann et al., 1991; Van Den Bosch et al., 2000).

Primary cultures of MN (Krugman et al., 1999) and cultures of different cell lines are used to study the molecular mechanisms of neurotoxicity induced by glutamate in MN. The most common cell line used in ALS research is the spinal cord neuron × neuroblastoma hybrid cell line (NSC-34), which was originally described as having several morphological and physiological properties of MN (Eggett et al., 2000; Rembach et al., 2004; Benkler et al., 2013; Maier et al., 2013; Valbuena et al., 2015). We required validation of this model to study the metabolic effects of glutamate excitotoxicity in a neurological disease. NSC-34 is an hybrid cell line produced by the fusion of MN from the spinal cords of mouse embryos with mouse neuroblastoma cells N18TG2 (Cashman et al., 1992). These cells exhibit properties of MN when subjected to protocols of differentiation and maturation. Although NSC-34 cells share many features with MN such as long processes and S-laminin with leucine-arginine-glutamate (LRE) adhesion motif (a specific basal lamina glycoprotein concentrated at the neuromuscular synapse; Hunter et al., 1991), the formation of contacts with myotubes in culture (Cashman et al., 1992) and increased survival in the presence of neurotrophic factors (NTFs; Turner et al., 2004), their functional properties when stimulated by glutamate have rarely been studied (Durham et al., 1993; Eggett et al., 2000).

The selective motor neuronal cell death in ALS is highly dependent on intracellular Ca2<sup>+</sup> and is insensitive to inhibitors of voltage-operated Ca2<sup>+</sup> and Na<sup>+</sup> channels (MacDermott et al., 1986; Van Den Bosch et al., 2000, 2006; Heath and Shaw, 2002). Thus, the permeability of glutamate receptors to Ca2<sup>+</sup> seems crucial to validate the NSC-34 model as appropriate for the exploration of glutamate excitotoxicity. In order to evaluate whether NSC34 represents a relevant model for studies on glutamate-induced neurotoxicity, we differentiated the NSC-34 cells by serum depletion in the presence or absence of all-trans retinoic acid (RA; Clagett-Dame et al., 2006; Maden, 2007), and analyzed the expression and functionality of glutamate receptors. We compared for the first time the usefulness of NSC-34 cells (using various differentiation protocols) and primary motor neuron cultures (used as the reference) as models for glutamate induced-excitotoxicity in the same study.

### MATERIALS AND METHODS

### Materials

Dulbecco's modified Eagle's medium (DMEM, Gibco), DMEM-Ham's F12 and Alpha-modified Eagle's medium (α-MEM), modified Eagle's medium non-essential amino acid (MEM-NEAA), penicillin/streptomycin (P/S), Trizol, Superscript II reverse transcriptase kit and Fura-2 acetoxymethyl ester were all obtained from Invitrogen (Life Technologies, Saint Aubin, France), L-glutamic acid and all-trans RA from Sigma Aldrich (Saint-Quentin Fallavier, France) and fetal calf serum (FCS) from Eurobio (Courtaboeuf, France).

### Culture of NSC-34

The NSC-34 was obtained from Cedarlane Laboratories (via Tebu-Bio, Le Perray en Yvelines, France). Cells were cultured as described previously (Madji Hounoum et al., 2015). Cultures were used 5–15 passages. Each type of experiment was performed on the same passage. No sub-culture passage was performed for differentiated NSC-34.

For differentiation, NSC-34 cells were grown to confluence and the proliferation medium (DMEM plus 10% FCS) was exchanged for fresh differentiation medium every 3 days. Cells were allowed to differentiate for up to 4 weeks. Three differentiation media were investigated: (1) 1:1 DMEM/Ham's F12 plus 1% FCS, 1% P/S and 1% MEM-NEAA, the most commonly used medium for NSC-34 differentiation (Kruman et al., 1999; Eggett et al., 2000; Rembach et al., 2004; Benkler et al., 2013); (2) α-MEM [the medium used for another neuron-like cell line (P19; MacPherson et al., 1997)] plus 1% FCS, 1% P/S and 1% MEM-NEAA; and (3) DMEM (the classic medium for NSC-34 culture) plus 1% FCS and 1% P/S.

Two conditions were investigated for all differentiation media, i.e., with or without RA [1 µM, as used previously (Johann et al., 2011; Maier et al., 2013)]. NSC-34 cells, maintained on proliferation medium, served as the undifferentiated control group. The average length of neurites in the differentiation media was quantified using Sholl's method for quantification of dendritic branching in hippocampal neurons (Sholl, 1956). Briefly, concentric circles at 25 µm intervals between adjacent circles were drawn on Powerpoint, at the same magnification of cell pictures. The center of the circle centered on the soma of the cell, the lengths of the processes were measured from the soma by multiplying the number of intersections (neuritecircle) every 25 µm. Cells with neurites longer than 50 µm were considered as differentiated. Neurite length was analyzed by imaging a minimum of 10 cells per experiment, four experiments for each condition.

### Primary Motor Neuron Cultures

Studies were conducted using primary cultures of motor neuron from the spinal cords of C57BL/6 mice at embryonic day 12.5 (Centre d'Elevage Roger Janvier, France). Cultures were grown as described previously (Camu et al., 2014; Dangoumau et al., 2015). MN were plated on poly-ornithine/laminin-treated wells in the presence of NTFs (0.1 ng/mL GDNF, 1 ng/mL BDNF, and 10 ng/mL CNTF in supplemented neurobasal medium (Invitrogen, Carlsbad, CA, USA)). Supplemented neurobasal medium contained 2% horse serum, L-glutamate (25 mM), β-mercaptoethanol (25 mM), L-glutamine (0.5 mM), and 2% B-27 supplement (Invitrogen, Life Technologies, Saint Aubin, France). The use of appropriate culture medium combined with multiple purification steps using density gradient (BSA cushion and Optiprep density centrifugation) and magnetic cell sorting with an indirect microbeads technique promoted the enrichment of MN as well as the elimination of astrocytes and microglia cells from the culture (Arce et al., 1999).

### Immunocytochemistry

To assess morphological characterization of primary motor neuron culture derived from embryonic spinal cord, expression of βIII-tubulin and p75 neurotrophic receptor were analyzed (Rembach et al., 2004). Cells were fixed in 4% paraformaldehyde and then incubated in a blocking and permeabilizing solution (10% donkey serum and 2% Triton X-100) for 1 h. The cells were stained with rabbit anti-βIII-tubulin (1:200, Covance, Princeton, NJ, USA) and mouse anti-nerve growth factor receptor (p75NTR 1:60, Chemicon MAB357) for 1 h. After being washed with phosphate buffered saline (PBS), the cells were incubated with secondary antibodies for 1 h. The secondary antibodies used were donkey anti-rabbit Alexa-488 (1:300, Life Technologies, Saint Aubin, France) and donkey anti-mouse Alexa-594 (1:300, Life Technologies, Saint Aubin, France). Nuclear DNA was observed by 4,6-diamino-2-phenylindole (DAPI) contained in ProLong<sup>r</sup> (Life Technologies, Saint Aubin, France). Cells were photographed using a confocal microscope (Olympus FV500).

### Cell Viability Assay

To assess the effect of glutamate-induced excitotoxicity in NSC-34D, cells were exposed to glutamate at different concentrations (100 µM, 500 µM, 1 mM and 10 mM) for 48 h (Eggett et al., 2000; Rembach et al., 2004; Maier et al., 2013) before determination of cell viability by trypan blue assay. A control condition with no excitotoxic agent (PBS) was included for comparison. Four replicates per condition were obtained. The cells were harvested with trypsin and re-suspended in culture medium before addition of trypan blue (0.4% w/v; Molecular Probes-Invitrogen, Carlsbad, CA, USA). Living and dead cells were counted accurately with a Countess Automated Cell Counter (Invitrogen, Carlsbad, CA, USA). Motor neuron survival was assessed by direct counting after exposing cells to 100 µM glutamate for 48 h.

### RNA Extraction, Reverse Transcription PCR (RT-PCR) and Real-Time Polymerase Chain Reaction (RT-qPCR)

Following a wash in PBS, RNA samples were extracted from NSC-34 cells and primary MN using Trizol and treated with DNase I (Proteigener). Reverse transcription was performed on 2.5 µg of treated RNA using the superscript II reverse transcriptase kit according to the manufacturer's instructions. cDNA samples were analyzed using primers for subunits of AMPA receptors (GluA1–4), NMDA receptors (GluN1, GluN2A-D, GluN3A), and standard markers for MN such as choline acetyltransferase (ChAT), p75 (Cashman et al., 1992; Matusica et al., 2008; Maier et al., 2013; see Table A1 in Supplementary Material). Actin was used as internal standard. cDNA samples were amplified using GoTaq<sup>r</sup> Flexi DNA polymerase (Promega) in 25 µL reaction mixture containing 125 ng cDNA following the manufacturer's protocol. Amplification consisted of 34 cycles at 95◦C for 1 min, 59◦C for 1 min and 72◦C for 1.5 min. Aliquots of 10 µL of these products underwent electrophoresis on 1.5% agarose gel before capture with a ChemiDoc XRS camera and quantification by Quantity One Software (BioRad). Mouse brain extract was used as positive control for amplifications.

The gene expression of glutamate receptor subunits, ChAT and p75NTR, was measured by semi quantitative real-time PCR (RT-qPCR) using the Brilliant III Ultra-Fast SYBR<sup>r</sup> QPCR MM kit (AgilentTM) and the same primers as for conventional RT-PCR (see Table A1 in Supplementary Material). The reaction was performed in a LightCycler 480 (Roche Diagnostic, Meylan, France) at Tm 60–62◦C. The efficiency of amplification was calculated on cDNA at concentrations ranging from 3.75 to 60 ng/mL. To ensure the absence of genomic DNA contamination, a control sample of non-reversetranscribed RNA was run for each set of RNA extractions. The expression's stability of reference genes was determined following MIQE guidelines (Bustin et al., 2009). We showed that β-actin and GAPDH genes were stably expressed (data not shown). Relative quantification was obtained by calculating the ratio between the values obtained for each gene of interest and the reference genes (β-actin, GAPDH). Melting curves were routinely performed to determine the specificity of the qPCR reaction. The 2−∆∆Ct method was used for analysis.

### Measurement of Calcium Influx

Cells were seeded on coverslips at a density of 3 × 10<sup>4</sup> cells/mL and allowed to grow for 2 days. They were loaded with a fluorescent probe (1 µM Fura-2 AM) in culture medium for 45 min at 37◦C. Cells were then washed with medium and allowed to de-esterify for at least 15 min at room temperature. The dish was then placed on the stage of a Nikon Eclipse TE2000-S inverted fluorescence microscope (Nikon, France) equipped with a 75 W Xenon Arc Lamp (Ushio, Japan), an Optoscan Monochromator (Cairn Optoscan, Kent, UK), and an ORCA-03G (CCD) camera (Hamamatsu, Japan). The coverslips were placed in a perfusion chamber containing physiological solution with 2 mM calcium (see Table A2 in Supplementary Material).

Excitation light was chopped by the monochromator at the two excitation wavelength maxima of fura-2 (340/380 nm). The excitation protocol used was a 500 ms excitation at each wavelength every 4 s. Fluorescence emission at 510 was detected by the CCD digital camera. The software (Imaging workbench 6) excited cells and allowed acquisition of images. The acute effects of glutamate at 100 µM and 1 mM (for primary MN and NSC-34D, respectively) on calcium entry were measured in physiological solution for 200 s. For primary motor neuron cells, we first determined the optimal time window allowing the experiment by measuring calcium entry every 2 days on independent replicates (cells that had not been exposed to fura-2 experiment before).

### Statistical Analysis

Data was expressed as the mean ± SEM values. Statistical significance was determined using the Mann-Withney test or one-way Anova. For multiple comparisons, Tukey's tests were used as post hoc tests or Kruskal-wallis test for appropriate analyses. For all analyses, p values < 0.05 were considered significant. The type of statistical test is specified in each figure legend.

### RESULTS

### Efficient Differentiation of NSC-34 Cells into Motor Neuron-Like Cells

When NSC-34 cells were cultured in the differentiation media (NSC-34D), we distinguished two morphologically distinct populations: cells with short neurites, and cells with phenotypic characterization of MN with long processes (**Figure 1A**). However, this differentiation was preceded by a massive loss of cells (>50%) as described by Eggett et al. (2000), with surviving cells sprouting extensive neuron-like projections.

We next investigated whether the potent morphogen RA could enhance differentiation of NSC-34 into motor neuron-like cells. We found that addition of RA (1 µM) in differentiation medium resulted in a significant decrease (60–70%) in proliferation at 4 days. Surviving cells showed a very low rate of proliferation compared to conditions without RA, as previously shown (Johann et al., 2011). However, we observed that a

test, according to the different media: DMEM/Ham's F12, α-MEM and DMEM. Values are means ± SEM (10–11 cells per experiment and four experiments for each condition), <sup>∗</sup>p < 0.001, ∗∗p < 0.0001, ns = not significant, Mann-Whitney test + RA vs. none.

significant proportion of surviving cells had a motor neuron-like phenotype (∼80%) in the presence of RA as determined by morphological criteria (**Figure 1A**).

We then assessed whether differentiation conditions could potentiate the effects of RA on process outgrowth of NSC-34 cells. We determined the average neurite length of NSC-34 cells when they are maintained for 4 weeks in three differentiation media (α-MEM, DMEM/Ham's F12 and DMEM) in the presence or absence of RA. We found that the average neurite length has increased by 55% when they are cultured in α-MEM media with Madji Hounoum et al. Glutamate Excitotoxicity: NSC-34, Unsuitable Model

RA compared to α-MEM media without RA (240.00 ± 15.89 and 154.16 ± 7.01, respectively, p < 0.0001). While it has decreased by 20% when cells were cultured in DMEM/Ham's F12 media with RA compared to media without RA (181.82 ± 11.29 and 229.69 ± 12.09, respectively, p < 0.001; **Figure 1B**). We did not find any significant change when NSC-34 cells were cultured in DMEM with or without RA (200.45 ± 11.95 and 206.25 ± 13.13, respectively, p = 0.46). Therefore, the effects of RA on NSC-34<sup>D</sup> cell morphology were highly dependent on the differentiation medium (**Figure 1B**).

### Differentiated NSC-34 Cells Expressed Glutamate Receptor Subunits

We first set up conditions for optimal and specific amplification of glutamate receptor subunits GluA1–4, GluN1, GluN2A-D and GluN3A using total RNA obtained from mouse brain extract (see Figure A1 in Supplementary Material). We next studied the expression profiles of all glutamate receptors promoted by NSC-34 differentiation (up for 4 weeks) and in primary MN. We found that the differentiation of NSC-34 in the conditions that we had previously described induced a significant increase in mRNA expression of glutamate receptor subunits GluN1, GluN2A and GluN2D compared to undifferentiated cells (**Figures 2B–I**). The example of GluN2A protein confirms this statement: the result obtained by western blot analysis showed that the GluN2A protein level in NSC-34<sup>D</sup> was higher than in undifferentiated NSC-34 (see Figure A2 in Supplementary Material). Levels of GluA4 subunits mRNA have increased in DMEM without RA compared to the other conditions without RA (**Figures 2C,E,G**). At more discrete levels, we also observed increases of GluA2 and GluA3 mRNA in DMEM/Ham's F12 and α-MEM but not in DMEM. We noted the expression of all glutamate subunits in primary MN, as observed in mouse brain extracts (**Figure 2A** and Figure A1, see Supplementary Material).

We found that the presence of RA promoted the expression of GluA4 mRNA in every differentiation medium (**Figures 2C–H**). However, mRNA expression of the other subunits of glutamate receptors was not modified by the presence of RA in the culture medium. The differentiation condition that promoted most of the glutamate receptors and that generated less heterogeneity bias was thus DMEM-Ham's F12 without RA. Indeed, this differentiation condition led to expression of all AMPR subunits (GluA1–4) and half of the NMDAR subunits (GluN1, GluN2A, and GluN2D), as summarized in Table A3 (see Supplementary Material). However, expression of AMPR subunits was lower compared to primary MN (**Figures 2A,C,I,J**). DMEM-Ham's F12 without RA medium is used for further explorations. Our results also showed that NSC-34<sup>D</sup> cells expressed many motor neuron properties such as ChAT and p75NTR (**Figures 2C–H**). RA-differentiated NSC-34 did not express more receptor subunits than NSC-34<sup>D</sup> without RA (**Figures 2C–H** and Table A3 in Supplementary Material). We next semi quantified expression of glutamate receptor subunits in NSC-34<sup>D</sup> differentiated in DMEM-Ham's F12 without RA condition compared to MN.

### Expression of Receptor Subunits

RT-qPCR analysis revealed high levels of expression of GluN1, GluN2A and GluN2D mRNA (CP values were 8.35 × 10−<sup>3</sup> ± 4.96 × 10−<sup>5</sup> , 3.02 × 10−<sup>3</sup> ± 6.30 × 10−<sup>8</sup> and 4.95 × 10−<sup>4</sup> ± 1.83 × 10−<sup>6</sup> , respectively) in NSC-34 differentiated in DMEM/Ham's F12 without RA, while expression of GluA1, GluA2, GluA4, GluN2C and GluN3A mRNA (9.85 × 10−<sup>7</sup> ± 4.02 × 10−<sup>9</sup> , 4.45 × 10−<sup>7</sup> ± 5.78 × 10−<sup>9</sup> , 7.45 × 10−<sup>6</sup> ± 5.77 × 10−<sup>9</sup> , 3.22 × 10−<sup>6</sup> ± 1.80 × 10−<sup>7</sup> and 9.74 × 10 <sup>−</sup><sup>6</sup> ± 2.01 × 10−<sup>8</sup> , respectively) was low in NSC-34<sup>D</sup> (**Figures 2I,J**), i.e., as for the results of conventional RT-PCR (**Figure 2C**). GluA3, and GluN2B mRNA levels were almost zero in NSC-34<sup>D</sup> (**Figures 2I,J**). It is of note that only 20% of the cells did not express motor neuron-like morphology. The results showed that primary MN expressed all glutamate receptor subunits at different levels, unlike to NSC-34D. Consistently with RT-PCR data, the transcript profile of NSC-34<sup>D</sup> glutamate receptor subunits is different from that of MN.

### Differentiated NSC-34 Cells did not Show Susceptibility to Glutamate-Induced Death

We investigated whether glutamate could trigger death of differentiated NSC-34 cells. NSC-34<sup>D</sup> cells were exposed to increasing concentrations of glutamate ranging from 0.1 to 10 mM. We found 102.04 ± 3.63%, 104.08 ± 8.03%, 94.90 ± 1.44%, 86.73 ± 4.17% and 64.28 ± 2.89% of relative cell viability when cells exposed to 0.1 mM, 0.5 mM, 1 mM, 5 mM and 10 mM glutamate, respectively (**Figure 3A**). We also evaluated excitotoxicity death of MN at 8 DIV (Days In Vitro) exposed to 0.1 mM glutamate and we found 53.99 ± 3.04% of relative cell viability (**Figure 3B**). We observed that 48 h exposure to 0–5 mM glutamate did not elicit any toxic response to NSC-34D, and a non-specific effect of about 30% on cells survival was found when glutamate was added at 100-fold higher concentrations (10 mM; **Figure 3A**) than that necessary to induce 50% death of primary MN (0.1 mM; **Figure 3B**). The results show that NSC-34<sup>D</sup> cells do not show the same susceptibility to glutamate-induced excitotoxicity compared to MN.

### Glutamate did not Elicit Sustained Calcium Entry into NSC-34<sup>D</sup>

Glutamate-induced excitotoxicity resulted in elevated intracellular calcium levels via ionotropic glutamate receptors and activation of cell death signaling pathways. To address the effects of glutamate on Ca2<sup>+</sup> transients in NSC-34D, we used ratiometric Ca2<sup>+</sup> measurement with the imaging dye Fura-2AM. To establish sensitive and accurate measurement of intracellular Ca2+, we used motor neuron-enriched cultures. When MN were purified from E12.5 embryos as described above, a culture containing about 70% MN was identified using the p75NTR generic marker (Figure A3-A see Supplementary Material). The phase-bright soma with long motor neuron processes made them clearly distinguishable from other cells (Figure A3 see Supplementary Material). Their maturity was tested every 2 days by measuring [Ca2+] influx to determine the optimal time window allowing the experiment. We found

that younger MN did not efficiently load Fura-2AM before 5 DIV (data not shown). We observed that MN showed robust fluorescence signal from 5 DIV and glutamate induced a robust Ca2<sup>+</sup> influx in MN at 12 DIV (Figure A3-B, see Supplementary Material). The maximum fura-2 fluorescence intensity relative to baseline was significantly higher (p < 0.005) in MN at 12 DIV (0.0293 ± 0.0026) and 13 DIV (0.0320 ± 0.0004) compared to those at 5 DIV (0.0019 ± 0.0002). There was no significant difference between MN at 12 DIV and 13 DIV. Subsequent experiments were therefore performed using 13-day-old cultures

from untreated cells <sup>∗</sup>p < 0.05, ∗∗p < 0.01. Data are mean ± SEM (n = 3).

(Figure A3-B, see Supplementary Material). At 13 DIV, the MN are fully mature with electrical activity (Jackson et al., 1982; Nicola et al., 1992; Chang and Martin, 2011).

We then investigated the functionality of glutamate receptors expressed in NSC-34<sup>D</sup> growing in DMEM/Ham's F12 medium without RA for 4 weeks through Ca2<sup>+</sup> permeability. The baseline recordings (first 0.5 min in all trace graphs) demonstrate stable cytosolic calcium levels in both MN and NSC-34<sup>D</sup> cells before treatment with glutamate. The addition of glutamate immediately increased the fluorescence intensity of the calciumsensitive dye Fura-2, indicating increased cytosolic Ca2<sup>+</sup> flux in MN (**Figure 4A**). Whereas MN presented a sustained calcium entry following acute stimulation by glutamate (0.1 mM), glutamate stimulation in NSC34<sup>D</sup> resulted in a small transient calcium entry (**Figure 4B**). However, in the presence of glutamate, Ca2<sup>+</sup> entry into MN was significantly higher than in NSC-34<sup>D</sup> cells, and the signal was as consistent as that obtained by fura-2 fluorescence intensity relative to baseline (0.032 ± 0.0004 vs. 0.009 ± 0.0012; <sup>∗</sup>p < 0.001; **Figure 4C**). Taken together, these results showed that an increased extracellular glutamate concentration (1 mM) did not elicit the same calcium response in NSC-34<sup>D</sup> cells (**Figure 4C**) as in MN.

### DISCUSSION

Although glutamate-mediated excitotoxicity is known to be involved in ALS pathogenesis (Beal, 1992; Mattson et al., 1992), the precise mechanisms leading to electrical, metabolic or signaling dysfunction in the motor neuron diseases remain to be determined. Excitotoxicity mechanisms have been described following in vitro studies using either primary motor neuron cultures (Kruman et al., 1999; Van Den Bosch et al., 2000; Dolga et al., 2011; Joshi et al., 2011) or cell lines (Rembach et al., 2004; Benkler et al., 2013). The advantage of primary cultures is the development of similar in situ counterpart morphology and physiology over time. The neurotoxicity of a variety of toxins in situ has also been reported (Durham et al., 1993). However, due to the low yields of motor neuron cultures, the NSC-34 cell line has been widely used and tacitly considered to be the most stable motor neuron cell line to model ALS pathophysiology (Veyrat-Durebex et al., 2014). Indeed, many studies have shown that the differentiation process can induce NSC-34 cells to express certain motor neuron properties and have shown some toxic effects following exposure to glutamate (Eggett et al., 2000; Benkler et al., 2013; Maier et al., 2013). The differentiation and the maturation of NSC-34 cells are generally characterized by extension of neurites and expression of certain motor neuronspecific proteins. In this study, we investigated the reliability of these cells as an in vitro model to study glutamate-mediated neurotoxicity, using an approach from receptor expression to the Ca2<sup>+</sup> influx.

### Differentiation of NSC-34 Cells into Motor Neuron-Like Cells

Differentiation of cell lines frequently requires modification of the culture medium through serum depletion and/or and use of chemical reagents or metabolites such as RA to obtain more neuron-like properties, including neurite outgrowth and morphological changes (MacPherson et al., 1997; Eggett et al., 2000; Maier et al., 2013). As reported in the literature, we found that the differentiation process induced NSC-34 cells to express phenotypic MN with long processes (Figure 1A and

Figure A3-A, see Supplementary Material). We found that the RA effects on NSC-34<sup>D</sup> cell morphology were highly dependent on the differentiation medium (**Figure 1B**). Importantly, we did not observe any effect on undifferentiated NSC-34 cells as previously demonstrated (Cashman et al., 1992; Maier et al., 2013). We also evaluated the RA effects on the expression of genes encoding certain specific motor neuron proteins and glutamate receptor subunits, and we found that RAdifferentiated NSC-34 did not express more receptor subunits than NSC-34<sup>D</sup> without RA (**Figures 2C–H** and Table A3, see Supplementary Material), thus suggesting that using RA is not required in NSC-34 differentiation processes in the context of excitotoxicity studies. This conclusion is supported by the results obtained by Cheung et al. (2009) who showed that RA differentiation conferred higher tolerance of SH-SY5Y cells to neurotoxin (6-hydroxydopamine) compared to undifferentiated SH-SY5Y cells, and they suggested that RA-differentiated SH-SY5Y was not appropriate for neurotoxicity or neuroprotection studies.

## The Receptor Subunits Expressed by NSC-34<sup>D</sup>

NMDARs and AMPARs have critical roles in excitatory synaptic transmission, plasticity and excitotoxicity in the CNS. We found that NSC-34<sup>D</sup> expressed more receptor subunits in DMEM/Ham's F12 medium without RA than in other media (**Figures 2C–H** and Table A3, see Supplementary Material). NSC-34<sup>D</sup> expressed all AMPAR (GluA1–4) and some NMDAR (GluN1, GluN2A/B) subunits, as reported in the literature (Eggett et al., 2000; Rembach et al., 2004). We therefore chose DMEM/Ham's F12 medium without RA for further experiments. Then we compared expression of glutamate receptor subunits on primary MN and NSC-34<sup>D</sup> using RT-qPCR (**Figures 2I,J**). The results showed that primary MN expressed all glutamate receptor subunits at different levels, in contrast to NSC-34<sup>D</sup> which expressed some of them at times and in very small quantities. This clearly showed that the transcript profile of NSC-34<sup>D</sup> glutamate receptor subunits was different from that of MN and very consistent with RT-PCR and RT-qPCR data, indicating that NSC-34<sup>D</sup> could not be used as a model of primary MN for glutamate-mediated excitotoxicity.

## No Glutamate-Induced Toxicity in NSC-34<sup>D</sup>

Glutamate concentrations in the extracellular fluid are normally around 0.8–2.9 µM (Lerma et al., 1986). A rise in the extracellular glutamate concentration to 2–5 µM is considered sufficient to cause degeneration of neurons through excessive stimulation of glutamate receptors (Rosenberg et al., 1992). When investigating glutamate-induced excitotoxicity at the cellular level, the glutamate concentrations used to obtain effects varied according to the in vitro models investigated and exposure times. For primary cultures, the range of glutamate concentrations used to act on ionotropic glutamate receptors (Van Den Bosch et al., 2000) was between 10 µM–1 mM, and cultures were exposed for less than 24 h (Mattson et al., 1995; Kruman et al., 1999; Sen et al., 2008). As in our study (**Figure 3B**), it had previously been demonstrated that exposure of primary embryonic MN to glutamate (0.1 mM) for 24 or 48 h led to approximately 50% cell death (Metzger et al., 1998; Urushitani et al., 1998), whereas for cell lines, this glutamate concentration range was increased to 10 mM with longer exposure times (24–72 h; Eggett et al., 2000; Rembach et al., 2004; Benkler et al., 2013; Maier et al., 2013). In our study, we investigated the toxicity of glutamate on NSC-34<sup>D</sup> and motor neuron survival. The results showed approximately 50% cell death at 100 µM glutamate for primary MN (**Figure 3B**), as described in the literature (Metzger et al., 1998; Urushitani et al., 1998). For NSC-34<sup>D</sup> cells, we found a significant loss of viable cells at a concentration of only 10 mM glutamate (**Figure 3A**). This result is consistent with previous excitotoxicity studies on NSC-34<sup>D</sup> where treatment with glutamate (up to 1 mM) failed to induce vacuolation of NSC-34D, in contrast to the effects on primary MN (Durham et al., 1993).

Among the different neuroprotective mechanisms in neurons, neurotrophin factors may play a key role. Literature reported that NTFs and glutamate interact to regulate developmental and adult neuroplasticity (Mattson et al., 1995; Mattson, 2008). Consequently, neurotrophin factors attenuated elevation of intracellular calcium concentrations induced by glutamate (Mattson et al., 1995), thus promoting the survival of neurons (McKay et al., 1996; Giménez y Ribotta et al., 1997; Vincent et al., 2004; Mattson, 2008) via the MAPK and PI-3K/Akt pathways (Vincent et al., 2004).

To validate the glutamate-sensitive motor neuron NSC-34 cell line, many studies have used modified growth conditions through serum depletion to induce differentiated NSC-34 cells and to investigate the cytotoxicity of glutamate or other excitotoxins (AMPA, 5-fluorowillardiine, H2O2, TNF-α, etc.) in the presence or absence of receptor antagonists by using viability tests (Rembach et al., 2004; Hemendinger et al., 2012; Benkler et al., 2013; Maier et al., 2013). In some cases, they evaluated the expression of certain genes which encode cholinergic phenotyperelated proteins (ChAT, acetylcholine esterase and vesicular acetylcholine transferase; Maier et al., 2013). It should be emphasized that most studies that used cell line models to investigate glutamate neurotoxicity focused predominantly on its toxic action and not on the functional properties of glutamatergic receptors such as Ca2<sup>+</sup> permeability (Rembach et al., 2004; Hemendinger et al., 2012; Benkler et al., 2013; Maier et al., 2013).

## No Sustained Calcium Entry in NSC-34<sup>D</sup>

The stoichiometry of the receptor complexes in mammalian cells seems to be largely controlled by the level of the individual subunits expression that determine their functional properties including Ca2<sup>+</sup> permeability via glutamate receptors (Hollmann et al., 1991; Monyer et al., 1992; Das et al., 1998; Dingledine et al., 1999; Sobolevsky et al., 2009). According to the literature evidence, the expression of GluA1–4, GluN1 and GluN2A or GluN2B subunits induces functional receptors. Surprisingly, in our study, despite the expression of GluA, GluN1 and GluN2A/D subunits in NSC-34<sup>D</sup> (**Figure 2I**), we did not observe a significant sustained calcium entry as demonstrated in MN (**Figure 4C**). In this study, we globally evaluated the intracellular calcium concentrations (including the interplay between extracellular and intracellular origin) provoked by glutamate, through glutamate receptors by using the Fura-2-AM. However, we cannot exclude the possibility that part of the increase of cytosolic calcium could come from internal stores. Using SH-SY5Y cells line and measurement of calcium release from different compartments, Jaiswal et al. (2009) provided evidences of the existence of two separate intracellular Ca2<sup>+</sup> stores (endoplasmic reticulum (ER) and mitochondrial intracellular pools) which may participate in the generation of intracellular Ca2<sup>+</sup> signals. Calcium permeability via glutamate receptors has also been shown to initiate a self-perpetuating process of intracellular Ca2<sup>+</sup> dysregulation with consecutive ER Ca2<sup>+</sup> release and mitochondrial Ca2<sup>+</sup> overload (Jahn et al., 2006; Grosskreutz et al., 2007, 2010). The ER and mitochondria form a highly dynamic interconnected network that is involved in the generation of Ca2<sup>+</sup> signals (Tadic et al., 2014). A recent article reported a Ca2<sup>+</sup> influx in NSC-34<sup>D</sup> cells compared to a ''negative control'' undifferentiated NSC-34 (Liu et al., 2015), but authors did not compare calcium influx of these NSC-34<sup>D</sup> to that of MN (as ''positive control'') which are the parent motor neuron for NSC-34, as performed here in our study. We postulate that this shift of Ca2<sup>+</sup> entry in NSC-34<sup>D</sup> could be due to the fact that some glutamate receptor subunits were not expressed or very weakly expressed, in contrast to MN which showed a sustained calcium entry following glutamate application.

### Limitations in the Use of the NSC-34 Cell Line

As mentioned above, Eggett et al. (2000) evaluated the use of NSC-34 cells as a glutamate-sensitive motor neuron model. Using immunocytochemistry, they demonstrated the presence of glutamate receptor proteins GluN1, GluN2A/B, GluA1–4, GluK2/3 and GluK5. Exposure to glutamate (1 mM) for 24 h induced significant cell death (∼30%), and changes in calcium cytosolic levels were observed. As in many other studies (Eggett et al., 2000; Rembach et al., 2004; Benkler et al., 2013; Maier et al., 2013; Liu et al., 2015) authors suggested that, because of their motor neuron origin, the NSC-34 cell line could be used to investigate excitotoxicity mechanisms. However, as reported by Durham et al. (1993), in our study we found limitations in the use of the NSC-34 cell line for neurotoxicity testing. Durham et al. (1993) have evaluated the value of these cells by following exposure of cultures to chemicals known to be neurotoxic for MN. Authors showed that NSC-34 responded to agents that affect voltage-gated ion channels, cytoskeleton organization and axonal transport. However, no electrophysiological evidence was shown, and exposure to glutamate (1 mM) had no effect on cell morphology or potential production. They concluded that the NSC-34 cell line was not a good model to investigate agents that affect synaptic transmission (Durham et al., 1993). Hemendinger et al. (2012) have demonstrated that riluzole in the neurorescue paradigm was unable to reduce cell death induced by neurotoxins that increased intracellular calcium levels independently of ER stress in NSC-34<sup>D</sup> cells, but they did not provide information on the molecular basis. This finding may potentially be explained by the fact that NSC-34 is not a suitable in vitro model to investigate excitotoxicity. All these results together suggest that NSC-34 cells may not be appropriate as an in vitro model to study glutamatergic toxicity in motor neuron degeneration. Motor neuronal properties evaluated on differentiated NSC-34 cells and literature findings were summarized in the Appendix (Table A4, see Supplementary Material).

Importantly, the phenotypes of NSC-34 may vary, according to the repeated culture passage or to the origin of the cell line [provided by Dr Neil Cashman (Durham et al., 1993; Maier et al., 2013; Liu et al., 2015) or purchased (Hemendinger et al., 2012)]. So, results may not be reproducible over time, from one laboratory to another, and even within the same laboratory. We cannot exclude the possibility that this contradictory findings may be explained by the differences between basal features of NSC-34 cells, rarely detailed in the studies (Eggett et al., 2000; Rembach et al., 2004; Benkler et al., 2013). There may be various differentiation processes with various media and chemicals known to induce various NSC-34-expressed motor neuron properties, and the culture conditions were not systematically explored to seek optimal toxicity to glutamate. However, in agreement with the literature, in our study we did not find any condition which promoted expression of all glutamatergic properties. The aim of this work was not to focus on optimization of differentiation processes which should involve several parameters (medium composition, differentiation time, time of changing medium, cell density, cell passage, etc.) but rather to evaluate the versatility of the NSC-34 cell line to model motor neuron susceptibility to glutamate-induced excitotoxicity. Reliable validation is a crucial step before using a cell line as an in vitro model in glutamate-induced excitotoxicity studies. Such validation requires the evaluation of the biochemical and electrophysiological processes of MN that are targets of glutamate such as glutamate receptors and Ca2<sup>+</sup> influx with a ''positive control''.

We could not exclude the possibility that the maturity was insufficient to induce calcium influx in NSC-34<sup>D</sup> cells under the present experimental conditions. Even if organotypic spinal cord cultures remain the models that closely reflect in situ reality, we suggest that the use of primary motor neuron culture is more suitable than NCS-34 cell line to explore the pathogenesis of glutamate-mediated excitotoxicity at the cellular level, in ALS

### REFERENCES


and other motor neuron diseases. However, different authors have shown that NSC-34 cells remain a suitable model of vulnerability to neurotoxins (Maier et al., 2013) and to investigate the effects of cerebrolysin (Keilhoff et al., 2014). NSC-34 cells line is a viable in vitro cell model for screening therapeutic candidates against nerve agents (Kanjilal et al., 2014), investigating the expression of transcription factors (Prell et al., 2012) and could thus be a relevant model for some specific experiments.

### AUTHOR CONTRIBUTIONS

BMH, RF, FP and MG conducted all experiments, analyzed the data and BMH wrote the manuscript. HB, SM, PV, CR and CRA designed the project, analyzed the data and wrote the manuscript. All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This work was supported by the ''Institut National de la Santé et de la Recherche'' INSERM, the University François-Rabelais de Tours, and ARSLA (''Association pour la Recherche sur la Sclérose Latérale Amyotrophique et autres maladies du motoneurone''). BMH was supported by ''La Région Centre'' with a PhD graduate grant.

### ACKNOWLEDGMENTS

We thank the staff of the PPF ASB platform of the Université François-Rabelais de Tours, France, and Catherine Cherpi-Antar for technical assistance.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2016.00118/abstract


aggregates in cultured motor neurons and reduces cell viability. Amyotroph. Lateral Scler. Frontotemporal Degener. 16, 131–134. doi: 10.3109/21678421. 2014.965179


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Madji Hounoum, Vourc'h, Felix, Corcia, Patin, Guéguinou, Potier-Cartereau, Vandier, Raoul, Andres, Mavel and Blasco. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Adducin at the Neuromuscular Junction in Amyotrophic Lateral Sclerosis: Hanging on for Dear Life

Charles Krieger <sup>1</sup> \*, Simon Ji Hau Wang1,2 , Soo Hyun Yoo1,2 and Nicholas Harden<sup>2</sup>

<sup>1</sup> Department of Biomedical Physiology and Kinesiology, Simon Fraser University, Burnaby, BC, Canada, <sup>2</sup> Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, BC, Canada

The neurological dysfunction in amyotrophic lateral sclerosis (ALS)/motor neurone disease (MND) is associated with defective nerve-muscle contacts early in the disease suggesting that perturbations of cell adhesion molecules (CAMs) linking the pre- and post-synaptic components of the neuromuscular junction (NMJ) are involved. To search for candidate proteins implicated in this degenerative process, researchers have studied the Drosophila larval NMJ and find that the cytoskeleton-associated protein, adducin, is ideally placed to regulate synaptic contacts. By controlling the levels of synaptic proteins, adducin can de-stabilize synaptic contacts. Interestingly, elevated levels of phosphorylated adducin have been reported in ALS patients and in a mouse model of the disease. Adducin is regulated by phosphorylation through protein kinase C (PKC), some isoforms of which exhibit Ca<sup>2</sup><sup>+</sup>-dependence, raising the possibility that changes in intracellular Ca<sup>2</sup><sup>+</sup> might alter PKC activation and secondarily influence adducin phosphorylation. Furthermore, adducin has interactions with the alpha subunit of the Na<sup>+</sup>/K<sup>+</sup>-ATPase. Thus, the phosphorylation of adducin may secondarily influence synaptic stability at the NMJ and so influence pre- and post-synaptic integrity at the NMJ in ALS.

Keywords: amyotrophic lateral sclerosis, neuromuscular junction, adducin, Hu-li tai shao, Discs large, neural cell adhesion molecule

### THE NEURODEGENERATIVE DISEASE AMYOTROPHIC LATERAL SCLEROSIS

The neurodegenerative disorder amyotrophic lateral sclerosis (ALS), also known as motor neurone disease (MND), is a motor system disease that causes progressive motoneuron loss in the spinal cord and brainstem leading to weakness and loss of muscle innervation (i.e., denervation), as well as the degeneration of descending motor tracts from the brain and subcortical structures resulting in spasticity (Eisen and Krieger, 2006; Su et al., 2014). Death typically occurs within 2–5 years of diagnosis, usually resulting from respiratory failure due to respiratory muscle weakness. At post-mortem, ALS patients have lost large numbers of spinal motoneurons, interneurons and other neuronal populations, but with considerable side-to-side asymmetry and variability at spinal and brain stem levels (Tsukagoshi et al., 1979; Swash et al., 1986). These findings are, by their very nature, end-stage effects and do not reflect the early stages of the disease. The large majority of ALS cases appear sporadically in the general population without evidence for an inherited gene mutation (''sporadic ALS'').

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Barbara Bardoni, Centre National de la Recherche Scientifique, Institut de Pharmacologie Moléculaire et Cellulaire, France Michael E. Hildebrand, Carleton University, Canada Maria Piotrkiewicz, Nalecz Institute of Biocybernetics and Biomedical Engineering, Polish Academy of Sciences, Poland

\*Correspondence:

Charles Krieger ckrieger@sfu.ca

Received: 26 October 2015 Accepted: 13 January 2016 Published: 29 January 2016

#### Citation:

Krieger C, Wang SJH, Yoo SH and Harden N (2016) Adducin at the Neuromuscular Junction in Amyotrophic Lateral Sclerosis: Hanging on for Dear Life. Front. Cell. Neurosci. 10:11. doi: 10.3389/fncel.2016.00011 However, genes associated with ALS susceptibility are being defined (Keller et al., 2014) and mutations in various genes are sometimes found in patients with sporadic ALS. In 5–10% of ALS patients, the disease has an autosomal dominant or, in rare cases, autosomal recessive inheritance, and is termed familial ALS (FALS; Andersen and Al-Chalabi, 2011). Gene mutations associated with FALS have been under intense scrutiny, and such mutations are currently identified in approximately 50% of FALS patients. Observed in the highest percentage of FALS patients are mutations in Cu/Zn superoxide dismutase 1 (SOD1) and chromosome 9 open reading frame 72 (C9ORF72; Vucic et al., 2014). However, mutations in TAR DNA-binding protein 43 (TDP-43, TARDBP), fused in sarcoma/translocated in liposarcoma (FUS/TLS), vesicle-associated membrane proteinassociated protein B (VAPB) and other genes have been reported (Andersen and Al-Chalabi, 2011). The functional diversity of the proteins associated with these gene mutations (e.g., SOD1, a cytosolic anti-oxidant; VAPB, a synaptic membrane-associated protein; and others) makes it unlikely that a single diseaseinitiating event will be present, but rather that a number of unique triggering events will result in a common pathogenic process leading to muscle denervation. Mutations in genes for isoforms of the cytoskeletal protein adducin have not been reported in FALS patients nor have mutations been found in genes for proteins which interact with adducin. Rarely, missense mutations are found in genes for synaptic proteins such as VAPB, both in FALS and in sporadic ALS (Millecamps et al., 2010). Although the expression of mutant VAPB in Drosophila melanogaster is associated with neuromuscular junction (NMJ) defects such as increased numbers of synaptic boutons at the NMJ and decreased bouton size, as well as locomotor defects, VAPB has numerous effects in neurons and it is not possible to conclude that VAPB mutations produce a phenotype that results exclusively from synaptic dysfunction (Sanhueza et al., 2014).

There are some differences in the clinical features of FALS due to distinct mutations in comparison to sporadic ALS (Stewart et al., 2012). However, in the absence of a family history, it is virtually impossible to distinguish FALS patients from those with sporadic ALS, suggesting that sporadic and FALS are mechanistically similar and that insights into FALS will be relevant for an understanding of sporadic ALS. Although it is clear that specific gene mutations result in ALS, it is not clear how these mutations result in disease. In this review, we will focus on the evidence that ALS results from synaptic dysfunction at the NMJ as well as at other synapses. We recognize that there is also evidence for other explanations as causative for ALS. We will also highlight the importance of cytoskeletal proteins such as adducin as being a modulator for the synaptic dysfunction in ALS.

### SELECTIVE VULNERABILITY OF MOTONEURONS IN ALS

It is likely that ALS develops as a consequence of the selective vulnerability of motoneurons or the axonal pathways from the brain and brainstem that descend to motoneurons to undergo dysfunction and cell death. The descending motor tracts that innervate motoneurons include the corticospinal tract and other descending motor tracts from the brain and brainstem. The vulnerability of motoneurons is produced by the large size of these cells, their large caliber and long axons, the abundant dendritic projections, and the high metabolic cost of action potential generation in motoneurons (Le Masson et al., 2014). Additionally, there is possibly a toxic environment surrounding motoneurons, including the presence of microglia and astrocytes that might be responsible for non-cell autonomous toxicity (Julien, 2007; Ikiz et al., 2015). Furthermore, motoneurons exhibit an abundance of excitatory amino acid (EAA) receptors including N-methyl-D-aspartate (NMDA) receptors, and some of these receptor channels have calcium permeability. Stimulation of EAA and/or NMDA receptors can lead to excitotoxicity which is likely accompanied by raised intracellular Ca2<sup>+</sup> concentrations ([Ca2+]i) in the cell body, but alternatively could occur in the post-synaptic region or pre-synaptic terminal. Motoneurons have lower levels of calcium buffering proteins, such as calbindin, than do other neuronal types, thus possibly enhancing the toxicity of elevated [Ca2+]<sup>i</sup> . Because of the vulnerability of the motoneuron to injury, previous hypotheses of neuronal dysfunction in ALS have focused on excitotoxicity, axonal dysfunction, or aberrant RNA processing in motoneurons. Elevations of [Ca2+]<sup>i</sup> are capable of activating Ca2+-dependent protein kinases such as some isoforms of protein kinase C (PKC). One consequence of PKC activation is the increased phosphorylation of cytoskeletal proteins such as adducin (Krieger et al., 2003). As described below, adducin is present at the synapse both presynaptically and postsynaptically and exists in phosphorylated and nonphosphorylated forms. Non-phosphorylated adducin links actin and spectrin filaments and provides stability to the NMJ. When adducin is phosphorylated by PKC or other protein kinase adducin will translocate from the membrane, where it likely stabilizes the synapse, to the cytosol. This will result in a de-stabilization of the synapse which can be important for synaptic plasticity such as in learning but may reduce the ability of the NMJ to form strong connections between pre-synaptic and post-synaptic contacts. Thus indirectly, the elevations of [Ca2+]<sup>i</sup> lead to destabilization of the synapse, especially at the NMJ, mediated in part through adducin. We recognize that this view is simplistic, as adducin effects are likely different at pre- and post-synaptic regions (Pielage et al., 2011; Wang et al., 2011, 2014) and that elevations in [Ca2+]<sup>i</sup> will have many effects. Nonetheless, it is the purpose of this review to highlight the potential role played by cytoskeletal proteins in general and adducin in particular as mediators for some of the reactions of motoneurons to cell stress including impaired synaptic connectivity.

Also important is that the motoneuron is surrounded by non-neuronal cells such as astrocytes, microglia and terminal Schwann cells which when activated could result in motoneuron damage (Julien, 2007). Evidence for non-cell autonomous damage to motoneurons in the pathogenesis of ALS is strong given studies in transgenic mice that neuron-specific expression of mutant proteins generally does not result in much motoneuron loss whereas ubiquitous expression of mutant proteins will produce profound motoneuron dysfunction (Ilieva et al., 2009; Ikiz et al., 2015). Although not well studied, it is likely that the effects of adducin phosphorylation in non-neuronal cells will be different from the effects of phosphorylated adducin at the NMJ.

## ALS INITIALLY AFFECTS THE DISTAL MOTONEURON AXON AND NEUROMUSCULAR SYNAPSE

Attempts to delineate early changes in ALS, as opposed to those at end-stages, have focused on biopsy material from ALS patients and controls, as well as studies in widely used animal models of ALS such as transgenic rodents over-expressing human mutant SOD (mSOD), which manifest motoneuron degeneration similar to that seen in ALS patients (Turner and Talbot, 2008). Studies involving mSOD mice have shown that motoneuron loss is prominent but not complete such that approximately 50% of motoneurons are still evident even in animals in the terminal stages of disease (Parkhouse et al., 2008; Turner and Talbot, 2008). These observations demonstrate that motoneurons and their proximal axons are still present in the late stages of the disease, but that many of the axons are incapable of forming functional NMJs with skeletal muscle (i.e., denervation) leading to muscle weakness. Furthermore, these axons are often unable to re-establish stable contacts with muscle once denervation has occurred (i.e., impaired re-innervation). This view is supported by evidence that motoneurons in mSOD mice with advanced disease can still be identified using morphological or immunohistochemical techniques, yet, they will not manifest labeling with retrograde markers such as FluoroGold delivered to the muscle, which would normally be taken up by the presynaptic terminals and transported in the retrograde direction to the cell bodies (Parkhouse et al., 2008). This is presumably due to loss of axon integrity at either the proximal or distal axon (Kennel et al., 1996). The capacity for compensatory axonal sprouting is strikingly reduced in ALS, even when compared to another motor neuron disease such as poliomyelitis. Interestingly in patients who develop poliomyelitis and have considerable motoneuron loss, the few surviving motoneurons are able to re-innervate muscle and compensate for the reduction in motoneuron number even when the motoneuron losses are as extensive as in some patients with ALS (Trojan and Cashman, 2005). Several studies that examined biopsies obtained from patients with ALS early in the disease have confirmed that muscle denervation is extensive, and develops prior to significant motoneuron loss (Fischer et al., 2004). The impairment in the ability of motoneurons to re-innervate muscle may also be the case in normal aging, but to a much lesser extent (Valdez et al., 2012). The degenerative process of ALS may initially lead to impaired function of the distal axon (so called ''axonopathy''), with relative preservation of the cell body and proximal axon, thus offering hope that if axonal function can be stabilized, then the axonal loss can be halted or even reversed and the disease progression can be attenuated. The role of axonal transport in neurodegenerative disease has been reviewed previously (De Vos et al., 2008).

The predominant features of ALS at the motoneuron, as distinct from axonal loss in the central nervous system (CNS), are weakness and fasciculations. Fasciculations are brief, spontaneous contractions of a limited number of muscle fibres which may arise from the spontaneous firing of motoneurons or their axons. Motor weakness is not exclusively found in ALS and can be present in many conditions due to impaired functioning of different regions of the motoneuron. For instance, some forms of neuropathy target the axon of the motoneuron (''axonopathy''). Some types of axonopathy have been referred to as a ''dying back'' process, where the longest and largest motor axons are targeted first, and the initial impairment affects the most distal portions of the axon (also called ''distal axonopathy''), such as at the pre-synaptic terminal. Subsequently, more proximal portions of the axon become affected in the disorder. This view is based on neuropathological evaluation of some types of toxin-induced peripheral nerve damage (peripheral neuropathy) such as that following extensive alcohol consumption, where the fine intramuscular nerve branches of motoneurons may be affected initially while more proximal axons are spared. Presumably, the most distal aspects of the nerve are more susceptible to some toxins or to axonal ''undernourishment'', or impaired axon transport associated with the actions of a toxin (Cavanagh, 1979). An additional clinical condition resulting in weakness that is likely due to impaired distal axon and NMJ function are some forms of critical illness polyneuropathy (Hermans and Van den Berghe, 2015). This evidence collectively supports the view that a loss of nerve-muscle contacts occurs early in the disease, which might represent the inability of presynaptic motoneuron terminals to re-innervate muscle at the NMJ (Moloney et al., 2014). However, it also should be kept in mind that transgenic mice with neuron-specific expression of mSOD do not develop disease with the same characteristics as with ubiquitous expression of mSOD, raising the likelihood that ALS is a disease not only of neurons, but also neurons and their non-neuronal neighbors (Julien, 2007; Ilieva et al., 2009; Meyer et al., 2014). These types of studies have not suggested that muscle cells were a cell type that is responsible for development of murine or human ALS (Miller et al., 2006).

Numerous factors are responsible for nerve-muscle interactions, both during development and at the mature synapse. The integrity of cell adhesion molecules (CAMs) that link the pre- and post-synaptic regions is undoubtedly relevant for the functionality of neuromuscular control. However, the link between structure and function at the NMJ is complex with numerous interactions between NMJ components and considerable functional redundancy for proteins at the NMJ (Koper et al., 2012). There are many evolutionarily conserved synaptic CAMs such as cadherins, integrins, neurexins and others. Physiological activation of the NMJ also has profound effects where pre- or post-synaptic cholinergic antagonism with botulinum toxin, or bungarotoxin, have similar effects to surgical denervation (Avila et al., 1989). It is also clear that perisynaptic Schwann cells (also called ''terminal Schwann cells'') exert important roles in synapse re-organization at the NMJ in ALS (Arbour et al., 2015). As will be discussed below there is evidence that one of these CAMs such as a Drosophila orthologue of neural cell adhesion molecule (N-CAM) interact indirectly with synaptic adducin.

Mammalian motoneurons have different physical and firing characteristics that can be roughly grouped into distinct classes, such as fast-fatigable (FF) and slow (S; Burke, 1967; Heckman and Enoka, 2012). Interestingly, there are differences in the susceptibility of various synaptic subtypes to degeneration in ALS, with synapses of FF motoneurons being vulnerable to synaptic loss and synapses of slow motoneurons being relatively resistant (Frey et al., 2000; Pun et al., 2006; Hegedus et al., 2008). In addition, with disease there may be an activity-dependent conversion of motoneuron types towards a slow phenotype (Frey et al., 2000). The progressive weakening of specific synapses with disease suggests synapse-specific mechanisms as being at least partially responsible for selective synaptic loss. This view is supported to some degree by the observation that the WldS gene, which protects against axonal injury, modestly prolongs survival in the mSOD mouse and other murine models of ALS (Ferri et al., 2003; Fischer et al., 2005). Synaptic terminals that are less susceptible to denervation also demonstrate more ability to generate stimulus-induced synaptic sprouting, which could arise from differences in the regulation of the actin cytoskeleton of the different synaptic types (Laux et al., 2000).

One additional piece of circumstantial evidence for the presence of distal axon or synaptic terminal dysfunction in ALS is the finding of decremental responses in muscle to repetitive nerve stimulation both clinically in ALS patients and in models of ALS (Kennel et al., 1996; Iwanami et al., 2011; Piotrkiewicz and Hausmanowa-Petrusewicz, 2013). That is, when motoneuron axons are stimulated repetitively, there is a reduction in the amplitude of the evoked response in muscle suggesting an impairment of neuromuscular transmission at the NMJ. ALS also includes involvement of CNS structures and it has not yet been determined whether involvement of the motoneuron precedes dysfunction in descending axons, or if the two processes occur concurrently, or in the reverse order (Dengler, 2011).

### MITOCHONDRIAL DYSFUNCTION IN ALS

Mitochondrial dysfunction has been linked with many neurodegenerative conditions, including ALS (Schon and Przedborski, 2011; Cozzolino et al., 2015). However, it is not clear how mitochondrial dysfunction arises in ALS and whether the impairment is an initiating feature of the disease, or a consequence of the motoneuron dysregulation that arises from other causes. Impaired mitochondrial function likely leads to defective bioenergetics, but could also produce defective mitochondrial trafficking, impaired mitochondrial fusion and fission, as well as altered mitochondrial quality control (Schon and Przedborski, 2011). For instance, mitochondrial disruption has been reported to contribute to SOD1-mediated motoneuron degeneration (Jaiswal and Keller, 2009). Motoneurons are large, actively firing cells having very demanding ATP-dependence to maintain the resting membrane potential, neurotransmitter release and other functions. Of relevance to this review are observations that the cytoskeletal protein adducin is associated with and directly interacts with the α-subunit of the Na+/K+- ATPase (Ferrandi et al., 1999; Torielli et al., 2008). Mutations in α-adducin are associated with hypertension in rats and humans which is presumably based on aberrant interactions between the mutant α-adducin and Na+/K+-ATPase (Torielli et al., 2008). Although it is beyond the scope of this review to discuss mitochondrial function in ALS, it is relevant that a major energy consumer in the neuron is the Na+/K+-ATPase which is essential for providing energy for neuronal repolarization following an action potential. Recent modelling studies of ATP utilization in model neurons have suggested that the amount of ATP required for ion homeostasis is closely matched by energy production (Le Masson et al., 2014). Under conditions where motoneurons produce high frequency action potential firing, especially in fast fatiguable motoneurons, reductions in steady state ATP levels can occur. Modelling studies of action potential firing in excess of 20 Hz suggest that motoneurons cannot support continued Na+/K+-ATPase activity and neurons will depolarize (Le Masson et al., 2014). This also might lead to aberrant action potential firing, such as is seen clinically in the appearance of ''fasciculations'', the spontaneous discharges of motor units that may be related to abnormal motoneuron discharge (de Carvalho et al., 2014). Furthermore, modelling studies also suggest that the energetic cost to support the vulnerable, ''FF'' motoneurons is higher than for the ''S'' motoneurons providing a possible explanation for the differential vulnerability of the two motoneuron subpopulations as is seen in ALS (Le Masson et al., 2014). Fasciculations arise from spontaneous motoneuron firing at the cell body, axon or other sites (de Carvalho et al., 2014).

### COULD MOLECULES MEDIATING CELL ADHESION BE RELEVANT FOR THE PATHOGENESIS OF ALS?

One approach to determine which molecules could be relevant for the pathogenesis of ALS would be to compare the biochemical profiles of synapses that are retained in ALS with those that are lost. This can be attempted by comparing NMJs from biopsy material from the limb muscles that are typically involved in ALS, with NMJs in extraocular muscles, which usually are not involved in ALS. One such study has claimed that significant reductions in the immunoreactivity of the extracellular matrix protein, α4 laminin, are seen in the NMJs of limb muscles in ALS patients, but are not as affected in the NMJs of extraocular muscles (Liu et al., 2011). It is possible that these immunohistochemical findings are related to the disruption of the NMJ, rather than being causative, but the results suggest that the integrity of the basement membrane surrounding the muscle fibre may be important for synapse stability. Interestingly, removal of Laminin A in Drosophila embryos causes a reduction in NMJ adhesion, and it has been proposed that the basement membrane contributes to adhesion of the motoneuron terminal to the muscle (Koper et al., 2012).

A potentially relevant CAM for ALS is the N-CAM and its homologue in the fly, Fasciclin 2 (Fas2). Mice lacking N-CAM in presynaptic terminals have impaired muscle re-innervation following denervation, emphasizing the important role played by N-CAM for re-innervation (Chipman et al., 2014a). It is also well established that at the simpler Drosophila NMJ, Fas2 is necessary for synaptic stabilization and plasticity (Schuster et al., 1996a,b). Furthermore, the proportion of denervated muscle fibres in transgenic mice overexpressing mSOD as identified using N-CAM immunoreactivity correlates well with the rate of decline of muscle contractile force in these mice (Gordon et al., 2009). When pluripotent stem cells are directed to differentiate into motoneurons, motoneuron stem cells lacking N-CAM have well defined structural and functional deficits at the NMJ (Chipman et al., 2014b). These deficits may be less pronounced than when N-CAM is lost from both pre- and post-synaptic sites. The mechanisms underlying the actions of N-CAM and Fas2 in modulating re-innervation at the NMJ are under investigation. However, it would seem important to study proteins that could have an impact on these CAMs and that these molecules deserve further scrutiny in ALS. In the following section we focus on a cytoskeletal protein that interacts with many other proteins at the NMJ and is likely relevant for understanding the pathophysiology of ALS.

### ADDUCIN AT THE NMJ

### The Distal Axon and Synapse are Regulated by the Actin Cytoskeleton

Actin and spectrin are cytoskeletal proteins whose interactions are critical for the morphology of the presynaptic terminal and the post-synaptic region. The association of these proteins is relatively weak, and spectrin requires accessory proteins to assemble and stabilize spectrin-actin interactions, which include the cytoskeletal accessory proteins adducin and protein 4.1 (Bennett and Baines, 2001). These interactions permit the cytoskeleton to provide several functions that have been clarified through study of the erythrocyte membrane: they allow structural support through the lattice-like arrangement of the fibrils of filamentous actin and spectrin; they provide a platform on which many accessory proteins can bind; and through the variation in the lattice sizes of actin and spectrin, they can likely modulate the structural strength of local regions of the membrane. Adducin is a tetramer composed of α, β and γ subunits, where interactions with actin and spectrin are mediated by the carboxy-terminal region of adducin that has homology to a myristoylated alaninerich C-kinase substrate (MARCKS)-related domain (Matsuoka et al., 2000). This highly basic, charged MARCKS domain likely permits adducin to associate (i.e., ''hang on'') to the cell membrane, and sequester highly charged phosphoinositides (PIs) such as PI(4,5)P<sup>2</sup> (Denisov et al., 1998; **Figure 1**). Adducin also associates with the sides and fast-growing ends of actin filaments, and exhibits high affinity for actin-spectrin complexes (Matsuoka et al., 2000; **Figure 2**). In this sense, adducin ''hangs on'' to the cytoskeleton and thus serves as a linker between the cell membrane and the cytoskeleton. The MARCKS region contains a phosphorylation site that is modulated by Ca2+/calmodulin and PKC, where the phosphorylation of adducin inhibits the stabilization of actin-spectrin complexes (Matsuoka et al., 2000; **Figure 2**). Interestingly, mammalian adducin and its invertebrate homologues have a number of defined synaptic protein-protein interactions including Na+/K+-ATPase, Discs large (Dlg) and Golden Goal (Gogo; Ohler et al., 2011; Wang et al., 2011, 2014; Gallardo et al., 2014; **Figure 2**).

### A Role for Adducin and Phosphorylated Adducin in ALS

Adducin is of interest in ALS as there are high levels of phosphorylated adducin (phospho-adducin) protein in spinal cord tissue from patients who died with ALS compared to individuals who died without neurological disease (Hu et al., 2003). Furthermore, immunocytochemistry of murine spinal cord tissue demonstrated that phospho-adducin immunoreactivity was significantly higher in mSOD spinal cord tissue compared to control tissue (Shan et al., 2005). Recently, both of these observations have been confirmed and the association between adducin and ALS extended (Gallardo et al., 2014). It is possible that excessive levels of phosphorylated adducin would create instability at the NMJ and knockdown of α-adducin by RNA interference (RNAi) in mSOD overexpressing mice results in greater motoneuron survival than in the absence of RNAi. The effects of elevations in phosphorylated adducin are likely not restricted to motoneurons as knockdown of α-adducin in the astrocyte population of co-cultured wild-type motoneurons with astrocytes over-expressing mSOD resulted in less motoneuron death. Gallardo et al. (2014) suggest that α-adducin mediates motoneuron degeneration in mSOD mice and potentially in human ALS as well. Evidence supporting involvement of adducin in ALS is shown in **Table 1**. Both we and Gallardo et al. (2014) have expressed the view that the effects of adducin in ALS are not mediated entirely by adducin, but are likely dependent on interactions between adducin and other proteins. Gallardo et al. (2014) have focused on interactions between α-adducin and the Na+/K+-ATPase, a protein that has been well-established to associate with adducin where this interaction has been of considerable interest with regard to the role of adducin polymorphisms in essential hypertension (i.e., high blood pressure). To our knowledge, the possible relevance of polymorphisms in α-adducin genes for the development of ALS has not been evaluated.

The work of Gallardo et al. (2014) indicates that a gain of adducin function may have a causative role in ALS, especially in astrocytes, whereas a recent study from Costessi et al. (2014) proposes that a loss of adducin function may be involved. TDP-43 is an RNA-binding protein found to be a major component of pathological cytoplasmic inclusions in ALS and as such, perturbed TDP-43 function may be linked to neurodegeneration. Costessi et al. (2014) demonstrated that

TDP-43 regulates adducin 2 gene expression by promoting adducin 2 mRNA stability, and suggest that loss of adducin function through disruption of TDP-43 activity may contribute to ALS.

These various studies indicate that misregulation of adducin activity through hyper-phosphorylation or changes in gene expression may be a contributing factor in the etiology of ALS. An alternative view could be that as many proteins are hyper-phosphorylated in ALS, adducin phosphorylation may be just one of many hyper-phosphorylated proteins in the disease (Krieger et al., 2003), and of little direct relevance. However, although not a ''primary cause'' for ALS, it is also possible that the hyper-phosphorylation or other misregulation of adducin is responsible for some of the particularly malicious aspects of ALS, namely the inability of denervated presynaptic terminals to re-innervate muscle and sustain life.

### Two Functions of Adducin at the Drosophila NMJ: Stabilizing the Presynaptic Terminal and Regulating Adhesion in the Post-Synaptic Region

Because of the strengths of Drosophila genetics, researchers have turned to the Drosophila larval NMJ for evaluation of adducin/Hts function in the nervous system. Presynaptically, spectrin-actin lattices are important for establishing a platform for regulating neurotransmitter release. Postsynaptically, these lattices are also present where they are responsible for the spacing and efficacy of the NMJ (Pielage et al., 2006). Two groups have shown that adducin/Hts is present in both the pre- and postsynaptic regions and that perturbations of adducin/Hts levels can influence the extent of synaptic branching of the NMJ (Pielage et al., 2011; Wang et al., 2011; **Figure 3**).

Adducin/Hts acts to maintain the presynaptic cytoskeletal organization, as well as to preserve contacts between the presynaptic terminal and muscle, as shown diagrammatically in **Figure 3**. Pielage et al. (2011) focused on the role of pre-synaptic adducin/Hts, where they provide evidence that it regulates synaptic structure through its role as an actincapping protein (**Figures 3A,B**). Of particular interest is their finding that loss of pre-synaptic adducin/Hts causes both overgrowth of the synaptic arbor and its retraction from the muscle, a phenotype particularly relevant to ALS (**Figures 3C,D**). By contrast, post-synaptic adducin/Hts has a distinct role from pre-synaptic adducin/Hts, which acts as an inhibitor of adhesion between nerve and muscle. The ability to break adhesion and then re-form contacts between nerve and muscle is important for growth of the synapse during development (Schuster et al., 1996a), and the loss of postsynaptic Hts prevents synaptic growth, possibly due to increased adhesion between nerve and muscle (**Figure 3E**). Conversely, post-synaptic overexpression of Hts may promote synaptic overgrowth by reducing adhesion between nerve and muscle (**Figure 3D**).

Over-expression of adducin/Hts in the muscle results in delocalization of the key postsynaptic protein, Dlg. Co-immunoprecipitation and proximity ligation assay experiments demonstrate that adducin/Hts is present in a complex with Dlg at the NMJ (Wang et al., 2011, 2014, 2015). These results suggest that the synaptic function of adducin/Hts may be mediated, at least in part, through interactions with Dlg. But how does Hts regulate Dlg postsynaptic targeting? A relevant finding is that adducin/Hts over-expression leads to increases in muscle immunoreactivity against PAR-1 and CaMKII, two protein kinases known to phosphorylate Dlg and disrupt its localization at the post-synaptic membrane (Koh et al., 1999; Zhang et al., 2007; Wang et al., 2011).

localized to actin-spectrin junctions and is found associated with Dlg, possibly through binding to Coracle (Cora). Homophilic adhesion between Fas2 molecules links the pre- and post-synaptic membranes, with the intracellular domain of Fas2 anchored to Dlg. Also associated with Hts is the α-subunit of the Na+/K+-ATPase. (Continued)

#### FIGURE 2 | Continued

Based on studies of mammalian adducin, we speculate that when phosphorylated, Hts translocates from the membrane and its subsequent dephosphorylation may enable Hts access to the nucleus. In the nucleus, Hts is involved in export of transcripts for the kinases PAR-1 and CaMKII. PAR-1 and CaMKII subsequently phosphorylate the scaffolding molecule Dlg, which delocalizes from the NMJ, leading to reduced adhesion between nerve and muscle. In this model Hts monitors the status of the NMJ and can promote synaptic plasticity by shuttling to the nucleus.

Thus, it appears that adducin/Hts regulates synaptic size by controlling Dlg localization at the postsynaptic membrane via PAR-1 and CaMKII-mediated phosphorylation. Dlg interacts with Fas2, which, as mentioned above, is the homolog of mammalian N-CAM (Thomas et al., 1997). Previous studies indicate that Dlg and Fas2 regulate synaptic plasticity and stabilization, and the Drosophila NMJ may prove ideal in addressing regulation of synaptic adhesion by adducin/Hts.

### Adducin is Present in Synapses of the Mammalian CNS

This review has focused on the action of adducin in relation to ALS through its action at the motoneuron and NMJ. However, ALS also affects the CNS, and adducin and phosphoadducin are widely distributed throughout the mammalian nervous system, especially in axons and presynaptic nerve terminals, regions that are likely relevant for the initiation of neurodegeneration in ALS (Seidel et al., 1995; Xu et al., 2013). Of interest is the observation that β-adducin knock-out mice have defects in learning, as well as long-term potentiation and long-term depression (Porro et al., 2010). β-adducin may be required specifically for the establishment of synapses in the CNS under conditions of an ''enriched environment'', an environment where animals are exposed to more sensory stimuli than normal which results in augmented synaptogenesis and improved learning ability (Bednarek and Caroni, 2011). These effects of β-adducin in learning are thought to be due to adducin augmenting the outgrowth of one type of synaptic input to inhibitory interneurons in the hippocampus (Ruediger et al., 2011). It is tantalizing to speculate that the specific effect of adducin in synaptic circuitry might occur through an interaction with Dlg. A mammalian homologue of Dlg is PSD-95, a post-synaptic protein found in excitatory, but not at inhibitory synapses. Thus, modulation of adducin at excitatory and inhibitory synapses would occur differently, as it is likely that adducin would also be found at inhibitory synapses. Adducin is also found in cerebellar Purkinje cells where it associates with an enzyme that generates inositol pyrophosphates (inositol hexakisphosphate kinase-3; IP6K3) and IP6K3 knockout mice have impaired motor learning and coordination, possibly related to impaired disposition of adducin (Fu et al., 2015).

Interestingly, the nematode homologue of adducin, Add-1, has also been drawn into the debate about whether the act of forgetting memories is an active or passive process. Studies in C. elegans have suggested that the long term retention of learned material in this animal is facilitated by Add-1 and inhibited by single stranded RNAs of the mushashi (msi-1) family (Hadziselimovic et al., 2014). These data underscore that adducin has been involved in synaptic architecture for a long time from an evolutionary point of view and is central to synaptic function even in simple organisms.

Circumstantial evidence for the importance of adducin/Hts in CNS function is also provided by recent studies showing that mutations in γ-adducin are associated with familial cerebral palsy (Kruer et al., 2013). Additionally, adducin has been reported to be a binding partner for chorein, a protein believed critical for development of the movement disorder chorea-acanthosis (Shiokawa et al., 2013).

### Adducin: A Reporter of Synaptic Activity at the Nucleus and a Regulator of Transcription at the Synapse?

Although this review has focused on the role of adducin/Hts at the NMJ, there are studies that indicate nuclear functions for adducin that might be relevant to its roles at the synapse. α-adducin contains nuclear localization and export signals, and has been shown to translocate to the nucleus in cultured epithelial


cells upon loss of cell-cell adhesions where it participates in mitotic spindle assembly (Chen et al., 2011; Chan et al., 2014). Interestingly, the nuclear localization and export signals are conserved in Hts, and Hts can localize to the muscle nucleus where it may regulate the export of transcripts that encode for the PAR-1 and CaMKII kinases, but not for transcripts that encode for Dlg (Wang et al., 2014; **Figure 2**). These various results raise the interesting possibility that Hts/adducin may shuttle between the NMJ and the nucleus in its regulation of the synapse. There is considerable evidence that protein trafficking occurs from the synapse to the nucleus especially with regard to activity-dependent gene expression (Kaushik et al., 2014). Furthermore, there is increasing recognition of misregulation of localized translation at the synapse in ALS, and adducin may also be involved in this process (Sephton and Yu, 2015).

## CONCLUSIONS

There are many independent primary causes for developing ALS. Mutations have been found in many genes in FALS involved in disparate processes including mutations in SOD1, TDP-43, C9ORF72 and other genes. Furthermore, the sporadic (non-familial) form of ALS has been associated with other risk factors and it is likely that ALS is a heterogeneous disease in which a variety of distinct initiating factors leads to common clinical manifestations. Current evidence suggests that the motoneuron loss in murine models of ALS are non-cell autonomous and require interactions between motoneurons, corticospinal tract neurons and their nonneuronal neighbors such as microglia, astrocytes and/or terminal Schwann cells. A particularly malicious aspect to ALS, and one that leads to the rapid decline of affected patients, is the inability of synaptic terminals to innervate their targets, particularly muscle, and especially respiratory muscle. This failure of synaptic connectivity might hinge on the ability of adducin to act with interacting proteins to permit strengthening of weakened synapses and allow the NMJ to ''hang on'' simultaneously to both terminal axons and muscle. The possible involvement of adducin and phosphoadducin in ALS suggests strategies for treatment. For instance, if the phosphorylation of adducin leads to some of the neurodegenerative changes, one approach would be to attempt to dephosphorylate adducin, or to inhibit the phosphorylation of adducin which could lead to a strengthening of the preand post-synaptic contacts, but which might exhibit less plasticity. This effect might be achieved pharmacologically by interfering with protein phosphorylation using compounds that block protein phospho-sites with sugar residues (Shan et al., 2012).

## REFERENCES


### AUTHOR CONTRIBUTIONS

CK wrote the review. CK and NH were jointly responsible for writing and editing the manuscript and coordinating the figure preparation. SJHW and SHY were responsible for carrying out experiments relevant for this review, writing of the manuscript and for figure preparation.

### ACKNOWLEDGMENTS

We thank the Natural Science and Engineering Research Council of Canada (NSERC), the Amyotrophic Lateral Sclerosis (ALS) Society of Canada and the Neuromuscular Partnership Program (NRP), a collaborative program between ALS Canada, Muscular Dystrophy Canada and the Canadian Institutes of Health Research (CIHR) for financial support. We also thank the Steel Fund of Simon Fraser University, and the ALS Clinic of Vancouver, Canada.


control of the Arp2/3 complex. Cell 156, 1153–1166. doi: 10.1016/j.cell.2014. 01.054


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Krieger, Wang, Yoo and Harden. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Gene expression profiling for human iPS-derived motor neurons from sporadic ALS patients reveals a strong association between mitochondrial functions and neurodegeneration

Chrystian J. Alves <sup>1</sup> , Rafael Dariolli <sup>2</sup> , Frederico M. Jorge<sup>1</sup> , Matheus R. Monteiro<sup>1</sup> , Jessica R. Maximino<sup>1</sup> , Roberto S. Martins <sup>3</sup> , Bryan E. Strauss <sup>4</sup> , José E. Krieger <sup>2</sup> , Dagoberto Callegaro<sup>1</sup> and Gerson Chadi <sup>1</sup> \*

<sup>1</sup> Department of Neurology, Neuroregeneration Center, University of São Paulo School of Medicine, University of São Paulo, São Paulo, Brazil, <sup>2</sup> Laboratory of Genetics and Molecular Cardiology/LIM13, Heart Institute, University of São Paulo School of Medicine, São Paulo, Brazil, <sup>3</sup> Department of Neurosurgery, Surgical Center of Functional Neurosurgery, Clinics Hospital of University of São Paulo, São Paulo, Brazil, <sup>4</sup> Viral Vector Laboratory, Center for Translational Investigation in Oncology/LIM24, Cancer Institute of São Paulo, University of São Paulo School of Medicine, São Paulo, Brazil

#### Edited by:

Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA

#### Reviewed by:

Stefan Hauser, Deutsches Zentrum für Neurodegenerative Erkrankungen, Germany Mu He, University of California, San Francisco, USA Xiaogang Wu, Institute for Systems Biology, USA

#### \*Correspondence:

Gerson Chadi, Department of Neurology, University of São Paulo, Av. Dr. Arnaldo, 455, 2nd Floor, Room 2119, 01246-903 São Paulo, Brazil gerchadi@usp.br

> Received: 07 May 2015 Accepted: 14 July 2015 Published: 04 August 2015

#### Citation:

Alves CJ, Dariolli R, Jorge FM, Monteiro MR, Maximino JR, Martins RS, Strauss BE, Krieger JE, Callegaro D and Chadi G (2015) Gene expression profiling for human iPS-derived motor neurons from sporadic ALS patients reveals a strong association between mitochondrial functions and neurodegeneration. Front. Cell. Neurosci. 9:289. doi: 10.3389/fncel.2015.00289 Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease that leads to widespread motor neuron death, general palsy and respiratory failure. The most prevalent sporadic ALS form is not genetically inherited. Attempts to translate therapeutic strategies have failed because the described mechanisms of disease are based on animal models carrying specific gene mutations and thus do not address sporadic ALS. In order to achieve a better approach to study the human disease, human induced pluripotent stem cell (hiPSC)-differentiated motor neurons were obtained from motor nerve fibroblasts of sporadic ALS and non-ALS subjects using the STEMCCA Cre-Excisable Constitutive Polycistronic Lentivirus system and submitted to microarray analyses using a whole human genome platform. DAVID analyses of differentially expressed genes identified molecular function and biological process-related genes through Gene Ontology. REVIGO highlighted the related functions mRNA and DNA binding, GTP binding, transcription (co)-repressor activity, lipoprotein receptor binding, synapse organization, intracellular transport, mitotic cell cycle and cell death. KEGG showed pathways associated with Parkinson's disease and oxidative phosphorylation, highlighting iron homeostasis, neurotrophic functions, endosomal trafficking and ERK signaling. The analysis of most dysregulated genes and those representative of the majority of categorized genes indicates a strong association between mitochondrial function and cellular processes possibly related to motor neuron degeneration. In conclusion, iPSC-derived motor neurons from motor nerve fibroblasts of sporadic ALS patients may recapitulate key mechanisms of neurodegeneration and may offer an opportunity for translational investigation of sporadic ALS. Large gene profiling of differentiated motor neurons from sporadic ALS patients highlights mitochondrial participation in the establishment of autonomous mechanisms associated with sporadic ALS.

Keywords: amyotrophic lateral sclerosis, sporadic ALS, induced pluripotent stem cells, motor neuron differentiation, microarray

### Introduction

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease that is characterized by a selective death of motor neurons in the brain and spinal cord of adults, leading to rapid respiratory muscle failure and patient death (Cleveland and Rothstein, 2001). Sporadic ALS is not genetically inherited and represents approximately 95% of all ALS cases. A minority of cases are familial forms and are associated with gene mutations (Gros-Louis et al., 2006; Turner and Talbot, 2008).

The generation of transgenic animals that carry mutated human genes associated with familial ALS, for instance superoxide dismutase (SOD1), tar DNA protein 43 (TDP-43), fused in sarcoma, and valosin-containing protein genes, has allowed the identification of basic mechanisms underlying neurodegeneration in the disease (McGoldrick et al., 2013). In fact, distinct processes of motor neuron death that are related to toxic glial paracrine (Nagai et al., 2007; Gowing et al., 2009) and autocrine (Ringer et al., 2012) signaling mechanisms have been described using animal models. Furthermore, the early peripheral pathological events of neuromuscular junction loss and motor axon retraction, which may start long before symptom onset, have also been described in these animal models (Rocha et al., 2013; Venkova et al., 2014). Indeed, recent evidence has pointed out the importance of close morphological and physiological relationships between peripheral non-neuronal Schwann cells (Keller et al., 2009; Chen et al., 2010; Maximino et al., 2014) and fibroblasts (Raman et al., 2015) with motor axons, as key participants in the early degeneration of motor neurons in ALS.

Despite significant progress, it is clear that shortcomings remain with regard to recapitulating the processes underlying the clinical form of sporadic ALS (Benatar, 2007b), based on mechanisms described in transgenic models carrying specific human gene mutations. Actually, these limitations inherent in animal models have been considered the major obstacle preventing the successful translation of results to clinical practice. No therapy has been effective in counteracting human disease progression or extending survival of ALS patients (Benatar, 2007a). Moreover, there is a general consensus that the development of therapeutic targets promoting positive clinical results have been hampered by lack of a relevant human disease model specific to sporadic ALS (Gordon and Meininger, 2011; Veyrat-Durebex et al., 2014).

One of the most promising approaches to provide a better understanding of the mechanisms underlying motor neuron degeneration in sporadic ALS is the generation of specific human-induced pluripotent stem cells (hiPSC) from fibroblasts of motor nerves from sporadic ALS patients. These hiPSCs can be differentiated into multiple cell types, including mature motor neurons which are normally inaccessible for in vitro studies (Veyrat-Durebex et al., 2014).

In fact, human motor neurons have been previously generated from human embryonic stem cells and hiPSCs carrying specific gene mutations, all related to the less prevalent, familial form of ALS (Di Giorgio et al., 2008; Marchetto et al., 2008; Mitne-Neto et al., 2011). Despite the substantial effort involved in generating iPSC-derived, differentiated neurons from familial ALS patients, such motor neurons developed from adult cells (i.e., skin fibroblasts and erythroblasts) harbor specific gene mutations, and thus may not represent an appropriate model from which to develop translational therapies specific to the more dominant, sporadic forms of ALS. Thus, this approach is presently restricted to the less prevalent form of ALS (Leblond et al., 2014).

In this study, hiPSC-derived motor neurons were obtained from sporadic ALS patients by reprogramming motor nervederived fibroblasts using a STEMCCA Cre-Excisable Constitutive Polycistronic Lentivirus System containing the four transcription factors OCT4, SOX2, KLF4, and CMYC (Somers et al., 2010). A large gene profiling analyses of these differentiated motor neurons was performed using a high-density oligonucleotide microarray linked to specific tools capable of identifying biological processes, molecular functions and pathways deregulated in the sporadic ALS form.

Our results demonstrate that the generation of hiPSCs by reprogramming motor nerve-derived fibroblasts from sporadic ALS patients, followed by their differentiation into adult motor neurons, is not only feasible, but more importantly, can be used to model key molecular mechanisms which may recapitulate those processes related to neurodegeneration in this disease state. This approach may thus provide one of the most promising platforms for the development of therapeutic targets specific to sporadic ALS.

### Methods

#### Human Tissue Samples

Sporadic ALS patients accompanied at the ALS Ambulatory Unit of Department of Neurology of Clinics Hospital of University of São Paulo School of Medicine signed informed consent and were included in the present study after neurological evaluation. Inclusion criteria were less than 6 months of disease evolution after diagnosis according to EL ESCORIAL (Brooks et al., 2000) and motor impairments not affecting substantially one of the lower limbs. Sporadic ALS patients showed preserved movements of the ankle and big toe in the foot corresponding to the site of biopsy. Exclusion criteria were the presence of associated pathologies, breathing disorders, swallowing and cognitive disorders. Patients were submitted to surgical procedures in one of their feet by means of local anesthesia to collect a fragment of their extensor hallucis brevis nerve which is a motor nerve that innervates the extensor hallucis brevis muscle in the dorsal region of the foot. As control, biopsies were taken from the distal accessory nerve performed during reconstructive surgery of traumatic brachial plexus injuries of non-ALS patients at the Clinics Hospital of University of São Paulo School of Medicine. Non-ALS patients reported no history of familial ALS and clinical evaluation failed to show any signs of ALS. Both procedures were performed in the Surgical Center of Functional Neurosurgery of Clinics Hospital of University of São Paulo in accordance with protocol number 0187/11 approved by the Ethics Committee for Analysis of Research Projects of Clinics Hospital of University of São Paulo School of Medicine.

#### DNA Sequencing to Exclude Most Frequent Gene Mutations Associated with Familial ALS SOD1 and TARDBP Analyses

The SOD1 and TARDBP genes were sequenced in ALS patients to verify the absence of DNA mutations in genes normally associated with familial ALS, thus providing further confirmation of the sporadic form of ALS in patients included in the study. Genomic DNA was extracted from peripheral blood using standard methods. For sequencing analysis, primers were designed so that they flanked all five exons of SOD1 and the first six exons of TARDBP, both within intronic regions (Hallewell et al., 1986; Xiong et al., 2010). Direct sequencing of amplified exons was performed using Big-dye R Terminator v3.1 sequencing (Applied Biosystems, USA). Reactions were performed by means of a ABI 3130 genetic analyzer (Applied Biosystems), and sequences were analyzed by using a Mutation Surveyor v5.0 (SoftGenetics, USA).

#### C9orf72 Repeat Expansions Analysis

Genomic DNA was isolated from peripheral blood according to standard protocols. The repeat number of the GGGGCC hexanucleotides was determined using genotyping primers. Repeat-primed polymerase chain reaction (PCR) was performed in order to provide a qualitative assessment of the presence of C9orf72 repeat expansions, as previously described (Dejesus-Hernandez et al., 2011). Briefly, 200 ng of genomic DNA were used as template in a final volume of 28µl containing 12.5µl of FastStart PCR Master Mix (Roche Applied Science, USA), and a final concentration of 0.25 mM 7-deaza-dGTP (Roche Applied Science), 5% dimetil sulfoxide (Sigma-Aldrich, USA), 1M betaine (Sigma-Aldrich) and 1µM of each primer. Sample analyses were performed on an ABI 3500 genetic analyzer (Applied Biosystems), and data evaluated using the GeneMapper software. Repeat expansions are known to produce a characteristic sawtooth pattern with a 6-bp periodicity, as previously described (Dejesus-Hernandez et al., 2011).

#### hiPS Cell Generation

Fibroblasts were obtained from fragments of extensor hallucis brevis peripheral motor nerve biopsies from sporadic ALS patients and from distal accessory nerves collected from control non-ALS patients. Fibroblasts of sporadic ALS and non-ALS subjects were submitted to Sendai and STEMCCA methods of reprogramming as described below.

#### Sendai Reprogramming

After purification, fibroblasts were expanded in culture (Seluanov et al., 2010) and reprogrammed based on the protocol described by Macarthur et al. (2012). Briefly, approximately 10<sup>5</sup> human fibroblasts were seeded per well in a 6-well plate and incubated at 37◦C and 5% CO2. The next day, cells were transduced with the CytoTune iPS Reprogramming Kit containing Sendai virus vectors (Life Technologies, USA; Cat. # A1378001) in a fibroblast growth medium at a multiplicity of infection of 3, as described in the manufacturer's protocols. The medium was replaced with fresh fibroblast growth medium 1 day after transduction and the cells were cultured for 7 days. Cells (2 × 10<sup>5</sup> cells per dish) were cultured on an inactivated mouse fibroblast feeder layer in 100 mm tissue culture dishes at day 8 of transduction. After an overnight period, the hiPSC medium was replaced daily. The hiPSC medium is composed of DMEM-F12 containing 20% knockout serum replacement (KSR), 1% MEM non-essential amino acids (NEAA) solution, 0.1% 2-mercaptoethanol. Medium was equilibrated at 37◦C and added with 4 ng/mL FGF-2 just prior the use. Reagents were purchased from Life Technologies. The hiPSC colonies were picked for expansion and characterization from days 27 to 30 of reprogramming.

#### STEMCCA Reprogramming

Fibroblasts of sporadic ALS and non-ALS subject groups were expanded in culture as described above. The reprogramming was performed based on the protocol described by Somers et al. (2010). Briefly, 10<sup>5</sup> fibroblasts were plated onto a GelTrex (Life technologies) pretreated P35 mm dish in 10% fetal bovine serum and DMEM. The fibroblasts were subjected to reprogramming 24 h after plating by STEMCCA Cre-Excisable Constitutive Polycistronic Lentivirus (Millipore, USA) expressing the pluripotency related genes OCT4, SOX2, KLF4, and CMYC. DMEM media was replaced by E6 medium (Life Technologies) supplemented with 10 ng/ml human FGF-2 (Life Technologies) (E6F) and 0.5 mM NaB (Sigma-Aldrich) 24 h after transduction. The medium was changed by fresh E6F on days 3, 5, 7, and 9. Subsequently, the medium was changed daily from day 11 to 25 by the E8 medium (Life Technologies) supplemented with 0.25 mM of NaB (Sigma-Aldrich). The colonies were visualized with the aid of a phase contrast microscope at the end of that period and were collected manually. The colonies were then seeded in GelTrex pretreated dishes for the expansion in E8 medium. The hiPSCs-derived from sporadic ALS and non-ALS subjects were characterized by means of immunocytochemistry (**Table 1**), reverse transcriptase polymerase chain reaction (RT-PCR) and quantitative polymerase chain reaction (qPCR) (**Table 2**) and alkaline phosphatase live stain described below.

#### Alkaline Phosphatase Live Staining

The culture media of hiPSCs from ALS and non-ALS subjects achieved with Sendai and STEMCCA methods of reprogramming were aspirated from colonies and were washed twice with DMEM/F-12 medium (Gibco). The colonies were incubated for 30 min at 37◦C with alkaline phosphatase live stain (Life Technologies) diluted in DMEM/F-12 medium (1:500) and then washed 3x with the same medium. Positive stained colonies were visualized by means a FITC filter under fluorescent microscopy (EVOS XL Core Imaging System, Life Technologies).

#### Karyotyping

Eighty percent confluent hiPSC colonies were incubated for 2 h in the E8 medium with 0.5µg/ml colchicine for karyotyping procedures. Cells were then harvested with TrypLE (Life Technologies) and incubated for 20 min at 37◦C in 0.075 M KCl solution, fixed in methanol/acetic acid (3:1), placed on pre-wet chilled microscope slides, air-dried, and incubated for 2 days at 37◦C. The microscope slides were stained for 90 s with Wright's

#### TABLE 1 | Antibodies for characterization of hiPSC, embryoid body, differentiated motor neurons and cardiomyocytes.


SSEA-4 (stage-specific embryonic antigen-4), TRA1-60 and TRA1-81 (podocalyxin-like), Oct-4 (octamer-binding transcription factor 4), Pck-26 (pan cytokeratin 26), Myosin LC 2 (myosin light chain 2), EphA2 (ephrin type-A receptor 2), MAP2 (microtubule-associated protein 2), ChAT (choline acetyltransferase), cTnT (troponin T), CD-31 (cluster of differentiation 31), TRA-1 (histone acetyltransferase TRA1), Myo (Myosin). hiPSC were obtained from motor nerve fibroblasts of sporadic ALS and non-ALS subjects.

#### TABLE 2 | RT-PCR and qPCR primers for hiPSC and differentiated motor neuron gene markers.


POUF51 (pou domain class 5, transcription factor), SOX2 (sex determining region Y-box 2), NANOG (homeobox transcription factor nanog), GDF3 (growth differentiation factor-3), ESG/DPPA5 (embryonic stem cell specific gene/developmental pluripotency-associated), DPPA4 (developmental pluripotency-associated protein 4), DPPA2 (developmental pluripotency-associated protein 2), REX1/ZFP42 (reduced expression protein 1/zinc finger protein 42), PAX6 (paired box 6), OLIG2 (oligodendrocyte lineage transcription factor 2), CHAT (choline acetyltransferase), HB9 (homeobox 9) and GAPDH (glyceraldehyde-3-phosphate dehydrogenase). hiPSC were obtained from motor nerve fibroblasts of sporadic ALS and non-ALS subjects.

stain in 2x buffer. Standard G-banding chromosome analysis was performed by Clinics Hospital Molecular Diagnostic Services (São Paulo, Brazil).

#### hiPSC Differentiation In vitro and Characterization

Differentiation of induced pluripotent cells was accomplished as described elsewhere (Evans and Kaufman, 1981; Martin, 1981).

#### Embryoid Body Differentiation

Briefly, hiPSCs were lifted using StemPro EZPassage (Life Technologies) and maintained for 4 days in suspension for embryoid body formation. The cells were cultured in embryoid body medium, composed of knockout DMEM/F12 (Life Technologies), supplemented with 20% KSR (Life Technologies), 2 mM GlutaMAX (Life Technologies), 100µM NEAA (Life Technologies), 1% antibiotic-antimycotic (Sigma-Aldrich) and 100µM 2-mercaptoethanol (Life Technologies). Addition of ROCK-inhibitor Y-27632 (Ascent Scientific, UK) for the first 24 h was used to improve the survival of hiPSC single cells (Lai et al., 2010) and to allow an adequate generation of embryoid bodies as well. The embryoid bodies were transferred to Geltrexcoated tissue-treated dishes to allow cell attachment and were maintained there for 1 week until their in vitro analysis of differentiation by means of immunocytochemistry with specific markers of three germ layers (**Table 1**).

#### Cardiomyocyte Differentiation

To further analyze their differentiation potential in vitro, hiPSCs were induced to directly differentiate into cardiomyocytes. The procedure was performed using a PSC cardiomyocyte differentiation Kit (Life Technologies), according to the manufacturer's protocol. Briefly, after splitting, cells were fed by cardiomyocyte differentiation medium A at day 4. Cells were then fed by cardiomyocyte differentiation medium B on day 6 and were maintained in a cardiomyocyte maintenance medium from day 8 to day 14. At that time point, cells were evaluated for their ability to show contraction as a normal feature of cardiomyocytes (Supplementary Video), characterized by immunocytochemistry for cardiomyocyte cell markers (**Table 1**) and, stained with fluorescent phalloidin actin binding (Life Technologies) performed according to the manufacture instructions.

#### hiPSC Differentiation In vivo

In order to evaluate the ability of hiPSC to develop teratomas, approximately 3x10<sup>6</sup> hiPSCs from ALS and non-ALS patients were injected subcutaneously into the dorsal flanks of nude rats anesthetized with ketamin (Cristália, Brazil)-xilazin (Vetbrands, Brazil; 62.5 mg/kg). Teratomas were allowed to expand for about 6 weeks. They were visualized and dissected by means of surgical procedure, fixed overnight in 4% paraformaldehyde and transferred to 70% ethanol until they were embedded in paraffin. Sections were stained with hematoxylin and eosin for microscopy analyses. Protocols for animal use were previously approved by the School of Medicine of the University of São Paulo Institutional Animal Care and Use Committee.

#### Motor Neuron Differentiation and Characterization

Motor neuron differentiation was performed according to a protocol described by Hu and Zhang (2009) and modified in our lab. Briefly, after reaching confluence, hiPSC colonies were cultured in suspension in the presence of embryoid body medium to achieve embryoid body formation as described above. The medium was replaced on day 4 by a neural differentiation medium containing DMEM/F12 (Life Technologies), N2-supplement (Invitrogen, USA), 100µM NEAA (Life Technologies), 1% antibiotic-antimycotic (Sigma-Aldrich) and 2µg/ml heparin (Cristália) to induce the formation of the neural progenitor cells. Clusters attached to laminincoated dishes (20µg/ml, Sigma-Aldrich) after 1 week in suspension. Primitive neuroepithelial cells were posteriorized by addition of 0.1µM retinoic acid (Sigma-Aldrich) at day 10 and ventralizated by the addition of 100 ng/ml sonic hedgehog (Shh; Sigma-Aldrich) and B27 supplement (Gibco, USA) at day 14. The cells were collected at differentiation day 20 for microarray experiments. Samples of hiPSC-derived motor neurons from ALS and non-ALS subjects were also obtained on small laminin-coated coverslips (13 mm) and characterized by immunocytochemistry (**Table 1**), RT-PCR, qPCR (**Table 2**) and Hb9::GFP live reporter.

#### Hb9::GFP Live Reporter Technique for Motor Neuron Visualization

The construct containing Hb9-GFP (Hb9::GFP) reporter was transferred to the differentiated motor neurons using a lentivirus system (Marchetto et al., 2008). Hb9 is a motor neuron-specific transcription factor expressed in mature cells that binds to the promoter region of the green fluorescent protein (GFP) sequence in the construct. HEK 293T cells were employed for virus production. Briefly, triple transfection with calcium phosphate was performed 24 h after plating using the PSPAX2 plasmids, the pCMV-VSVg and the pLenti-Hb9-GFP (Addgene plasmid # 37080) or pEGIP (constitutive lentiviral vector for expression of GFP; Addgene plasmid # 26777) (Marchetto et al., 2008; Zou et al., 2009). The supernatant containing viral particles was collected 24 and 48 h after transfection. In parallel, differentiated motor neurons were plated at a concentration of 2 × 10<sup>6</sup> cells per dish on 60 mm plates. Transduction was achieved by adding the lentivirus to hiPSC-derived motor neuron cell cultures after the treatment with Shh. The differentiated motor neurons were maintained on small glass coverslips in 24-well plates and transduced using 200µl viral supernatant plus 100µl of neural differentiation medium containing 8 mg/ml polybrene. The medium was replaced 6 h after incubation at 37◦C and 5% CO2. The cells were fixed 72 h after transduction and were evaluated for GFP expression using an Olympus AX-70 microscope (Olympus, JP).

#### Quantification of Differentiated Motor Neurons

The amount of differentiated motor neurons was quantified by the counting of ChAT immunoreactive profiles and also Hb9::GFP profiles counterstained with nuclear DAPI. Briefly, a sample of cells were seeded onto small laminin-coated (Sigma) glass coverslips and immunostained for ChAT (Millipore) or labeled with Hb9:GFP methodology (the construct Hb9::GFP as described above) followed by a DAPI nuclear labeling in order to determine the number of differentiated motor neurons in the sample fields. Evaluation of motor neuron vitality was performed by means of Fluoro-Jade C (FJC) analysis as described in the Figures S4E,F of the Supplementary Material (Schmued et al., 1997, 2005).

#### Immunocytochemical Characterization

The hiPSC colonies, the embryoid body differentiation in vitro and the differentiated motor neurons were characterized by indirect immunofluorescence, as described elsewhere (Maximino et al., 2014). Briefly, cells were fixed with 4% paraformaldehyde and incubated with primary antibodies shown in **Table 1**. Primary antibodies were detected using Alexa Fluor <sup>R</sup> 488 or 594-conjugated secondary antibodies specific for mouse, rabbit and goat (all from Invitrogen). Preparations were mounted on microscope slides and counterstained with nuclear 4 ′ ,6-diamidino-2-phenylindole dihydrochloride (DAPI; Vector, USA). Digital images of immunofluorescence staining were obtained by means of an Olympus AX-70 microscope (Olympus).

#### RT-PCR and qPCR Characterization

Molecular characterizations of hiPSC and of differentiated motor neurons of the chamber culture were performed by means of RT-PCR and qPCR as described elsewhere (Maximino et al., 2014). Briefly, total RNA of cells contained in the 24-well plates was extracted using Trizol (Life Technologies) according to the manufacturer's protocol. The quantity (NanoDrop 1000 Spectrophotometer; Thermo Scientific, USA) and quality (Agilent 2100 bioanalyser, RNA 6000 Pico LabChip; Agilent Technologies, USA) of RNA were determined. cDNAs were amplified by PCR using the primers specific for hiPSC and differentiated motor neurons (**Table 2**). PCR products were submitted to electrophoresis in 2% agarosis gels containing ethidium bromide for 60 min at 100 volts, and then visualized under ultraviolet light exposure. qPCR reactions were performed as described below. hESC (Fraga et al., 2011) and neural progenitor cells were used as positive controls for hiPSC and negative controls for motor neuron characterization, respectively (Marti et al., 2013; Jha et al., 2015).

#### RNA Isolation and Microarray Experiments

The differentiated motor neurons that were achieved from a pool of neuronal progenitor lines of the two sporadic ALS and two non-ALS patients, this process were performed in three different times. A total of three ALS and three non-ALS (control) samples of differentiated motor neurons obtained were subjected to large gene profiling by means of microarray analysis. The same sampling procedure was employed in other quantification methods of the study. The procedures of RNA isolation and the methodology of microarray experiments were described in our previous publication (Maximino et al., 2014). Briefly, total RNA was extracted from differentiated motor neurons and linear amplification of RNA was performed using the RiboampHSplus kit (Arcturus) according to the manufacturer's protocol, control of RNA quality was them assessed, as described above. A representative electropherogram from a Bioanalyzer evaluation of RNA integrity of the differentiated motor neurons is shown in the Supplementary Material (Figure S4H). RNAs of samples (25 ng) and reference (100 ng) were reverse transcribed by the low-input RNA linear amplification kit and then transcribed to Cy5-labeled (samples) or Cy3-labeled (reference) according to the manufacturer's instructions (Agilent Technologies) and to our previous descriptions (de Oliveira et al., 2013). A total of 300 ng of Cy5-labeled cRNA was hybridized together with the same amount of Cy3-labeled reference to Whole Human Genome Oligo 8 × 60 K. All steps were performed according to the manufacturer's instructions (Agilent Technologies) and to our previous descriptions (de Oliveira et al., 2013; Maximino et al., 2014).

#### Microarray Analysis

The slide were scanned by a Microarray Scanner (Agilent) and the raw image data were converted to numerical data using the Agilent Feature Extraction Software, version 11.0.1.1 (Agilent Technologies), as described in our previous study (Maximino et al., 2014). Microarray raw data (.txt files) were imported into GeneSpring v.13 GX software package (Agilent Technologies). Raw signal intensities were normalized using this software. The probes were tested for differential expression using a moderated t-test for the comparisons between differentiated motor neurons that were achieved from a pool of neuronal progenitor lines of the two control patients and the two ALS patients, using GeneSpring v.13 GX Software. Genes with p < 0.05 were accepted as differentially expressed genes. The raw data from hybridizations are available on the Gene Expression Omnibus Database, and the GEO accession number is GSE68240. The data were registered in the GEO according to the following description: control samples (control sample 1, control sample 2, control sample 3) and ALS samples (ALS sample 1, ALS sample 2, ALS sample 3) of differentiated motor neurons that were achieved from a pool of neuronal progenitor lines of the two control patients and the two ALS patients, respectively.

### Functional Enrichment Analysis

The functional annotation analysis of differentially transcribed genes was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) web-server v.6.7 (http://david.abcc.ncifcrf.gov) (Huang Da et al., 2009). Gene Ontology (GO) terms of Biological Process and Molecular Functions, and also Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways included in the DAVID knowledgebase were considered. We have also focused on the GO terms mitochondrion of the Cellular Component and their related genes. Furthermore, the mean of gene expression normalized signals related to the mitochondrion was compared between the two groups. GO terms were submitted to REVIGO, a web server that takes long lists of GO terms and summarizes them in categories and clusters of differentially expressed genes by removing redundant entries (Supek et al., 2011). The analysis was conducted on the list containing the up-regulated and downregulated genes. High stringency (EASE score) parameters were selected to improve confidence in those values designated as enriched (Ashburner et al., 2000; Kanehisa et al., 2006).

### Quantitative PCR

qPCR was carried out for the analysis the expression of pluripotency markers, NANOG, SOX2, OCT4, and CHAT in the hiPSCs and also the expression of the PAX6 (neural stem cell marker), OLIG2 (motor neuron precursor marker), and the CHAT, HB9 (mature motor neuron markers) in the differentiated motor neurons. The details of the normalization procedures were described in the legends of the **Figures 2**, **5** and Figure S4. Furthermore, a sample of differentially expressed genes described by microarray was selected for verification by qPCR, as described in our previous publication (de Oliveira et al., 2014). The genes were chosen for verification based on their high or low fold relative expression levels and possible involvement in ALS or other neurodegeneration related mechanisms. Thus, dual specificity phosphatase 6 (DUSP6), diacylglycerol o-acyltransferase 1 (DGAT1), potassium channel, subfamily K, member 12 (KCNK12), kinesin family member C1 (KIFC1); keratin associated protein 4-11 (KRTAP4-11); leucine zipper-EF-hand containing transmembrane protein 1 (LETM1); succinate dehydrogenase complex assembly factor 1 (SDHAF1); caspase 9, apoptosis-related cysteine peptidase (CASP9); vacuolar protein sorting 35 homolog (S. cerevisiae) (VPS35); and insulin-like growth factor 2 (IGF2) were verified by qPCR.

Briefly, cDNA of differentiated motor neurons, as described above was synthesized from 50 ng of total RNA by using a Maxima First Strand cDNA Synthesis Kit (ThermoScientific,USA) according to manufacturer. qPCR reactions were carried out in duplicate with 5 ng cDNA, the DyNAmo Color Flash SYBR Green qPCR kit (Thermo Scientific) and 400 nM of each primer in a final volume reaction of 20µl, by using the Applied Biosystems 750 Real-Time PCR System (Applied Biosystems). The SYBR primers used in qPCR verification can be found in **Table 3**.

The cycling in the SYBR reactions was composed of an initial denaturation at 95◦C for 10 min. Templates were subsequently amplified by 40 cycles of 95◦C for 15 s and 60◦C for 30 s. A dissociation curve was then generated to ensure amplification of a single product, and the verification of no primer dimers formation. A standard curve was generated for each primer pair in order to determine the efficiency of the PCR reaction over a range of template concentrations from 0.032 ng/µl to 20 ng/µl, using cDNA synthesized from human differentiated motor neuron reference RNA. The efficiency for each set of primers was 100 ± 5%. Gene expression was normalized to GAPDH, and was determined using the 11Ct mathematical model (ABI PRISM 7700 Sequence Detection System protocol; Applied Biosystems). GAPDH was chosen as a housekeeping gene to normalize the qPCR values because the microarray analysis showed that its expression was stable across samples.

#### Statistical Analyses

The statistical method employed in the microarray analysis is described above. The One-Way ANOVA was applied with Tukey's multiple comparisons post-test to identify statistical significances between samples within groups. Furthermore, onetailed unpaired t-test was used to determine the statistical significance of differences in qPCR experiments. All analyses were performed using Graphpad Prism 5 (GraphPad, USA). Data were presented as means ± SEM and significance level was set at p < 0.05.

### Results

#### Clinical Evaluation of ALS Patients and DNA Sequencing

Sporadic ALS patients showed clinical and laboratory features of ALS according to EL SCORIAL (**Table 4**). The evolution of the disease did not exceed 29 months after initial symptoms and ALS patients showed preserved movements of the ankle and the big toe in the foot corresponding to the site of biopsy. The amplitude of the big toe movement diminished after the biopsy of the extensor hallucis brevis nerve, but they showed a progressive recovery after 4 weeks of surgical intervention. Unfortunately, due the natural evolution of the disease, ALS patients lost movements of the entire inferior limbs in the subsequent months. ALS patients did not present with any cognitive impairments or familial history of neurodegenerative disease.

Non-ALS patients were adult subjects that had suffered a unilateral injury of the brachial plexus months before the procedure of plexus surgery. The distal fragments of the preserved accessory nerve were obtained when the nerve was cut and transferred to one of the major nerves of the brachial plexus. That means the donor nerve was functional and did not show signs of local inflammation at the time of surgery.

Sporadic ALS patients showed no SOD1 and TARDBP gene mutations within the coding regions of those genes. Furthermore,



DUSP6 (dual specificity phosphatase 6), DGAT1 (diacylglycerol o-acyltransferase 1), KCNK12 (potassium channel, subfamily K, member 12), KIFC1 (kinesin family member C1), KRTAP4-11 (keratin associated protein 4–11), LETM1 (leucine zipper-EF-hand containing transmembrane protein 1), SDHAF1 (succinate dehydrogenase complex assembly factor 1), CASP9 (caspase 9, apoptosis-related cysteine peptidase), VPS35 (vacuolar protein sorting 35 homolog), IGF2 (insulin-like growth factor 2) and GAPDH (Glyceraldehyde-3-Phosphate Dehydrogenase). hiPSC were obtained from motor nerve fibroblasts of sporadic ALS and non-ALS subjects.


#### TABLE 4 | Patient's information.

RLL, right lower limb; LUL, left upper limb.

<sup>a</sup>Time from symptom onset to nerve biopsy. Patients from the ALS Ambulatory Unit of Department of Neurology were diagnosed accordingly to EL ESCORIAL and were submitted to a biopsy of extensor hallucis brevis nerve. Non-ALS patients were followed at Neurosurgery Division of Department of Neurology and were submitted to a biopsy of motor branches of accessory nerve during the reconstructive surgery of brachial plexus. The strength in the big toe of 3/5 indicates that functionality of segment was substantially preserved.

C9orf72 repeat expansions were not present in the sporadic ALS patients (Figure S1).

#### Generation and Characterization of hiPSCs from Motor Nerve Fibroblasts

Integration-free iPS cell lines were generated using a Sendai virus -based on individual delivery of the 4 Yamanaka reprogramming factors, i.e., OCT4, SOX2, KLF4, and CMYC (Takahashi et al., 2007). Fibroblasts of all subjects showed successful reprogramming, with no morphological changes observed up to 12 days of reprogramming. The colonies of stem cells were tightly packed, showed a round morphology, and lacked well-defined sharp edges (Figures S2A,B). Unfortunately, several colonies regressed 3 weeks later and expansion of embryonic markers (see below) occurred in only a few colonies. The emerging colonies expressed the surface embryonic stem cell marker, SSEA4, and the pluripotency marker, alkaline phosphatase as shown in Supplementary Material (Figures S2C,D).

Based on the above results regarding the low level of reprogramming achieved by the Sendai method in our laboratory, we subsequently transduced fibroblasts using the STEMCCA Cre-Excisable Constitutive Polycistronic Lentivirus (Somers et al., 2010) in order to increase efficiency. In fact, morphological changes were observed as early as 6 days after reprogramming. Thus, 20 embryonic stem cell-like colonies were picked and expanded under feeder-free conditions 4 weeks after transduction. The best candidates for hiPSC clones from sporadic ALS and non-ALS subjects were selected based on their morphology and were employed in further characterization and motor neuron differentiation experiments.

All hiPSC colonies obtained from sporadic ALS and non-ALS control subjects showed classical morphology and their cells expressed the surface embryonic stem cell markers SSEA-4, TRA1-60, and TRA1-81, as well as the transcription factor OCT4 (**Figures 1A–F**) and alkaline phosphatase (**Figure 1G**). Karyotyping analysis of samples of hiPSC lines from all subjects showed an absence of alterations in the hiPSC chromosomal number and morphology (**Figure 1H**). Furthermore, colonies from ALS and non-ALS subjects showed NANOG, SOX2, OCT4, DPPA4, ESG/DPPA, DPPA2, REX1, GDF3, and CHAT signals which were similar to those of an hESC line and substantially different from the iPSC-differentiated motor neurons (**Figure 2A**). Furthermore, hiPSC colonies obtained from sporadic ALS and non-ALS control subjects expressed the higher levels of NANOG and OCT4 as evidenced by the qPCR experiments. Interestingly, the expression of SOX2, as seen in the hiPSCs and also in the differentiated motor neurons could be related to the fact that we have not performed the silencing of exogenous factors after lentiviral induction or to a natural timedependent decrease of these pluripotency factors (Muratore et al., 2014; Raitano et al., 2015) (**Figure 2B**).

The pluripotency of the hiPSCs from ALS and non-ALS subjects was also evaluated in vitro through the formation of embryoid bodies. All hiPSC lines differentiated spontaneously into cell types of the three embryonic germ layers, as indicated by expression of the specific markers; Pck-26 (ectoderm), Myosin LC 2 (mesoderm), and EphA2 (endoderm) (**Figures 3A–C**).

Furthermore, the pluripotency of hiPSC from ALS and non-ALS subjects was confirmed by the analysis of their ability to develop teratomas in vivo, after subcutaneous injection in nude rats. Histopathological analyses demonstrated tissues derived from the three germ layers, including neural rosettes from ectodermal origin, muscle tissue from mesodermal origin and gut-like cells from endodermal origin (**Figures 3D–F**) in the tumors 6 weeks after injection. These findings indicated a successful reprogramming of fibroblasts to a pluripotent state from sporadic ALS and non-ALS subjects.

extensor hallucis brevis nerve obtained by biopsy from sporadic ALS patients and from distal fragments of the accessory nerve from a non-ALS subject. Immunostaining of cultured hiPSC for SSEA-4 (B), TRA1-60 (E), TRA1-81 (C), and OCT-4 (F). The presence of OCT-4 immunoreactivity in the nucleus

#### Differentiation of hiPSC into Cardiomyocytes

hiPSCs were also differentiated into cardiomyocytes in order to further characterize their potential to generate additional mature cell types for further functional analyses. At 14 days, ALS differentiated cardiomyocytes were present as spontaneously contracting syncytia (Supplementary Video) expressing troponin T cardiac type 2 (cTnT) (Figures S3A,B), but were negative for CD-31 (an endothelial cell marker; Figure S3C), and TRA-1 (a stem cell marker, Figure S3D). hiPSCdifferentiated cardiomyocytes expressed positive double-staining against myosin and actin to show the identity and the structural pattern of cardiac differentiated cells (Figures S3E–H).

#### Differentiation of hiPSCs into Motor Neurons

hiPSC lines from ALS and non-ALS subjects were also manipulated in order to generate mature motor neurons. hiPSC colonies (**Figure 4A**) first generated embryoid bodies in suspension (**Figure 4B**) and neuronal progenitor cells. Neural progenitor cells expressing the PAX6 marker (**Figure 5K**) were analyzed for alkaline phosphatase activity using the Alkaline Phosphatase Live Stain (G). Karyotyping sampling of hiPSC clone from an ALS patient, in which metaphase plates showed the normal male chromosomal content (H; 42,XY). The same pattern was observed for other subjects of the study, including the non-ALS patients (data not shown). Scale bars, 50µm.

differentiated into post-mitotic motor neurons (**Figures 4C,D**) expressing MAP2 (**Figure 5A**) and ChAT (**Figure 5B**). Motor neurons were characterized further by means of a Hb9 (promoter/enhancer)-GFP (Hb9::GFP) reporter construct (**Figure 5C**), an early marker of these maturing motor neurons. Negative controls are shown in **Figures 5D–F**. There were no differences in the neuronal differentiation process between hiPSCs derived from sporadic ALS or non-ALS subjects. Samples of differentiated hiPSC-mature motor neurons from ALS patients (**Figure 5G**) and their non-ALS controls (**Figure 5H**) showed axonal projections. The results of the quantification of the number of differentiated motor neurons in the sample regions of the coverslips revealed the presence of 95.39 and 95.3% of ChAT positive and Hb9 positive profiles, respectively, in relation to the total amount of DAPI nuclear profiles in these regions (**Figures 5I,J**).

RT-PCR and qPCR analyses of neuronal gene markers of neural progenitor cells and differentiated motor neurons contained in the 24-well plates are shown in the **Figures 5K,L**

and Figure S4G of Supplementary Material. The marker PAX6 is present only in neural progenitor stage. Despite the very low OLIG2 signal seen by RT-PCR (**Figure 5K**), a higher level of signal amplification of OLIG2 was detected in differentiated motor neurons by qPCR (**Figure 5L** and Figure S4G of Supplementary Material). Furthermore, qPCR showed high expressions of CHAT and HB9 in the differentiated motor neurons compared to neural progenitor cells (**Figure 5L** and Figure S4G). Of note, an absence of Fluoro-Jade C staining was found in the 20 day-differentiated motor neurons, indicating that age as an adequate period for molecular study in order to avoid natural stress conditions of in vitro analysis.

#### Microarray Analysis

Microarray analysis demonstrated differentially expressed genes in the hiPSC-derived motor neurons from sporadic ALS and non-ALS subjects (Table S1). Statistical analyses have identified 1591 deregulated genes in differentiated motor neurons from sporadic ALS compared to control subjects. Of these 1591 genes, 526 were up-regulated and 1065 down-regulated in sporadic ALS compared to non-ALS subjects. The complete list of differentially expressed genes can be found in Supplementary Table S1.

#### Gene Ontology Terms Grouped by REVIGO

Differentially expressed genes based on GO and grouped by REVIGO could be characterized in clusters according to molecular function, e.g., mRNA binding, GTP binding, transcription (co)-repressor activities and low-density lipoprotein particle receptor binding (Figure S5); as well as gene clusters representing various biological processes such as macromolecule catabolism, synapse organization, aging, intracellular transport, negative regulation of DNA binding, death, cell death, mitotic cell cycle and response to inorganic substance (Figure S5). Detailed results for GO terms regarding the GO numbers and number of genes as well as the representative genes in each of these categories can be found in Tables S2, S3 in Supplementary Material. The Cellular Component indicated four GO term related with mitochondrion, specifically the mitochondrion (105 genes), mitochondrion part (65 genes), mitochondrion matrix (30 genes), and mitochondrion lumen (30 genes). The 105 genes of mitochondrion GO term are overlapped with all other GO terms (**Table 5**). The mean of gene expression normalized signal of 105 genes related to the mitochondrion of differentiated motor neurons from sporadic ALS and non-ALS samples are shown in **Figure 6**. Of note, the pattern of deregulated gene expression was consistent among the 3 lines of differentiated motor neurons of ALS and non-ALS groups, as exemplified by the absence of statistical differences of the mitochondrion gene expression within the each one of the two studied groups.

### KEGG Analysis

Enriched KEGG terms were identified among differentially expressed genes (p < 0.05) in hiPSC-derived motor neurons of sporadic ALS and non-ALS subjects (**Table 6**). Importantly, differentially expressed genes associated with Parkinson's disease and oxidative phosphorylation are present as over-represented KEGG pathways. These pathways, and respective deregulated genes associated with them, are seen in **Table 6**. The prostate cancer pathway was omitted from the table because it was composed of genes unrelated to ALS.

### Verification of Microarray Results by qPCR

The results of qPCR for the 10 representative genes deregulated in differentiated motor neurons from sporadic ALS compared to non-ALS subjects are shown in **Figure 7**. Both the up (SDHAF1; 4.88-fold; CASP9; 7.64-fold; VPS35; 8.59-fold and IGF2; 13.52-fold) and down (DUSP6; –3.24-fold; DGAT1; – 1.23-fold; KCNK12; –2.97-fold; KIFC1; –1.07-fold; KRTAP4-11; – 2.46-fold; and LETM1; –3.03-fold) regulation observed for these selected genes, as determined by qPCR, were coincident with and supported our microarray results.

### Discussion

Sporadic ALS is a complex disorder and patients present with a wide range of diverse clinical outcomes regarding disease

layers in vitro and in vivo. Immunocytochemical staining of anti-Pan cytokeratin (A; Pck-26, a marker of ectoderm layer), Myosin Light Chain (LC) 2 (B; a marker of mesoderm layer) and Ephrin A2 (C; a marker of endoderm layer) after spontaneous differentiation in vitro of human-induced pluripotent stem cells (hiPSC). hiPSC were derived from fibroblasts which were obtained

onset, rate of progression and survival (Burkhardt et al., 2013). High throughput gene expression analysis using microarrays has not yet been described for iPSC-derived motor neurons from sporadic ALS patients, thus contributing to a lack of understanding of the mechanisms underlying neurodegeneration in this most common form of that human disease. The absence of clear knowledge regarding the etiopathogenesis and physiopathology of human sporadic ALS has been considered the major cause for failure to translate potential therapeutic strategies to the clinic. Indeed, most experimental studies resulting in therapeutic targets rely fundamentally on experiments using animal models carrying mutated genes of familial ALS (Beghi et al., 2007; Musarò, 2010). Of note, the patients of this study showed an early stage of ALS development, and the biopsy of the motor nerve was performed in a still functioning limb segment, thus the size of samples had to follow the roles of human ethical research. Despite that limitation, the results demonstrated the absence of internal variability.

#### Generation of hiPSC-derived Motor Neurons to Study Molecular Mechanisms in Sporadic ALS

Peripheral motor nerves of ALS patients were used for the first time to provide adult fibroblasts, in contrast to numerous previous studies using skin fibroblasts (Almeida et al., 2013; Zhang et al., 2013; Devlin et al., 2015), to be reprogrammed to hiPSCs. Motor nerves, as a source of adult cells for reprogramming, represent a logical choice with which to model ALS particularly since retraction of motor axons and (blue; A–C). Hematoxylin and eosin staining of sections from a teratoma formed after a subcutaneous injection of hiPSC into nude rats (D–F). Neural rosettes (arrow, D), muscle tissue (arrow; E), and gut-like epithelium (arrow, F) indicate tissues into the teratomas with ectoderm, mesoderm and endoderm features, respectively. Scale bars: 50µm.

denervation of the neuromuscular junction are the earliest events in the disease (Fischer et al., 2004; Schaefer et al., 2005; Pun et al., 2006). In fact, recent evidence describing neuronal dye-back mechanisms, combined with our own unpublished data, suggest that peripheral cells associated with motor axons, such as Schwann cells, may play a role in motor neuron degeneration in ALS (Keller et al., 2009; Maximino et al., 2014), particularly in the sporadic form. Furthermore, previous microarray studies also showed a differential pattern of gene expression in peripheral fibroblasts from ALS patients (Raman et al., 2015). These non-neuronal cells associated with peripheral motor nerves may provide additional information related to the pathogenic mechanisms of sporadic ALS once hiPSC-derived motor neurons could be produced from them. Of note, the growth and differentiation of peripheral motor nerve fibroblastderived hiPSCs shown in this work seemed to be comparable to those hiPSCs derived from adult skin fibroblasts described elsewhere (Chestkov et al., 2014).

In support of this approach, in the present study, we initially generated integration-free iPS cell lines using Sendai virus -based individual delivery of each of the four Yamanaka reprogramming factors. However, a greater efficiency was achieved by using STEMCCA Cre-Excisable Constitutive Polycistronic Lentivirus. The combination of all factors in a single transcript using STEMCCA technology may be responsible for the increased efficiency of reprogramming (Awe et al., 2013; Kadari et al., 2014). In our study, this may have allowed the efficient production of hiPSCs from peripheral motor nerve fibroblasts

(embryoid bodies; B). Axonal lengths project from the clusters in an adherent culture (C). The hiPSC-derived motor neurons are showing classical morphology with axonal projections (D). Scale bars: 50µm.

and may have contributed to the subsequent success of motor neuron differentiation. Moreover, as far as we know, it is the first time that the STEMCCA methodology has been applied to generate hiPSC-derived motor neurons.

The potential for motor neuron differentiation achieved by means of the embryoid body method has been previously discussed by Hu and Zhang (2009). The neuronal differentiation process was initiated with the formation of cell aggregates in suspension (free-floating spheres). Furthermore, the subsequent stages involved in the acquisition of the precise motor neuron identities established by the neural progenitor cells, in a process that recapitulated those defined during normal mammalian embryonic development (Dasen and Jessell, 2009). The first step is the acquisition of a "spinal cord" character, i.e., caudalization, by means of extrinsic signals, i.e., retinoic acid. The cell specification along the dorsoventral axis, i.e., ventralization, occurs in response to Shh signaling (Jessell, 2000; Li et al., 2008). After that, differentiating neurons are maintained in the culture medium supplemented with specific neurotrophic factors to undergo further maturation, including the ability to produce specific neurotransmitters and to show specific morphological characteristics of adult motor neurons (Nizzardo et al., 2010). Because our objective was to assess early changes in the gene expression profile of differentiated motor neurons of sporadic ALS patients, microarray analyses were performed on day 20 of differentiation, a time point at which obtained motor neurons already expressed the specific markers ChAT and Hb9. Of note, we did not perform any electrophysiological tests related to mature neuron function on our iPSC-derived motor neurons. However, despite the OLIG2 expression shown by qPCR, the majority of differentiated motor neurons showed ChAT and also Hb9 markers at that culture age. Also, the cells possessed morphology of mature cells at that time. Thus, the results of CHAT and HB9 by qPCR together with high number of ChAT and Hb9 positive cells in our differentiated motor neuron culture conditions emphasize the state of early mature motor neurons in our system. Furthermore, the expression of OLIG2 seen in our system has been mentioned previously. In fact, expression of progenitress markers can be encountered in different types of transformed mature cells (Lian et al., 2012). The same should be true for IPSC-derived mature motor neurons because the coexpression of OLIG2, HB9, and CHAT were found at different stages of maturation (Hester et al., 2011; Jha et al., 2015), a process that may last for several weeks in culture. Nevertheless, the persistent presence of OLIG2 expression in motor neurons may not impair the differentiation process because generated motor neurons over-expressing Olig2, which were generated by a pLPCX-Olig2 retroviral transfection, were able to acquire morphological, biochemical and electrophysiological properties of mature cells (Lee et al., 2014). Nevertheless, it should be a matter of a next investigation whether a non-excisable system, as it is the case of the present study, may be responsible for the presence of OLIG2 in the mature differentiated motor neurons. Of note, despite the high number of differentiated motor neurons found in the coverslip, that sample analysis should not be taken as a parameter of reprograming/differentiation efficiency, which was not the scope of the present paper. A highly variable efficiency has been described in the literature, which is due to several factors related to differences of the protocols, iPSCs clones, substratum and also the limitation of the methodology to evaluate the final results (Hu and Zhang, 2009; Takazawa et al., 2012; Amoroso et al., 2013; Qu et al., 2014; Toli et al., 2015). Our qPCR analysis on regulation of transcripts of differentiated motor neuron markers may be a better method to document the type of cultured cell RNA that was evaluated in the microarray.

Furthermore, we decided to evaluate the gene profiling in the 20 day cultured ChAT and Hb9 expressed differentiated motor neurons in order to minimize the stress condition of in vitro assay and also because we were interested in the early molecular changes in the disease that could be related to the triggering of motor neuron death. Our decision to study molecular regulation on early mature differentiated motor neurons was also based on the fact that cultured motor neurons from ALS mouse start showing morphological alterations after 14 days in culture (Nagai et al., 2007), the expression of the mature marker may decrease in these cells from culture days 20–30 (Hester et al., 2011), and also because human derived motor neurons could show molecular instability in the process of maturation (Almeida et al., 2013). Despite the early age of differentiated mature motor neurons, that time point seems to be a safer stage to study molecular feature of the disease model in reprogrammed cells. Nevertheless, a detailed analysis of molecular changes at different phases of motor neuron maturation should be a matter of a future investigation.

are arranged in clusters and show axonal projections. Cells are positive for MAP2 (A; green), ChAT (B; red) and express GFP under control of motor neuron-specific gene Hb9 (C). The negative controls for each motor neuron marker mentioned above are seen in (D–F), respectively, and no specific labeling was found. Human induced pluripotent stem cells (hiPSC)-differentiated mature motor neurons from ALS patients (G) and their non-ALS controls (H) are arranged in clusters and show axonal projections. Number of differentiated motor neurons was quantified by the of differentiated motor neurons and neural progenitor cells (K) contained in the 24-well plates. The markers of PAX6 (neural progenitor stage), OLIG2 (motor neuron progenitor stage), CHAT (mature motor neuron stage) and GAPDH (endogenous control gene) are evidenced. The expression of PAX6, OLIG2, CHAT, and HB9 was quantified in differentiated motor neurons contained in the 24-well plates by qPCR and the graphic shows the relative fold-change values; a pool of the neural progenitor cells was used as reference samples with a reference value of 1 (L).



The Cellular Components GO terms mitochondrion (0005739), mitochondrial part (0044429), mitochondrial matrix (0005759) and mitochondrial lumen (0031980) were composed by 105, 65, 30, and 30 genes, respectively. All genes of GO terms mitochondrion, mitochondrial part, mitochondrial matrix and mitochondrial lumen were also present in the GO term mitochondrion.

ANOVA followed by Tukey post-test, whereas differences between non-ALS and ALS groups were analyzed according to unpaired t-test \*\*p < 0.01.

The adequacy of sporadic ALS-derived iPSC-differentiated motor neurons for molecular studies is also highlighted by our results of the large gene profiling analysis, which is in line with profiles previously described using cells or tissues from post-mortem human ALS subjects or from animal models (Olsen et al., 2001; Dangond et al., 2004; Malaspina and De Belleroche, 2004; Jiang et al., 2005; Perrin et al., 2005; Ferraiuolo et al., 2007, 2011; Yamamoto et al., 2007; Offen et al., 2009; Brockington et al., 2010; D'Arrigo et al., 2010; Guipponi et al., 2010; Saris et al., 2013). Of note, we have not performed the excision of inserted cassette in our hiPSC cell lines to avoid an additional step of cell manipulation, what has not impaired the differentiation of mature Hb9 expressing motor neurons. There is a lack of detailed information regarding the presence and effects of residual reprogramming transgene expression the iPSCs-derived motor neurons. Based on the fact that residual reprogramming transgene expression from integrated viruses may alter the biological properties of iPSCs (Sommer et al., 2012), future work should analyze the gene profiling in hiPSC-derived motor neurons after silencing the exogenous factors. Despite the novelty of the methodological approach, the results of the present gene profiling in the differentiated motor neurons should be taken into attention before in vivo experiments because it is difficult to control all consequences of cell reprograming and also to avoid the stress conditions of any in vitro studies.

Importantly, the present gene profiling analysis of iPSCderived motor neurons from sporadic ALS patients provides evidence of possible novel and autonomous mechanisms underlying neurodegeneration in the most abundant form of human ALS. It should be emphasized that previous descriptions on the regulation of gene expression in ALS are based solely on studies with animal models or human post-mortem tissue.

#### GO Molecular Function and Biological Processes from Sporadic ALS Motor Neuron Deregulated Genes

The microarray analyses identified 1591 deregulated genes in differentiated motor neurons from sporadic ALS patients compared to control subjects. The differentially expressed genes with a p-value lower than 0.05 were submitted to enrichment analyses based on GO and KEGG databases, according to our previous publications (de Oliveira et al., 2013; Maximino et al., 2014). Differentially expressed genes based on GO and grouped

#### TABLE 6 | KEGG pathways obtained from microarray analysis.


Microarray analyses of hiPSC-derived motor neurons from sporadic ALS and non-ALS subjects. Positive and negative values correspond to up regulation and down regulation fold changes, respectively. The significance of enrichment (EASE score) was considered p < 0.05. The Parkinson disease and oxidative phosphorylation pathways were obtained. The gene names and symbols are seemed.

by REVIGO displayed molecular functions and biological processes consistent with mechanisms potentially contributing to the autonomous motor neuron cell death characteristic of sporadic ALS.

A wide range of terms of GO molecular function and biological processes grouped by REVIGO have already been implicated in ALS etiopathology. Moreover, their wide ranging and diverse activities underline the high degree of complexity of molecular and biochemical events taking place in the sporadic form of this disease. For instance, within the mRNA binding cluster (Takanashi and Yamaguchi, 2014; Štalekar et al., 2015), hnRNPA1, which codes for a heterogenous nuclear ribonucleoprotein A1, was up-regulated, and has been related to frontotemporal dementia and ALS (Seelen et al., 2014). The GTP binding cluster was previously correlated with ALS (Otomo et al., 2003; Droppelmann et al., 2014). However,

the possible involvement of the down regulation of FKBP-4 in the cell-autonomous neurodegeneration characteristic of the sporadic form of ALS is an original contribution of this study. Furthermore, the deregulated FKBP-4, which belongs to the immunophilin chaperone protein family, may contribute to anti-aggregation processes of proteins widely described in ALS (Steiner et al., 1997; Gold et al., 1999; Manabe et al., 2002). Of note, the transcription (co)-repressor activity cluster (Tan et al., 2012; Dini Modigliani et al., 2014) revealed by REVIGO, identified NKX2 as an up-regulated gene product. NKX2 encodes a natural antisense RNA that modulates the expression of sense transcripts and mRNA processing, and might be involved in intracellular processes related to neuronal degeneration in sporadic ALS (Werner and Sayer, 2009). The synapse organization cluster (Shefner et al., 2006; Venkova et al., 2014) contains up-regulated MAP1B, which encodes a molecule linked to microtubule and synaptic stabilization and was found altered in spinal cords from ALS patients (Coyne et al., 2014).

Perhaps most significantly, REVIGO analysis has revealed a possible critical role for mitochondria in the diverse mechanisms contributing to neurodegeneration in ALS. Of note, the synapse organization cluster included the mitochondrion organization category in which the strongly deregulated genes, LETM1 and SDHF1, were identified (discussed below). Also, the intracellular transport cluster (Alami et al., 2014; Foran et al., 2014) contains the highly down-regulated gene, APOE, whose product is involved in mitochondrial oxidative stress in ALS, and its variants have been correlated to an increased risk of bulbar-onset of human ALS (Praline et al., 2011). Furthermore, the intracellular transport category also contains the significantly down-regulated KIF1A and KIFC1 genes (also included in several GO terms) whose protein products are responsible for anterograde axonal transport of mitochondria (De Vos et al., 2008; Hou and Yang, 2013), a process that is altered in ALS (De Vos et al., 2008). Additionally, the mitotic cell cycle cluster (Cova et al., 2010) revealed an up-regulation of DCTN1, encoding the p150 subunit of the axonal transport protein dynactin. Importantly, mutations in this gene have been associated with sporadic and familial ALS (Münch et al., 2004, 2005).

Significantly, REVIGO analysis provided direct evidence of mitochondrial involvement in the autonomous mechanisms underlying neurodegeneration in ALS with data associated with the cell death and death clusters (Martin et al., 2000; Tapia, 2014). These clusters contained the up-regulated gene, ATXN10, which has previously been associated with ALS (Figley et al., 2014), possibly involving mitochondria and activation of caspase 3 (White et al., 2010). Interestingly, the down-regulation of DYNLL1, also associated with this cluster, may be associated with impairment of dynein induced retrograde axonal transport of mitochondria in ALS (Chen et al., 2010).

Admittedly, it is difficult to separate the initial triggering from secondary events associated with stress of oncoming neurodegeneration based solely on results of the GO analysis of molecular function and biological process. Moreover, despite a huge diversity of events revealed by GO analysis, the general processes involving mitochondria may represent possible early autonomous mechanisms associated with sporadic ALS motor neurons. It should be stressed that mitochondrial functions also appear in several terms associated with cellular components of GO (data not shown).

#### KEGG Pathways from Sporadic ALS Motor Neuron Deregulated Genes

The DAVID analysis identified two specific KEGG pathways that might be related to ALS mechanisms of motor neuron degeneration in sporadic ALS, the Parkinson's disease and oxidative phosphorylation pathways. ALS and Parkinson's disease share some features and mechanisms of neuronal degeneration (Bosco et al., 2011; Shulman et al., 2011), thus emphasizing the relevance of the present molecular analysis employing human motor neurons. In fact, there is growing evidence from work of the past two decades suggesting that ALS may show a complex combination of events at both molecular and cellular levels that are common to Parkinsonism and frontotemporal dementia (Espay et al., 2011). For instance, TARDBP mutations (Mosca et al., 2012), TDP-43 proteinopathy (Mackenzie et al., 2010) and C9orf72 expansion (Lagier-Tourenne et al., 2013) are common to all three pathologies.

The oxidative phosphorylation pathway identified in the KEEG analysis is consistent with mitochondrial participation in sporadic ALS pathology (Ladd et al., 2014). For example, impairments of mitochondrial bioenergetics and energy metabolism, oxidative stress and apoptosis, among other features have been previously described (Cassina et al., 2008; Hedlund et al., 2010; Dupuis et al., 2011; Reddy et al., 2011; Cozzolino and Carrì, 2012; Federico et al., 2012; Brockington et al., 2013; Heath et al., 2013).

The importance of the above two KEEG pathways in the autonomous mechanisms related to neurodegeneration in sporadic ALS is highlighted by recent descriptions of the involvement of their genes in the degeneration of motor neurons in ALS and other neurodegenerative disorders related to ALS. For instance, LRRK2 mutations are associated with Parkinsonism, ALS and dementia (Wszolek et al., 1997; Zimprich et al., 2004), leading to mitochondrial fragmentation, mitochondrial dysfunction and increased reactive oxygen species (Whittle et al., 2007; Wang et al., 2012). Thus, the down regulation of LRRK2 in iPSC-derived motor neurons from sporadic ALS subjects in the present study provides further confirmation of the importance of mitochondrial dysfunction in this neurodegenerative disorder.

The down regulation of ATP5B we have identified in differentiated motor neurons from sporadic ALS patients in the present microarray analysis correlates with previous findings of reduced cytochrome oxidase and ATP synthase activities described in neurodegenerative disorders (Schägger and Ohm, 1995; Bosetti et al., 2002; Sergeant et al., 2003; Basso et al., 2004; Pamplona et al., 2005). For example, ATP5B deregulation was previously characterized in western Pacific ALS-parkinsonismdementia complex (Shiraki, 1975).

Mitochondrial involvement in the autonomous motor neuron cell death in sporadic ALS is also supported by the up-regulation of VDAC1 seen in the microarray analyses. VDAC1 encodes for a voltage dependent anion channel in the outer mitochondrial membrane, and thus is able to regulate local ATP/ADP flux (Colombini, 2004; Lemasters and Holmuhamedov, 2006). Interestingly, VDAC1 is also a key player in mitochondria-mediated apoptosis and has been implicated in apoptotic-relevant mitochondrial events (Shimizu et al., 1999; Shoshan-Barmatz et al., 2006, 2008; Tajeddine et al., 2008; Abu-Hamad et al., 2009; Arbel and Shoshan-Barmatz, 2010). Misfolded mutant SOD1 directly binds and inhibits VDAC1 conductance, thus impacting the onset and survival in the ALS mouse model (Israelson et al., 2010). Furthermore, the deregulation of VDAC1 may also trigger neuronal damage by a direct interference in the regulation of mitochondrial reactive oxygen species (Liu et al., 2004; Vande Velde et al., 2004) or by a direct activation of mitochondrial apoptotic pathways (Tsujimoto and Shimizu, 2002; Shoshan-Barmatz et al., 2006; Yagoda et al., 2007; Abu-Hamad et al., 2008).

Additional evidence in support of mitochondrial dysfunction in autonomous sporadic ALS is provided by the demonstration that significant up-regulation of HFE may be associated with iron homeostasis. HFE mutations have been linked to human ALS (Rothstein, 2009; Kwan et al., 2012) and changes in iron metabolism may confer susceptibility to this disease in both humans (Li et al., 2014) and animals (Nandar et al., 2014). Because mitochondria regulate intracellular iron homeostasis (Horowitz and Greenamyre, 2010), HFE deregulation may represent an additional mitochondrial-based pathogenic mechanism in sporadic ALS. Mitochondrial dysfunction leading to cell autonomous neuronal damage in sporadic ALS is also supported by the significant deregulation of both TOMM20L (Ryan and Gogvadze, 1999) and HtrA2/Omi (Srinivasula et al., 2003), as seen in our microarray analysis.

Finally, many of the genes identified as being dysregulated in our microarray analysis have not yet been previously associated with ALS in any context. However, some (eg. ATP5D, ATP5E, NDUFA4, NDUFB5, NDUFB6) have been implicated in other neurodegenerative disorders (Chen et al., 2014; Suszynska- ´ Zajczyk et al., 2014). Additionally, NDUFC2, ND2, COX7C, NDUFV3, SDHC, ATP5C1, COX6A1, ATP4B, ATP6V1D, and COX17 are associated with mitochondrial function or impaired bioenergetic metabolism, and thus further support a role for mitochondrial dysfunction in the cell autonomous degeneration of motor neurons in sporadic ALS (Burman et al., 2014).

### GO Mitochondrion Genes of the Cellular Component

Regulation of some of the mitochondrial genes and their related proteins of the cellular component mitochondrion described in the microarray analysis have been mentioned in the context of neuronal degeneration and ALS. For instance, C1QBP, CAV2, CTSB, DIABLO, GLRX2, NDUFA4, NLRX1, OPA1, SDHC, SIRT3, SURF1 have been correlated to neuronal death and to excitotoxic events (Manfredi and Beal, 2000; Weiergräber et al., 2007; Hayashi et al., 2009; Bräutigam et al., 2011; Gray et al., 2013; Tian et al., 2013; Imbeault et al., 2014). Some of them were related to complex cellular mechanisms of neurodegenerative disorders, also involving other cellular organelles as it is the case of CTSB in lysosomal function (Sevenich et al., 2006; Cho et al., 2013). Further indication of the importance of mitochondrion gene regulation in the mechanisms of ALS is highlighted with the descriptions of mitochondrial oxidative stress leading SDHC mutation-induced apoptosis (Ishii et al., 2011) and of neuroprotection after inhibition of mitochondrial release of the pro-apoptotic protein Smac/DIABLO (Soustiel and Larisch, 2010; Luan et al., 2012).

The genes ATP5B, E2F1, HTRA2, LETM1, LRRK2, POLG, SIRT3, SOD1, VDAC1 have been specifically described in the context of neuronal death in ALS. For instance, the gene polymerase gamma (POLG), which is related to mitochondrial biogenesis, was investigated in the context of ALS (Ladd et al., 2014). The mitochondrial protein ATP5B was also correlated to pathogenesis of neurodegenerative disorders (Yoon et al., 2009), actually the Western Pacific ALS-parkinsonism-dementia complex (Kisby et al., 2006). Mitochondrial proteins involved in cell-cycle and transcriptional regulation, i.e., E2F1, participate in the neuronal death pathways (Wu et al., 2012) and in ALS (Ranganathan and Bowser, 2003). Also, impairments of the apoptotic intermembrane mitochondrial serine protease HTRA2 lead to its accumulation in motor neuronal inclusions in ALS (Kawamoto et al., 2010) and neuronal death (Patterson et al., 2014). Importantly, the mitochondrial Ca(2+) transporter LETM1 is elevated in death resistant motor neurons of ALS mice. Conversely, LRRK2 was implicated in early neuronal death (Chen et al., 2012; Chou et al., 2014) while SIRT3 may exert neuroprotection by avoiding mitochondrial fragmentation in ALS mouse (Song et al., 2013), indicating the complex role of mitochondria in neurodegenerative diseases (Mühling et al., 2014). Furthermore, dysfunction of the mitochondrial voltage-dependent anion channel (VDAC) was associated with neuroprotection failure (Fernandez-Echevarria et al., 2014) and neurodegeneration in ALS (Israelson et al., 2010). Of note, a specific pathogenic LRRK2 mutation was described in ALS (Whittle et al., 2007). It should be noticed that the role of SOD1 in mitochondrial mechanism related to ALS has been largely described (Dupuis et al., 2004).

Finally, the PSMC4, GRN, and ND2 encoded genes of mitochondrial proteins have also been described in ALS, as it is the case of the 26S proteasome regulator PSMC4 (Trippier et al., 2014). Interestingly, a co-occurrence of GRN/C9ORF72 (Testi et al., 2015) and of NADH dehydrogenase subunit 2 (ND2) mutations were also detected in ALS patients (Lin et al., 1992).

It should be noticed that mutations of CHCHD10 have been described in ALS, a gene that encodes an intermembrane mitochondrial protein involved in mitochondrial genome stability and morphology (Bannwarth et al., 2014; Johnson et al., 2014).

It is well known that the regulation of mitochondrial gene expression is not restricted to pathological conditions but instead to a large range of cellular physiological conditions. However, we have demonstrated in our work no variations of mitochondrial gene expression within the groups but a significant difference between non-ALS and ALS subjects, emphasizing a possible participation of deregulated genes in the motor neurons disease. Actually, statistical significance between groups and small intragroup variability indicated an homogeneity of the samples.

#### qPCR Verification of Sporadic ALS Motor Neuron Deregulated Genes

The results of the qPCR and microarray analyses correlated well for all selected genes, confirming the efficacy of large gene profiling in the search for therapeutic targets in neurodegenerative disorders (de Oliveira et al., 2013; Maximino et al., 2014). Notably, a relationship between these deregulated genes with various molecular/biological processes related to intracellular trafficking, intercellular signaling, and neurotrophic functions (Vucic and Kiernan, 2009; Martin, 2012; Chaturvedi and Flint Beal, 2013) is supportive of a role for mitochondrial dysfunction in ALS, as previously postulated for other neurodegenerative diseases (Pagano et al., 2014) Furthermore, a critical analysis of the up and down regulated genes by qPCR supports their roles in both the events initiating neuronal damage and subsequent reactive events leading to neuroprotection.

Of note, LETM1, which encodes a mitochondrial Ca2+/H<sup>+</sup> exchanger protein (Waldeck-Weiermair et al., 2011; Nowikovsky et al., 2012) was found to be up-regulated in iPSC-developed motor neurons of sporadic ALS patients. This mitochondrial ion exchange protein has been linked to protection of rescued neurons in the ALS mouse model (Mühling et al., 2014), possibly by its ability to regulate Ca2<sup>+</sup> homeostasis. The high down regulation of LETM1 in sporadic ALS motor neurons supports a role for mitochondrial dysfunction in the cell-autonomous failure of neuroprotection. Moreover, DUSP6 was the most significantly down regulated gene identified by the microarray analyses. This gene encodes for a dual specificity phosphatase-6 protein, which is a critical determinant of ERK signaling (Bhalla et al., 2002), and is thus able to influence cell survival after neurotoxicity (Hetman and Gozdz, 2004). As ERK signaling has been implicated in ALS (Kim and Choi, 2010), a strong downregulation of DUSP6 in sporadic ALS motor neurons is worthy of further investigation. Of note, despite of microarray and qPCR have shown a down regulation of DUSP6, the degree of regulation demonstrated by microarray was higher than that obtained by qPCR analysis The use of qPCR analysis to qualitatively verify the microarray results is largely accepted in the literature. However, small discrepancies in the degree of gene regulation between the two methods have been reported (Ferraiuolo et al., 2007). These methods have their specific quantitative differences (Chuaqui et al., 2002), which are thought to be related to the variation in the hybridization kinetics of the technologies, low fold changes or lack of concordance between transcripts accessed in each method (de Oliveira et al., 2013).

Dysfunction in lipid metabolism has also been associated with ALS (Schmitt et al., 2014; Palamiuc et al., 2015). The present study is the first to demonstrate a significant deregulation of DGAT1, a gene which encodes a multipass transmembrane protein that functions as a key metabolic enzyme in the conversion of diacylglycerol and fatty acetyl CoA to triacylglycerol, in the context of human ALS.

Also, specific processes involving potassium channels have been described in non-autonomous glial mechanisms of ALS (Bataveljic et al., 2012; Sato et al., 2014 ´ ). However, this work contributed to the original description of this potassium channel's (subfamily K, member 12 gene, KCNK12) deregulation in differentiated human motor neurons from sporadic ALS patients.

The contribution of kinesins in the deregulated axonal transport (Landers et al., 2009; Shi et al., 2010; Kuzma-Kozakiewicz et al., 2013) as well as the participation of cytoskeleton components (Maximino et al., 2014) in ALS have been well documented. However, the significant down regulation in the genes, KIFC-1 and KRTAP4-11, point to the possibility of abnormalities in intracellular trafficking and cytoskeletal structural changes respectively, in the differentiated motor neurons from sporadic ALS patients.

The up-regulation of CASP9, which encodes for caspase-9 protein, might activate mitochondria-dependent apoptotic signaling, an event that has been described in neurodegenerative diseases (Andreoli et al., 2009; Darwish and Amiridze, 2010) including ALS (Pasinelli et al., 2000; Guégan et al., 2001, 2002; Inoue et al., 2003). Interestingly, caspase-9 levels have been correlated with clinical severity in ALS patients (Ilzecka, 2012). Considering the disease-specific phenotype of the iPSC-derived motor neurons in the present analysis, an early dysfunction of caspase-9 mediating motor neuron death may take place in sporadic ALS. Furthermore, the up regulation SDHAF1, which encodes for succinate dehydrogenase complex, an enzyme that participates in the respiratory chain and Krebs cycle (Rutter et al., 2010) further implicates mitochondrial participation in sporadic ALS.

Conversely, distinct from the putative roles of CASP9 and SDHAF-1, the up-regulation of IGF2 and VPS35 may indicate the occurrence of reactive autocrine neuroprotective responses in the sporadic ALS-differentiated motor neurons after an initial damage. Thus, it is likely that up-regulation of IGF2 may be involved in neuroprotection of motor neurons (Leventhal et al., 1999; Silva et al., 2009), in line to its previously suggested role in non-autonomous glial ALS mechanisms (Kihira et al., 2007; Dagvajantsan et al., 2008). The fact that IGF-1 treatment led to neuroprotection in an ALS animal model (Jablonka et al., 2011; Saenger et al., 2012) and to partial benefits in ALS clinical trials (Lai et al., 1997; Borasio et al., 1998), further supports our finding that up-regulation of IGF2 expression in the differentiated motor neurons from sporadic ALS patients highlights the importance of IGF family proteins as potential therapeutic targets for this disease. Accordingly, the regulation of VPS35 corroborates our findings with IGF2. The VPS35 gene product is involved in retrograde transport of proteins from endosomes to the Golgi complex (Zhang et al., 2001). VPS35 mutations and disruption of endosomal trafficking has previously been implicated in other neurodegenerative disorders (Small et al., 2005; Zimprich et al., 2011; Ando et al., 2012; Vilariño-Güell et al., 2014; Perrett et al., 2015). Thus, up-regulation of VPS35 might account for a neuroprotective response in ALS before degeneration of motor neurons begins.

In conclusion, the large gene profiling of early stage mature motor neurons differentiated from hiPSCs, which in turn, were derived from adult motor nerve associated fibroblasts from sporadic ALS subjects, has provided evidence for a wide range of molecular and cellular mechanisms possibly involved in the cell autonomous neuronal death in the disease. Of note, evidence supporting impairment of intracellular

### References


trafficking, intercellular signaling, oxidative phosphorylation and neurotrophic factor function among others, strongly implicate mitochondrial dysfunction as a key factor in this cellautonomous neurodegeneration process. Interestingly, evidence is also provided that autocrine neuroprotective mechanisms coexist with those related to cell toxicity in these human sporadic ALS differentiated motor neurons. These results thus emphasize the importance of the use of human iPSC-derived motor neurons, obtained from motor nerve fibroblasts, as a reliable method to model disease mechanisms associated with neuronal degeneration in sporadic ALS. Importantly, the approach may also provide a viable opportunity to translate these results to the development of potential therapeutic targets specific to sporadic ALS.

### Author Contributions

CA, RD, MM, and JM performed the experiments. FJ, GC, RM, and DC selected the patients and performed the biopsies. CA, JM, BS, JK, and GC designed the study, analyzed the results and wrote the manuscript. GC is responsible for the ALS Brazil Project of the University of São Paulo School of Medicine. All authors read and approved the final manuscript.

### Acknowledgments

We thank professor Lygia da Veiga Pereira (Bioscience Institute of University of São Paulo) for providing hESC and valuable advice during the generation of hiPSCs. We thank professor Robert Carlone of Brock University for reading the manuscript and for his valuable suggestions and we also thank Leonardo Jensen and Graça Correa Rosas for helping with the preparation of iPSC cardiomyocytes. This work was supported by grants from the São Paulo Research Foundation (FAPESP; #2010/20457- 7, #2010/20467-2) and National Council for Scientific and Technological Development (CNPq).

### Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2015.00289


caused by a complex I mutation that uncouples proton pumping from electron transfer. Dis. Model. Mech. 7, 1165–1174. doi: 10.1242/dmm.015321


axonopathy: evidence in mice and man. Exp. Neurol. 185, 232–240. doi: 10.1016/j.expneurol.2003.10.004


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Alves, Dariolli, Jorge, Monteiro, Maximino, Martins, Strauss, Krieger, Callegaro and Chadi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# ALS Patient Stem Cells for Unveiling Disease Signatures of Motoneuron Susceptibility: Perspectives on the Deadly Mitochondria, ER Stress and Calcium Triad

Anjoscha Kaus 1,2 and Dhruv Sareen1,2,3 \*

<sup>1</sup> Board of Governors-Regenerative Medicine Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA, <sup>2</sup> Department of Biomedical Sciences, Cedars-Sinai Medical Center, Los Angeles, CA, USA, <sup>3</sup> iPSC Core, The David and Janet Polak Stem Cell Laboratory, Cedars-Sinai Medical Center, Los Angeles, CA, USA

Amyotrophic lateral sclerosis (ALS) is a largely sporadic progressive neurodegenerative disease affecting upper and lower motoneurons (MNs) whose specific etiology is incompletely understood. Mutations in superoxide dismutase-1 (SOD1), TAR DNAbinding protein 43 (TARDBP/TDP-43) and C9orf72, have been identified in subsets of familial and sporadic patients. Key associated molecular and neuropathological features include ubiquitinated TDP-43 inclusions, stress granules, aggregated dipeptide proteins from mutant C9orf72 transcripts, altered mitochondrial ultrastructure, dysregulated calcium homeostasis, oxidative and endoplasmic reticulum (ER) stress, and an unfolded protein response (UPR). Such impairments have been documented in ALS animal models; however, whether these mechanisms are initiating factors or later consequential events leading to MN vulnerability in ALS patients is debatable. Human induced pluripotent stem cells (iPSCs) are a valuable tool that could resolve this "chicken or egg" causality dilemma. Relevant systems for probing pathophysiologically affected cells from large numbers of ALS patients and discovering phenotypic disease signatures of early MN susceptibility are described. Performing unbiased 'OMICS and highthroughput screening in relevant neural cells from a cohort of ALS patient iPSCs, and rescuing mitochondrial and ER stress impairments, can identify targeted therapeutics for increasing MN longevity in ALS.

Keywords: ALS, motoneurons, C9orf72, TDP43, mitochondria, endoplasmic reticulum, calcium homeostasis, iPSCs

### INTRODUCTION

Amyotrophic lateral sclerosis (ALS), or Lou Gehrig's disease, is one of the most common neurodegenerative disorders and adult-onset motoneuron disease (MND), with a global incidence of 2.5 cases per 100,000 individuals each year, and is classified as a rare disease. The disease is primarily characterized by a relentless degeneration of upper motoneurons (UMNs) in the cerebral cortex and lower motoneurons (LMNs) in the brain stem and spinal cord,

#### Edited by:

Manoj Kumar Jaiswal, Columbia University Medical Center, USA

#### Reviewed by:

Angelo Poletti, Università Degli Studi di Milano, Italy Gerson Chadi, University of Sao Paulo School of Medicine, Brazil Sónia S. Leal, Universidade de Lisboa, Portugal

> \*Correspondence: Dhruv Sareen dhruv.sareen@gmail.com dhruv.sareen@cshs.org

Received: 22 July 2015 Accepted: 02 November 2015 Published: 19 November 2015

#### Citation:

Kaus A and Sareen D (2015) ALS Patient Stem Cells for Unveiling Disease Signatures of Motoneuron Susceptibility: Perspectives on the Deadly Mitochondria, ER Stress and Calcium Triad. Front. Cell. Neurosci. 9:448. doi: 10.3389/fncel.2015.00448 causes muscle weakness, progressive paralysis, speech and swallowing disabilities, and ultimately leads to death by respiratory failure (Al-Chalabi and Hardiman, 2013). A definite cause for ALS is not known except for a minority of cases with a familial history pointing to a genetic mutation as an origin (**Table 1**). ALS exists in two forms: heritable (familial, fALS) and idiopathic (sporadic, sALS), with the former featuring autosomal dominant or recessive inheritance patterns in 10% of total ALS cases and the latter represent the overwhelming majority of ALS cases (Cozzolino and Carri, 2012).

Since the discovery of the first gene responsible for familial forms of ALS, superoxide dismutase-1 (SOD1), fundamental advances in our understanding of the disease have been made with regard to biological, genetic and clinical processes (Cozzolino and Carri, 2012). Currently, ALS is considered a multi-factorial syndrome with motor system degeneration only being part of a more widespread disease process (Ilieva et al., 2009). Although, it is still not understood why particular populations of neurons are particularly vulnerable, increasing evidence over recent years emphasized that in addition to MNs, other cell types such as astrocytes, microglia, oligodendrocytes (OGs) and muscle cells may also actively participate in disease processes, such as neuroinflammation and muscle degeneration (Robberecht and Philips, 2013).

An ever-growing number of genes have been identified that are associated with ALS pathogenesis and account for a large fraction of familial cases of the disease, even in sporadic forms (**Table 1**). Interestingly, many of the same genetic mutations reported to be involved in familial and some sporadic forms of ALS have also been found to be associated with non-MN phenotypes, such as dementia in frontotemporal lobar degeneration (FTLD; Cooper-Knock et al., 2014), emphasizing a great overlap in the molecular pathogenesis of both disease entities (Goldstein and Abrahams, 2013).

In recent years the major focus of ALS research has moved to RNA related control of MN function and pathogenesis. Mutations in transactive response (TAR) DNA-binding protein (TARDBP or TDP43; Kabashi et al., 2008) and fused in sarcoma/translocated in liposarcoma (FUS/TLS, ALS6; Kwiatkowski et al., 2009; Vance et al., 2009) have been linked to fALS. The two proteins are involved in the regulation of RNA transcription, splicing, transport and translation and this enabled the establishment of new experimental models, both, in vivo and in vitro (Sreedharan et al., 2008; Vance et al., 2009;


A list of genes associated with classical and atypical ALS. The left column groups the genes according to their cellular functions, the second to left column lists the corresponding chromosomal locus, and the middle column gives the gene name, if applicable. In the second to right column the time of disease onset and means of inheritance are displayed, while the right column links each gene to the disease phenotype in the clinic. AD, autosomal dominant; AR, autosomal recessive; FTD, frontotemporal dementia; PD, Parkinson's disease.

Zhou et al., 2010a). A more recently identified gene, C9orf72, is profoundly affected by expansions of GGGGCC (G4C2) repeats. C9orf72 encodes for a protein with two different isoforms: (i) a mainly diffuse in the cytoplasm localized C9-L and (ii) a nuclear membrane localized C9-S (Xiao et al., 2015). The latter one shows redistribution to the cytoplasm of diseased MNs in ALS and an interaction with the nuclear pore complex components importin β1 and Ran-GTPase. The finding of reduced levels of at least one C9orf72 transcript in expanded-repeat-carriers suggests a potential loss-of-function mechanism (Dejesus-Hernandez et al., 2011; Renton et al., 2011; Gijselinck et al., 2012; Donnelly et al., 2013; Sareen et al., 2013). Alternatively, the accumulation of transcripts containing the G4C<sup>2</sup> transcripts as nuclear RNA foci are considered to confer the mutant gene with a toxic feature via an RNA-dependent gain of function mechanism (Dejesus-Hernandez et al., 2011; Achsel et al., 2013).

These three genes alone account for more than half of the reported fALS cases making RNA dys-metabolism one of the central issues of ALS pathogenesis. Several additional genes have been found to cause rare or atypical forms of fALS (**Table 1**). Nevertheless, based on their biological role and acquired cellular phenotypes, mutations in these genes have been also linked to oxidative stress, protein-misfolding and aggregation, endoplasmic reticulum (ER) and cytoskeleton alterations, ubiquitin proteasome pathway malfunctions, glutamatemediated excitotoxicity, calcium (Ca2+) imbalance, and axonal transport defects (Cozzolino et al., 2012). Intriguingly, whether and how these recently described nuclear pore-mediated and RNA dysmetabolism-related pathways are intricately linked to oxidative stress pathways and mitochondrial damage phenotypes observed in many of ALS experimental models should be a focus of further studies.

Mitochondria play a key role in cellular respiration by converting nutrients into ATP thereby providing cellular processes with energy. They are also the main source of reactive oxygen species (ROS) and act as gatekeepers in intrinsic apoptotic pathways. Mitochondrial dysfunction can lead to oxidative stress, failure of cellular bioenergetics and ultimately to cell death (**Figure 1**). Thus, an alteration of their properties could confer an intrinsic susceptibility to stress and aging of long-lived post-mitotic MNs in MNDs (Cozzolino et al., 2008; Cozzolino and Carri, 2012). ALS patient tissue and animal models exhibiting mitochondrial alteration and dysfunction have frequently been found to also exert ER stress. As the cellular compartment where secreted and membrane proteins are synthesized and folded, the ER is equipped with foldases, chaperones and co-factors to process these proteins and to prevent misfolding or aggregation. Stress conditions can interfere with ER function and result in abnormal folding and aggregation of proteins as has been observed in TDP-43 and FUS/TLS mutations (Andersen and Al-Chalabi, 2011; Turner et al., 2013) thus provoking a state of ER stress (Boillée et al., 2006a; Pasinelli and Brown, 2006; Matus et al., 2013; **Figure 2**). Given the large size and long neuritis of MNs, mitochondria and ER dysfunctions significantly interfere with their normal electrophysiological function as described in the following chapter.

Despite enormous advances in understanding the pathogenic events occurring in ALS, only a small fraction of all reported cases can be linked to genetic causes (**Table 2**). Thus the cause of sALS is largely unknown, and while environmental parameters have been a subject of many epidemiological studies (Seelen et al., 2014; Ingre et al., 2015; Malek et al., 2015), unidentified genetic factors are also thought to be a contributing factor towards susceptibility of MN degeneration. Considerable efforts have been made to unravel the genetic etiology of ALS with genomewide association studies and other next-generation sequencing approaches (Renton et al., 2014; Cady et al., 2015). Whole exome sequencing studies, for instance, have identified de novo risk genes and pathways, i.e., TBK1 (Cirulli et al., 2015), thereby shedding new light on established ALS associated gene mutations and how they converge in pathways in fALS and sALS.

In this review, we will outline the main factors that have been associated with ALS disease progression and pathology, the impact of organelle dysfunction underlying dysfunction in MNDs. Mitochondrial dysfunction and ER stress have been implicated as major events in MN degeneration and we will highlight findings from animal and cell-based models and how those contribute to degenerative phenotypes of neuronal and non-neuronal cells. With an increasing number of studies utilizing patient-derived induced pluripotent stem cells (iPSCs) to recapitulate disease phenotypes we will outline some of the observed phenotypes and to what extent they resemble hallmarks of disease progression observed in the clinic.

### MOTONEURONS INVOLVED IN ALS

The risk of developing ALS peaks between 50–75 years of age, which is in contrast to other neurodegenerative diseases such as Parkinson's disease (PD) or Alzheimer's disease (AD), thus emphasizing that it is not necessarily a disease of aging, but that age is one of the chief risk factors. In a majority of patients, ALS manifests as asymmetric, painless weakness in a limb, referred to as spinal onset. About 20% of patients present with symptoms of weakness in bulbar muscles, difficulties swallowing (dysphagia) and speech impairment, referred to as bulbar onset. In about 3–5% of patients, ALS is characterized by onset affecting the respiratory system, with shortness of breath and difficulty breathing (orthopnea and dyspnea). While these symptoms of onset can be overlapping, the average survival is 2–5 years, with bulbar and respiratory onsets having a worse prognosis (Swinnen and Robberecht, 2014).

MN diversity is generated during CNS development by members of the hox gene family and an array of inductive signaling pathways. Although, it is not yet completely understood how different subtypes of MNs are specified during development, work over the past decades has significantly advanced our understanding of hox genes conferring regional identity to MN groups along the rostro-caudal axis (Philippidou and Dasen, 2013). With over 300 bilateral pairs of muscles comprised of over a 100 million muscle fibers, over 120,000 MNs in the spinal cord alone account for sensation of muscle, contraction, respiration and other vital functions. Upper MNs (UMNs) that

are affected in ALS originate either in the motor region of the cerebral cortex or in the brainstem and carry motor information down to the lower MNs (LMNs). LMNs are either located in the anterior gray column and anterior nerve roots—termed spinal LMNs, or the cranial nerve nuclei of the brain stem and cranial nerves with motor function, summarized as cranial nerve MNs. Voluntary movements rely on spinal LMNs, which innervate skeletal muscle fibers, thereby acting as a link between UMNs and muscles. In turn, cranial nerve LMNs control

input from their corresponding MN and are prone to atrophy, manifesting in increased muscle weakness of ALS patients.

movement of the eyes and tongue as well as contributing to swallowing, chewing and vocalization (Swinnen and Robberecht, 2014).

Based on the type of muscle LMNs innervate, they can be further classified into alpha-, beta-, and gamma MNs. Alpha MNs (αMNs) innervate extrafusal muscle fibers, the most abundant type of muscle fiber and the type being involved in muscle contraction. Beta MNs (βMNs) innervate intrafusal fibers of muscle spindles with collateral extensions to extrafusal slow

factor remains to be determined. Activated astrocytes and microglia can trigger inflammatory responses resulting in further MN stress. Finally, a prion-like hypothesis posits that misfolded protein aggregates can spread from surrounding cells, either affected dying neurons or astrocytes, and infects MNs thereby initiating stress responses and ultimately apoptosis in the infected cells.

twitch fibers. Gamma MNs (γMNs) innervate intrafusal muscle fibers and together with sensory afferents, compose a muscle spindle and are part of proprioception, the sensing of body positioning. According to the contractile properties of the motor units (MU) they form with their target muscles, αMNs can be further classified into fast-twitch fatigable (FF), fasttwitch fatigue-resistant (FR) and slow-twitch fatigue-resistant (SR; Burke et al., 1973).

### Subtypes of MNs are Differently Affected in ALS

Not all MN subtypes are affected equally in ALS. In SOD1 transgenic mice the FF MU undergo atrophy the earliest, with a nearly total loss of FF terminals from muscle fibers in the triceps area (Frey et al., 2000). The almost synchronous dieback is followed by delayed, but likewise rapid fallout of the FR units, while S MU remain well preserved until late in disease progression (Pun et al., 2006). Electrophysiological studies and fiber type analysis confirm the overall sequence of degeneration; however, they also suggest an initial switch in MU phenotype from FF to FR to precede the loss of FF axons (Kieran et al., 2005; Hegedus et al., 2008; Gordon et al., 2010). The relative resistance observed in FR and S MU may by explained be their high sprouting capacity, which could enable them to innervate motor endplates that have been vacated by FF axons (Frey et al., 2000). In accordance with this electromyogram (EMG) patterns reflect cycles of denervation/reinervation (de Carvalho et al., 2008) and the FF MU being affected first in sALS patients as well (Dengler et al., 1990).

A selective vulnerability of particular neuronal subsets to the neurodegenerative process is a key feature, not only of ALS, but of other neurodegenerative diseases. Why certain subtypes of MNs are particularly susceptible to injury in the presence of mutations affecting ubiquitously distributed proteins such as TDP-43 and SOD1, is not well understood. The cellspecific features of LMNs, i.e., a large cell size, long axons and large terminal arbors, have great implications for their energy metabolism, the requirement of mitochondrial and cytoskeletal support, and the distribution of mRNA for protein synthesis (Ferraiuolo et al., 2011). LMNs vulnerable to injury in ALS exhibit particular sensitivity to glutamatergic toxicity via a-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptor activation. Unlike other neuron groups, they exert high expression levels of calcium-permeable AMPA receptors that lack the GluR2 subunit (Williams et al., 1997) and have

#### TABLE 2 | Animal models of ALS.


A representation of some of the vertebrate and invertebrate models utilized to study the cellular pathology of MND. The left column group's spontaneous mutations in animal models found to closely resemble MND pathology, Mendelian and atypical forms of ALS associated mutations. In the middle column disease associated genes are listed. The right column summarizes how far ALS-associated phenotypes resemble the designated models.

fairly low amounts of cytosolic calcium-buffering proteins (Ince et al., 1993). In addition these MNs have a high threshold for initiating a protective heat-shock protein response, an increased sensitivity to ER stress (Saxena et al., 2009) and mitochondrial features that predispose them to calcium overload and oxidative damage (Sullivan et al., 2004; Panov et al., 2011).

Data obtained in ALS patients and animal models alike initially suggested an early axonal dieback and selective susceptibility of FF axons to occur during human disease progression (Fischer and Glass, 2007). With a selective dieback of FF MU in ALS, intrinsic molecular mechanisms conferring susceptibility to these units lie at hand. In case of the many different point mutations reported for mutant SOD1, research has mainly focused on how misfolded SOD1 toxicity impacts the cell. For instance excessive accumulation of misfolded proteins leads to stress responses in the ER in an effort to restore cellular homeostasis. When this process is not successful, as is the case in SOD1 mutants, ER pathways can trigger apoptosis (Matus et al., 2013). Interestingly, a recent study targeting mSOD1 in UMN via an AAV9-shRNA in a presymptomatic SOD1 (G93a) rat model reported significant delay of disease onset and improved survival of spinal MNs (Thomsen et al., 2014). These findings suggest differential vulnerability of MN subtypes alongside early UMN dysfunction during the presymptomatic phase of disease course.

With the recognition of non-neuronal cells such as glia and activated microglia also have a major contribution to the progression and outcome of the disease three scenarios have been hypothesized: (I) The ''dying forward'' hypothesis suggests that events within the MN are the primary determinant initiating MN damage, while alterations in closely associated non-neuronal astrocytes, Schwan cells, microglia and skeletal muscle cells only contribute to the disease progression; (II) In a counter-proposed ''dying backward'' mechanism, ALS is considered to begin within the muscle cells and/or neuromuscular junction as a result of one or more neurotrophic factors being released by postsynaptic cells. When retrogradely transported via the presynaptic axon to the cell body those factors may initiate MN damage and degradation (Dupuis and Loeffler, 2009); and (III) Alternatively, the ''prionlike propagation'' scenario could explain the focal onset and propagation of misfolded ALS protein seeds via the neuronal network or the extracellular matrix (Polymenidou and Cleveland, 2011).

### CONTRIBUTION OF OTHER CELL TYPES TO ALS

### GABA-Ergic Interneurons

During the last 20 years, structural and functional neuroimaging studies have greatly revised our longstanding notions of the pathophysiology of ALS. In fact, anatomical lesions extend beyond precentral cortices and corticospinal tracts, and include the corpus callosum, the frontal, sensory and premotor cortices, thalamus and midbrain. Theories of glutamate excitotoxicity and hyperexcitability in ALS suggest the contribution of excessive synaptic excitation, for instance through AMPA receptors (Kwak et al., 2010) and Na<sup>+</sup> channels. Disease-specific changes in RNA editing of the AMPA glutamate receptor subunit 2 (GluR2), for instance, have been found in spinal MNs of human sALS autopsy cases (Kawahara et al., 2004). Clinical data, however, suggest the possibility of insufficient synaptic inhibition. Magnetic resonance imaging (MRI) and positron emission tomography (PET) studies in patients revealed an early and diffuse loss of inhibitory cortical interneurons and diffuse gliosis in white matter tracts (Chió et al., 2014). In fact, MN excitability is strongly modulated by synaptic inhibition mediated by presynaptic glycinergic and GABAergic innervations and postsynaptic glycine receptors (GlyR) and GABA<sup>A</sup> receptors (Martin and Chang, 2012).

Abnormal glycine and GABA levels have been observed in ALS patients (Malessa et al., 1991; Niebroj-Dobosz and Janik, 1999; Eisen and Weber, 2000). In human ALS autopsy samples of the spinal cord, glycine binding sites are reduced in the anterior horn and in human ALS autopsy motor cortex (Hayashi et al., 1981; Whitehouse et al., 1983). In agreement with this, mRNA levels of the alpha1-subunit of the GABA receptor of inhibitory interneurons was significantly reduced, while the GABA synthesizing enzyme glutamic acid decarboxylase (GAD), in turn, was dramatically upregulated in the prefrontal and temporal cortices of ALS patients (Petri et al., 2006). In the spinal cord of adult transgenic G93A-hSOD1 transgenic mice (Schutz, 2005), as well as organotypic spinal cord-slice cultures of embryonic G93A-hSOD1 transgenic mice (Avossa et al., 2006), signs of imbalance excitatory and inhibitory innervations, indicative of aberrant neuroplasticity, are evident. Thus aberrant or even failed inhibition may be involved in ALS mechanisms of disease.

### Glia—Microglia

The immune system plays a crucial role in the maintenance of tissue homeostasis and the response to injury or infection. Within the central nervous system (CNS) microglia, the major resident immune cells, constantly survey the microenvironment and produce factors that influence other glial cells such as astrocytes. While microglia remain in a deactivated state during physiological conditions (Streit, 2002) they switch to an activated state upon pathogen invasion or tissue damage detected by their Toll-like receptor 4 (TLR4) and promote inflammatory responses via mediators such as TNF-α or IL-1β, engaging the immune response and tissue repair (Saijo et al., 2009). Usually this response is self-limiting and is resolved when infection has been eradicated or tissue damage has been repaired. Sustained inflammation, however, leading to tissue damage implies persistent inflammatory stimuli or failure in resolution mechanisms. A persistent stimulus can either result from environmental factors or chronic persistence of endogenous factors, being perceived by the immune system as ''stranger'' or ''danger'' signals and the subsequent inflammatory responses may overwhelm normal resolution mechanisms. While some inflammatory stimuli produce beneficial effects, i.e., phagocytosis, uncontrolled inflammation may as well result in generation of neurotoxic factors that amplify the underlying disease states (Glass et al., 2010).

Signs of neuroinflammation can be readily observed in pathologically affected areas in the brain and spinal cord in both human ALS patients as well as mouse models of the disease. Characteristic for ALS are gliosis and accumulation of large numbers of activated microglia and astrocytes marked by elevated production of inflammatory mediators (i.e., COX-2), proinflammatory cytokines (i.e., IL-1β, IL-6 and TNF-α) and potentially cytotoxic molecules such as ROS. For instance, major histocompatibility complex molecules and complement receptors were found highly expressed by activated microglia in the primary motor cortex and the anterior horn of the spinal cord in ALS patients (McGeer and McGeer, 2002). Receptors of the innate immune response could be potential sensors of molecules that induce or amplify inflammation in ALS (Letiembre et al., 2009), given that CD14, a protein facilitating TLR4 responses to lipopolyssaccharides (LPS), and TLR2 are upregulated in the spinal cords of mouse models (Nadeau and Rivest, 2000; Nguyen et al., 2001) and ALS patients (Letiembre et al., 2009; Liu et al., 2009). Indeed, a study conducting chronic infusion of presymptomatic ALS mice with LPS documented an enhanced innate immune response and exacerbated disease progression (Nguyen et al., 2004). Another piece of evidence for an active involvement of microglia in disease progression is the finding that mutant SOD1, but not wild-type protein, is presumably secreted into the extracellular space via chromogranin vesicles, resulting in activation of microglia and MN death in culture (Urushitani et al., 2006). Consistent with this is the toxicity of ALS patient derived CSF for exposed rat spinal cord cultures and the presence of mSOD1 within this CSF (Tikka et al., 2002) and intracerebral infusion of mutant SOD1 into wild-type mice can induce microglia activation and cytokine production (Kang and Rivest, 2007).

#### Glia—Astrocytes

Even though MNs are the primary cells affected in ALS, accumulating evidence points to a role of supporting astroglia during pathogenesis. MNs isolated from transgenic mSOD1 mice were found to be more sensitive to Fas or NO triggered cell death than in wild-type neurons suggesting a MN specific death pathway (Raoul et al., 2002). Binding of the ligand FasL to the Fas receptor leads to its intracellular portion recruiting the adaptor protein FADD which in turn activates a caspase cascade that culminates ultimately in MN death. While this pathway was found activated in presymptomatic ALS mice as well (Raoul et al., 2006), it is noteworthy that mSOD1 astroglia produce significant NO and FasL (Barbeito et al., 2004), thus implying that astroglia could in fact be the executioner of MN death in ALS. In line with this, the p75 neurotrophin receptor (p75NTR) has been associated with ALSassociated MN death, as nerve growth factor (NGF) secreted by mSOD1 astrocytes induced death of p75NTR expressing MNs via the formation of NO and peroxide (Pehar et al., 2004).

Cell-specific targeting of mSOD1 deletion in astrocytes and microglia utilizing GFAP-cre and CD11b-cre transgenes, respectively, was found to reduce severity of the disease phenotype and dramatically improve survival of ALS mice (Boillée et al., 2006b; Yamanaka et al., 2008b). While concerns related to the timing and efficiency of gene targeting in these mice have been raised, additional supporting evidence for an involvement of non-neuronal cells comes from studies with chimeric mice selectively expressing mSOD1 in MNs and non-neuronal cells. In such chimeras the presence of wild type non-motor neurons significantly delayed the onset of MND despite a high penetrance of mSOD1 MNs and OGs, increasing disease free life by 50% (Yamanaka et al., 2008a), and in co-cultures mSOD1 astrocytes exerted toxic effects on healthy wild-type MNs (Di Giorgio et al., 2008; Marchetto et al., 2008). These experiments suggest a noncell-autonomous scenario, for SOD1-meidated ALS at least, in which sick or mutant astrocytes directly contribute to MN degeneration.

That non-neuronal cells have a crucial impact on MN degeneration is also supported by the finding of mutant protein aggregates in neighboring astrocytes (Zhang et al., 2008). In a series of elegant experiments dramatic impact of ALS associated TDP-43 mutations were evident in human astrocytes. The authors generated a significantly purified population of astrocytes from patient-derived iPSCs with specific TDP-43 mutations (M337V) characterizing the cellular phenotype and found the astrocyte cultures to accumulate 30% more cytoplasmic TDP-43 than control astrocytes. Control-astrocytes that were transfected with mutant TDP-43 in turn displayed a similar amount of cytoplasmic TDP-43 accumulation (Serio et al., 2013). This emphasizes a direct responsibility of the specific TDP-43 mutation in the observed subcellular mislocalization.

### Glia—Oligodendrocytes

OGs are of particular importance for neuronal function and axonal transmission as they insulate axons with an essential myelin sheath. This myelin sheath allows for the fast and efficient propagation of action potentials and neuronal electrical transmission over long distances through saltatory conduction (Nave, 2010a,b; Simons and Lyons, 2013). Apart from this, OGs are a principal supplier of metabolic energy to axons and neurons.

Until recently, the involvement of OGs and their precursors in the pathogenesis of ALS was fairly unexplored. Abnormalities of OGs, such as pathological inclusions immunoreactive for smooth muscle alpha-actin, but neither angiotensin nor TDP-43, were observed in affected post mortem tissue of ALS patients. Pathological aggregates sequestering the ALS-linked protein TDP-43 occur in the cytoplasm of OGs from both sporadic and familiar ALS patients (Seilhean et al., 2009; Murray et al., 2011; Philips et al., 2013). Another ALS-linked protein, FUS, was found sequestered into cytoplasmic inclusion of patients with FUS-linked ALS. Interestingly, patients with late onset of FUSlinked ALS are characterized by highly abundant FUS inclusions in OGs, while early onset patients preferentially have neuronal cytoplasmic inclusions (Mackenzie et al., 2011; Nonneman et al., 2014).

Studies on ALS-linked mSOD1 transgenic mice revealed that gray matter OGs in the spinal cord degenerate before the first symptoms of MN degeneration become apparent (Kang et al., 2013; Philips et al., 2013). Indicatory for this OG degeneration is an abnormal morphology manifested in thickened, irregularly shaped cell body, enlarged cytoplasm and elongated reactive processes. Such dysmorphic OGs are in an apoptotic state characterized by cleaved caspase-3 immunoreactivity and altered chromatin condensation (Kang et al., 2013; Philips et al., 2013) and are significantly increased in SODG93a mice already before disease onset and are progressively increasing throughout (Philips et al., 2013). In addition, dense clustering of reactive microglia around degenerating OGs can be observed (Kang et al., 2013). Lineage tracing of OGs with a tamoxifeninducible OG-specific Cre line in mSOD transgenic mice (Plp-CreER; Rosa26-EYFP, SOD1G93A) confirmed the progressive decrease in the number of OGs as a function of disease progression in SOD1G93A mice (Kang et al., 2013; Philips et al., 2013). Morphological examination, however, revealed dysmorphic somata and branches and altered lipid and protein composition of the myelin sheath. Besides myelin defects, the newly generated OGs also fail to supply neurons with metabolic energy substrates, including lactate, pyruvate and ketone bodies (Philips et al., 2013). In addition, activated microglia and astrocytes exert OG toxicity as microglia for instance produce ROS and secret pro-inflammatory cytokines (Nonneman et al., 2014).

Although inflammation may not necessarily represent an initiating factor in neurodegenerative diseases, intriguing evidence obtained both from patient tissue and from animal models over the recent years stressed that sustained inflammatory responses involving microglia, astrocytes and OGs contribute to disease progression (**Figure 3**). This is the case not only in ALS and FTLD but in many other neurodegenerative diseases such as Alzheimer's, Parkinson's and multiple sclerosis (Glass et al., 2010). Whether inhibition of sustained inflammatory responses can be a safe and successful means to reverse or to at least slow down the course of disease progression should be a focus of further investigations. Disease-in-a-dish models utilizing ALS patient iPSC-derived co-cultures of MNs with astrocytes or OG could be used to study whether pharmacological inhibition of inflammation in the latter two cell-types would improve MN survival in vitro. As a second step such astrocytes and OG with pharmacological inhibition of inflammation could be transplanted into the CNS of ALS disease animal models to evaluate whether reducing inflammation in non-neuronal cells in vivo would ultimately slow MN degeneration or even improve MN health. Should this be the case, it will point to inflammation being the key contributor to trigger mitochondrial dysfunction, Ca2+-imbalance and ER stress responses.

#### MITOCHONDRIAL INVOLVEMENT IN ALS

Mitochondria perform diverse functions, such as producing ATP via oxidative phosphorylation, and regulating cellular levels of metabolites, amino acids and co-factors for various regulatory enzymes. As a main source of ROS and buffer of spatiotemporal Ca2<sup>+</sup> homeostasis, they integrate signals with other cellular compartments and contribute to cellular stress responses such as autophagy and apoptosis (Nunnari and Suomalainen, 2012). Further, mitochondrial fission and fusion specifies health in metabolic disorders and neurodegenerative diseases (Chen and Chan, 2009; Westermann, 2010).

The unique cytological properties of MNs, such as their extended cell size and the large volume of the axonal compartment, require a high metabolic demand and crucially depend on a robust cytoskeleton and axonal transport mechanisms. Due to that, MNs show increased generation of ROS and a tendency towards oxidative stress, making them particularly prone to other factors, i.e., ER stress, altered calcium levels and excitotoxicity (**Figure 1**). Developmental defects due to genetic mutations may confer mitochondria with an inert susceptibility towards developing MND (Ferraiuolo et al., 2011) as discussed.

#### Structural Alterations

In samples of ALS patients an altered ultrastructure is evident, with swollen and vacuolated mitochondria found in MNs, muscles, and intra-muscular nerves (Afifi et al., 1966; Atsumi, 1981; Siklos et al., 1996; Sasaki et al., 2007). Axonal swellings in MNs, Bunina bodies, comprised of accumulated neurofilaments, organelles and secondary lysosomes have been observed in the remaining LMNs and constitute a hallmark of ALS-specific histopathology (Hart et al., 1977; Okamoto et al., 1990). Imaging experiments on MNs derived from ALS transgenic mSOD1 mice revealed a reduced mobility (Magrane et al., 2009) and a net increase in retrograde transport of mitochondria away from the axon terminal, resulting in a depletion (De Vos et al., 2007, 2008).

A common feature of mitochondrial morphology changes in cellular and animal models of ALS is fragmentation of the mitochondria. Expression of mSOD1 in neuronal cells per se induces mitochondrial fragmentation. This pattern of alteration in mitochondrial morphology is for instance observed when interfering with the expression of aptic atrophy 1 (OPA1), a pro-fusion factor, dynamin related protein 1 (Drp1; Ferri et al., 2010) and other factors such as PINK1 (Matsuda et al., 2010; Deas et al., 2011). While mitochondrial alteration has been extensively studied in mSOD1 animal models, it remains to be determined, whether such changes are consistently found in other ALS associated mutations, both in animal models and human stem cell derived MNs.

### Reactive Oxygen Species (ROS) and Oxidative Stress

Mitochondria are particularly vulnerable to oxidative stress given that: (i) excessive ROS leads to lipid peroxidation interfering with mitochondrial membrane integrity; (ii) a plethora of mitochondrial proteins, such as essential proteins of the ETC complexes, contain highly oxidizable iron-sulfur clusters; and (iii) due to the lack of protective histone complexes mitochondrial DNA is far more susceptible to mutations in organelle encoded proteins (Cozzolino et al., 2012). Thus, conditions that interfere with mitochondrial homeostasis or lead to mitochondrial dysfunction are self-sustaining due to increased ROS production and accumulation. Experimental evidence obtained in mouse models have the focus of other reviews (Cozzolino et al., 2012).

Localized oxidative stress in ALS mitochondria, specifically ETC impairments, has been documented in patients and SOD1 mice (Kong and Xu, 1998; Miana-Mena et al., 2005). This was accompanied by energy deficits in the spinal cord (Jung et al., 2002; Mattiazzi et al., 2002; Kirkinezos et al., 2005), primary astrocytes and primary motor cortex (Cassina et al., 2008; Loizzo et al., 2010). A reduction in mitochondrial repair enzymes in patients and SOD1 mice (Kikuchi et al., 2002; Murakami et al., 2007), and markers of oxidative stress such as lipid peroxidation, protein glycoxidation and altered protein carbonyl levels in patient samples (Shibata et al., 2002) emphasize an exacerbating mitochondrial dysfunction in ALS (Mahoney et al., 2006). In mSOD1 mice phenotypes manifest in muscles with hypermetabolism, reduced adipose tissue accumulation and elevated energy expenditure already in the asymptotic phase (Dupuis and Loeffler, 2009; Dupuis et al., 2009). Such findings, however, have not been reported in patients and other ALS models, raising the question whether the early oxidative stress associated muscle phenotype in SOD1 animal models is an ALS specific early phenotype or, instead, rather a sub-form of ALS, specific to mutations in SOD1, and finally, whether such phenotypes may only be specific to the SOD1 animal models themselves.

### Calcium Buffering and Homeostasis

MNs express a large number of Ca2<sup>+</sup> permeable receptors lacking the GluR2 subunit (Greig et al., 2000; Heath et al., 2002; Van Damme et al., 2002; Kawahara et al., 2003), and in addition to that, have particularly low Ca2<sup>+</sup> buffering abilities (Lips and Keller, 1998; Palecek et al., 1999). This could make MNs particularly sensitive to glutamate toxicity exerted through AMPA receptors, however would mostly contribute to disease severity in later stages (Grosskreutz et al., 2010; Cozzolino and Carri, 2012). In turn selective loss of the astroglial glutamate transporter GLT1 is a more likely scenario of neurotoxicity due to impaired glutamate clearance from the synaptic cleft (Rothstein et al., 1995; Bendotti et al., 2001; Howland et al., 2002).

Toxic Ca2<sup>+</sup> dysmetabolism in ALS is probably mostly evoked by a deficiency in intracellular Ca2<sup>+</sup> mishandling, as a consequence of low levels of Ca2+-binding proteins or impaired mitochondrial Ca2<sup>+</sup> buffering. Supporting this, calbindin D-28K immunoreactivity was found reduced in the frontal cortex of patients with FTLD (Ferrer et al., 1993), and in ALS patients, with significant differences in the densities of calbindin positive neurons within cortical layers V and VI (Maekawa et al., 2004). A loss of calbindin and parvalbumin immunoreactivity from subsets of MNs in presymptomatic mSOD1 mice suggests calbindin loss as an early event in ALS pathogenesis (Sasaki et al., 2006).

Evidence for defective Ca2<sup>+</sup> storage as a result of mitochondrial dysfunction initially came from cell models for SOD1-linked ALS, which displayed elevated cytoplasmic Ca2<sup>+</sup> levels in conjunction with compromised mitochondrial gradient potential (Carri et al., 1997; Jaiswal et al., 2009). In animal and cell models of mSOD1, MNs expressed markers of oxidative stress and mitochondrial dysfunction, elevated Ca2<sup>+</sup> levels and vulnerability to glutamate toxicity (Kruman et al., 1999). Further in vivo studies manifested prolonged Ca2+-induced membrane polarization in an early decrease in mitochondrial Ca2<sup>+</sup> loading capacity, impaired Ca2<sup>+</sup> uptake in neurons and muscle tissue (Damiano et al., 2006; Jaiswal and Keller, 2009; Nguyen et al., 2009; Zhou et al., 2010b). Adding to this it has been found that Ca2<sup>+</sup> is able to bind to SOD1 immature states and promote its aggregation (Leal et al., 2013). This emphasizes that, apart from mutations in ALS-related proteins disrupting calcium homeostasis, Ca2<sup>+</sup> itself affects ALS-critical proteins and cellular processes as well.

In summary, Ca2<sup>+</sup> dysregulation and mitochondrial pathology are tightly interconnected. Apart from SOD1, however, many other ALS-critical proteins have been shown to potentiate Ca2<sup>+</sup> dysregulation thereby increasing cellular Ca2<sup>+</sup> vulnerability as well (Leal and Gomes, 2015). Altered Ca2+-influx can increase ROS production and accumulation and cytochrome c release, making this process self-sustaining and steering towards activation of cell death (Dykens, 1994). Although it is tempting to speculate that Ca2<sup>+</sup> could be the one factor shifting MN integrity out of balance, future studies will need to address early Ca2<sup>+</sup> phenotypes further. Intriguingly, Ca2<sup>+</sup> dysregulation and mitochondrial pathology are also intimately connected to protein aggregation as we will discuss in the next chapter.

### Non-SOD1 Mutation Derived Mitochondrial Alteration and Damage

In recent years, a number of studies have described mitochondrial alteration and dysfunction in non-mutantSOD1-linked cases of ALS. TDP-43, a protein with dual functions as DNA-binding protein with nuclear export sequence and as RNA-binding protein, is found to be predominantly localized to the cytosol instead of the nucleus in ALS spinal MNs, found in punctate immunoreactive granules and as dense skeins (Kabashi et al., 2008). Transgenic mice overexpressing a human TDP-43 (hTDP-43) under the control of the mouse prion protein reporter in the brain and spinal cord, produce an ALS-like phenotype exhibited by reactive gliogenesis, axonal and myelin degeneration, gait abnormalities and early lethality. The hTDP-43 overexpression induces abnormal perinuclear aggregates of mitochondria accompanied by enhanced levels of fission and fusion components, Fis1, phosphorylated Dlp1 and mitofusin 1, respectively (Xu et al., 2010). Accumulation of mitochondria in TDP-43-negative cytoplasmic inclusions in MNs was also found in a Thy1.2; hTDP-43 mouse model that notably lacked mitochondria in motor axon terminals. This, however, could be due to an alteration in elements of the transport machinery, kinesin-associated proteins Kif3a and KAP3, found in inclusions (Shan et al., 2010). Alternatively, it has been observed that TDP-43 stabilizes the RNA of human low molecular weight neurofilament protein via interaction with the 3 <sup>0</sup>UTR, altering its location in spinal MNs and thereby possible favoring the formation of NF aggregates in ALS (Strong et al., 2007) and mitochondria mis-localization.

The RNA-binding protein FUS/TLS is normally localized in the nucleus where it participates in process involved in RNA maturation, among those snRNP biogenesis, alternative splicing, and mRNP localization. Most of the research on the role of mutant FUS in ALS has focused on its mislocalization and aggregation, and little is known about its involvement in mitochondrial dysfunction. Mutant forms of FUS/TLS were found to accumulate in the cytoplasm, similarly to TDP-43, and form protein inclusions in spinal horn neurons derived from ALS patients (Deng et al., 2010). In a few juvenile ALS cases post-mortem tissue analysis of MNs revealed basophilic inclusions containing abnormal aggregates of FUS proteins and disorganized intracellular organelles, among those mitochondria and endoplasmic reticuli (Huang et al., 2010).

While the SOD1 model may only represent a small fraction of the heterogeneous disease population and posits a note of caution when interpreting mSOD1 mouse phenotypes at large across idiopathic forms of ALS, recent findings supporting the view of OS and mitochondrial damage to play a role in non-SOD1 ALS also come from cell and invertebrate models (Carri et al., 2015). The phenotypes found in mutant TDP-43 and FUS/TLS support the concept of mitochondrial dysfunction and damage as a major event in ALS progression. Further investigation is needed to elucidate if mitochondrial alteration and dysfunction phenotypes in non-SOD mutants occur at early disease stages as a key gateway in ALS pathogenesis.

### Mitochondrial Susceptibility as Cause vs. Result of ALS

While ALS-associated mutations can to some extent explain the circumstances leading to ROS and Ca2<sup>+</sup> dysregulation in MN ultimately resulting in apoptosis, they do not account for the overwhelming majority of sALS cases in which no genetic inheritance has been found. A limited ability of MNs to balance intracellular Ca2<sup>+</sup> levels, however, may very well be attributable to sALS cases as well. Given the large cytoarchitecture of MN axons this raises the question whether mitochondrial phenotypes in ALS could be a matter of inert susceptibility being an early event in the onset of MN degeneration. Other than heritable genetic mutations of fALS cases, factors interfering with mitochondrial function and Ca2<sup>+</sup> homeostasis, such as environmental perturbagens and/or yet to be identified modifier genes, could provoke prolonged mitochondrial stress and affect cell function in sALS cases. Compelling evidence exists pointing to approaches that artificial manipulation and rescue of mitochondrial dynamics may lead to new therapeutic approaches that would, to the least, extend MN life and improve disease conditions in patients (Chen and Chan, 2009).

### Screening and Reversing Mitochondrial Phenotypes

A major hurdle in clinical progression of ALS still is that the time for a definite diagnosis of patients, which may take 15 months or longer. Given an average life expectancy of 3–5 years, it is crucial to develop faster means of clinically detecting ALS. As a complex disease with a wide range of phenotypical presentation, in the clinic as well as in animal models, mitochondrial stress and dysfunction represent a common feature of MN degeneration (Talbot, 2014) and may proof valuable as molecular markers in future diagnostics. Rescuing mitochondrial dysfunction in ALS ultimately represents a valuable target for many other neurodegenerative diseases as well. Human stem cell derived disease-in-a-dish models could provide additional insight. Patient-derived fALS iPSCs have been shown to reproduce some of the neuronal stress and degeneration phenotypes (Donnelly et al., 2013; Sareen et al., 2013; Kiskinis et al., 2014; Wainger et al., 2014). In vitro screening in human MNs and reversing associated mitochondrial dysfunction signatures may thus represent an invaluable means of unraveling molecular signatures of human neurons.

### ENDOPLASMIC RETICULUM (ER) STRESS

The ER is the primary compartment in which secreted and membrane proteins are processed to ensure proper folding via different classes of folding mediators. A cohort of molecular chaperones assists the folding process of unfolded or misfolded proteins preventing aggregation. Such chaperones are found in multiple sub-cellular compartments, and each of those compartments possesses its own subset of chaperones (Zapun et al., 1999). A number of stress conditions can interfere with the normal function of the ER leading to abnormal oxidative folding of proteins at the ER lumen inducing ER stress (Boillée et al., 2006a; Pasinelli and Brown, 2006; Matus et al., 2013). Mutations in ALS disease causative genes, such as TARDBP/TDP-43, FUS/TLS, and vesicle-associated membrane protein-associated protein B (VAP-B), for instance, trigger aggregation of the mutant proteins leading to a gain of neurotoxic activity and provoking ER stress (Andersen and Al-Chalabi, 2011; Turner et al., 2013). These conditions engage the UPR, a signal transduction pathway responding with increased protein folding capacity and quality control mechanisms of the ER to restore homeostasis (Hetz et al., 2011; Walter and Ron, 2011). Chronic ER stress, however, yields irreversible damage of the neuronal cells and ultimately MN death (Hetz, 2012).

Three major stress sensors activate the UPR. (1) The inositol-requiring transmembrane kinase/endonuclease (IREI1), the PKR-like ER kinase (PERK), and activating transcription factor 6 (ATF6). IREI1 initiates splicing of an mRNA coding for the transcription factor X-Box-Binding protein 1 (XBP1), converting it into a potent activator of multiple UPRresponsive genes, the XBP1s (Calfon et al., 2002). XBP1s in turn control the expression of genes responsible for protein folding, protein quality control, and secretion or ER-associated degradation (ERAD; Acosta-Alvear et al., 2007), while IREI1 further engages in degradation of specific subsets of RNAs and pathways involving Janus Kinases (JNK), mediating apoptosis and autophagy levels (Ron and Walter, 2007). (2) PERK is a stress sensor and once activated, reduces protein translation into the ER by phosphorylation of the eukaryotic initiation factor 2 alpha (eI2Fα), leading to a decrease in misfolded protein overload (Harding et al., 1999). The eI2Fα enables expression of the gene ATF4, which in turn upregulates a cascade of UPR-targeted genes responsible for amino acid and redox metabolism, protein folding, autophagy, and apoptosis. One of those target genes, CHOP, is a crucial mediator of apoptotic processes in response to ER stress, supposedly via driving the expression of pro-apoptotic Bcl-2 family members, such as BIM and PUMA (Tabas and Ron, 2011; Walter and Ron, 2011), in addition to GADD45. Sustained PERK activation, however, contributes to apoptosis by increasing oxidative stress and resuming protein synthesis after prolonged ER stress (Marciniak et al., 2004; Verfaillie et al., 2012; Han et al., 2013). (3) The third major stress sensor, ATF6, is activated in the ER and translocates to the Golgi apparatus. Following processing, its released cytoplasmic domain acts as a transcription factor (Walter and Ron, 2011). The activated ATF6 in turn controls certain sets of target genes related to protein folding and quality control (Haze et al., 1999; Chen et al., 2002).

### Misfolded Proteins and Protein Aggregates in Motoneuron Disease

Familial ALS mutations have been associated with an accumulation of misfolded proteins (Andersen and Al-Chalabi, 2011; Turner et al., 2013), increased ER stress and dysfunction as well as pathogenic structural alterations of the ER in ALS patient tissue (Oyanagi et al., 2008; Sasaki, 2010; Vijayalakshmi et al., 2011). In human postmortem tissue of sALS cases ER stress was evident by distended and fragmented ER cisternae in affected cells of the anterior horn of the spinal cord. Ribosome-free membranous structures extending from the ER and electron dense material resembling Bunina bodies, Hirano bodies and honeycomb-like structures were observed in samples as well (Oyanagi et al., 2008; Sasaki, 2010; Vijayalakshmi et al., 2011). Interestingly, a recent study immunolabeling patient tissue with GRP78 for BiP expression found even normal appearing MNs labeled and display dilated ER lumen with amorphous and granular material in electron microscopy analysis (Oyanagi et al., 2008). However, many technical limitations exist in interpreting causative processes and the precise kinetics in many cellular neuropathological observations made in post-mortem ALS tissue.

TDP-43, a member of the heterogeneous nuclear ribonuclearprotein (hnRNP) subclass of 2xRBD-Gly proteins, shuttles between the nuclear and cytoplasmic compartments, binding both DNA and RNA, and is involved in exon splicing of several genes (Chen-Plotkin et al., 2010). In both ALS and FTLD pathogenesis TDP-43 was found as the most frequent component of ubiquitinated protein inclusions. The clinical histopathology is characterized by cytoplasmic inclusions of skein-like or dense granular appearance and by clearance of TDP-43 from the nucleus (Geser et al., 2008). TDP-43 can bind to RNA in a singlestranded and sequence-specific manner, required for many RNA processes, and its most C-terminal portion bears unique characteristics making it prone to misfolding and aggregation (Buratti and Baralle, 2008; Johnson et al., 2009; Achsel et al., 2013; Al-Chalabi and Hardiman, 2013).

Expansions of a hexanucleotide G4C<sup>2</sup> repeat in the first intron/promoter region of the C9ORF72 gene were identified as the most frequent genetic cause of both f/sALS cases and FTLD (Dejesus-Hernandez et al., 2011; Renton et al., 2011; Gijselinck et al., 2012). Reports suggested that the repeat is transcribed and leads to accumulation of repeat-containing RNA foci in patient's tissues (Dejesus-Hernandez et al., 2011). Studies using stem cell models of ALS, one by the Sattler and Rothstein groups and another by our group (Donnelly et al., 2013; Sareen et al., 2013), generated iPS cells from patients with ALS caused by C9ORF72-associated repeat expansion. While patient-iPSC derived MNs were not adversely sensitive to cell death processes, they exhibited certain electrophysiological phenotypes. The mutant C9ORF72 transcript containing G4C<sup>2</sup> repeat expansions conferred a gain of function neurotoxicity, which was abolished upon correcting gene expression profiles by anti-sense oligonucleotides (ASOs) introduction (Donnelly et al., 2013; Sareen et al., 2013). C9ORF72 mutant animal models, with one exception (Koppers et al., 2015), reported to recapitulate key molecular signatures unique to the intronic G4C<sup>2</sup> repeat expansion, including RNA-foci and inclusions of dipeptide repeat (DPR) expansion inclusions, TDP-43 pathology, and behavioral abnormalities similar to symptoms observed in C9ORF72/ALS patients (Chew et al., 2015). Furthermore, some of these studies were able to associate C9ORF72 repeat expansions to interfere with nucleocytoplasmic shuttling and identified elements of the nuclear import/export machinery to either enhance or suppress DPR mediated toxicity in Yeast and fly models (Freibaum et al., 2015; Jovicic et al., 2015; Zhang et al., 2015).

Many fALS mutation occur in genes encoding key components of protein quality control. This includes components essential for the retrograde transport of autophagosomes from axons to the cell body such as dynein and dynactin (Moughamian and Holzbaur, 2012; Ikenaka et al., 2013), the autophagic adaptor p62 (Fecto et al., 2011), as well as the UBA-containing proteins Ubqln2 and Optineurin (Fecto and Siddique, 2012). Elevated levels of ER chaperones, foldases, cell death signals and upregulation and activation of the IREI1, PERK and ATF6 branches of UPR signaling have been reported by multiple studies in sALS patient spinal cord tissue (Ilieva et al., 2007; Hetz et al., 2009; Ito et al., 2009). Misfolded proteins are constantly generated in many compartments of the cells, and are usually removed by quality control systems composed of the ubiquitin (Ub)-proteasome system (UPS), chaperone mediated autophagy (CMA) and macroautophagy. Proteins that escape the surveillance of the UPS and CMA or tend to form aggregates are subjected to macroautophagy for degradation into amino acids (Hariharan et al., 2011; Koga and Cuervo, 2011; Ciechanover and Kwon, 2015). SOD1 and TDP-43 may additionally perturb their folding, leading to the formation of β-sheet enriched aggregates (Andersen and Al-Chalabi, 2011) that are resistant to all known proteolytic pathways and can further grow into inclusion bodies and extracellular plaques (**Figure 2**; Koga and Cuervo, 2011).

Aggregating proteins form intracellular inclusions containing Ub and Ub ligases, as found in animal models (Bruijn et al., 1997; Bendotti et al., 2004) and patient tissue (Leigh et al., 1991; Bendotti et al., 2004; Sasaki, 2010) becoming visible in the brain stem and spinal cord at onset of the manifestation of ALS symptoms (Watanabe et al., 2001). Although these large inclusions are clinical hallmarks of ALS symptoms, they are unlikely to be toxic to neurons. In fact they may be a neuroprotective phenomenon, as monomeric and oligomeric misfolded ALS proteins seem to be toxic to MNs (Guo et al., 2010). Monomeric and oligomeric misfolded proteins in turns may exert the actual toxicity in ALS according to recent findings suggesting a ''prion-like'' focal propagation of the disease (Polymenidou and Cleveland, 2011; Sugaya and Nakano, 2014). In support of that mutant and wild-type SOD1 have been found to spontaneously fibrillize (Grad et al., 2011; Munch et al., 2011) and both TDP-43 and FUS were proposed to contain a prion-like glutamine/asparagine (Q/N)-rich domain found in inclusions. In vitro these inclusions exhibited ordered, self-perpetuating aggregation and were transmissible from one cell to its progeny (Ito et al., 2011; Polymenidou and Cleveland, 2011).

Despite an ever growing number of ALS-causative genes having been identified, yet, difficult to identify one key factor driving disease progression in MNs. Protein misfolding and the effects of mono- and oligomeric forms on sequestering functional RNAs and proteins from cellular processes constitute a plausible cause of MND and are in alignment with many of the ALS typical hallmarks found in ALS post mortem tissue. Given the sensitivity of RNA and protein metabolism, rapid development of disease phenotypes (i.e., TDP43, C9orf72) are often observed in animal models such as mice and flies. The corresponding genetic mutations in ALS patients however, display later in life with median onset age greater than 50. Many more elements of ER stress responses, i.e., the UPR, have been found to impact phenotypes of animal models and reviewed elsewhere (Matus et al., 2013).

### Screening Assays for Testing and Reversing UPR Phenotypes

A number of studies have shown that ER chaperones can be secreted into the extracellular space upon cellular stress (Dorner et al., 1990). In line with this is the finding of PDIA1 levels being upregulated in the cerebrospinal fluid (CSF) of ALS patients (Atkin et al., 2008) and a study exposing spinal MNs to CSF of ALS patients documented an induction of ER stress in the MNs (Vijayalakshmi et al., 2011). Furthermore proteomic biomarkers in blood samples of ALS patients revealed the up-regulation of chaperones, such as the ER-stress responsive chaperones PDIA1 and ERp57. This was also confirmed for mononuclear cells from blood of mSOD1 mouse models. Finally, in a longitudinal study TDP-43, cyclophillin A, and ERp57 were strongly associated with disease course in ALS patients and control subjects (Nardo et al., 2011).

UPR signaling responses convert information about the nature and intensity of the occurring stress into a network of partially overlapping target genes that orchestrate an adaption to the stress, or trigger cell death programs as a last resort. Given the need for biomarkers to quantitatively assess disease stage, prognosis and efficacy of clinical trials, these studies suggest measuring stress factors released into the CSF or the blood as a valuable tool for diagnosis and monitoring ALS disease progression. Regardless of the origin of cellular stress in MNs, it is increasingly clear that modulating stress due to protein (mis-)folding and the proteostatic capacity of MNs represent a promising therapeutic target to delay the symptomatic phase of ALS. With regard to this, utilizing gene therapy or small molecule approaches to reinforce the capacity of coping with ER stress mechanisms may be an applicable means for disease intervention (Matus et al., 2013). While animal models of ALS provide valuable insight into many aspects of genetically linked disease progression in vivo, there are, however, limitations to these mouse models with regard to specificity of the human disease phenotypes (**Table 3**). To this end, ALS patient-derived iPSCs and fluorescent reporter cell lines can serve as an important complementary tool to monitor and

#### TABLE 3 | Hallmarks of ALS in animal models.


Cellular and molecular features presenting in clinical ALS are listed in the far left column. Some of the most frequently found mutations found associated with fALS and sALS are represented at the top and the phenotypes found in animal models summarized for each category. Color coding: Black, mouse/rat; Red, fruit-fly; Green, worm; Blue, zebrafish; +, phenotype; −, no phenotype; ±, variable phenotype among different animal lines; Res, restricted phenotype; ND, not determined (yet).

discover molecular markers of ER stress and UPR, specifically in different subclasses of MNs from patients bearing the patient genotype.

### HUMAN STEM CELL DISEASE MODELS FOR ALS

The majority of disease processes involve multiple cell and tissue types and in principle, are best modeled utilizing animals. Not surprisingly, animal models have been instrumental to our understanding of human disease pathology, including neurodegenerative diseases such as PD, AD and ALS. Nonetheless, many neurodegenerative diseases, such as those affecting MNs, in fact do not naturally occur in commonly used laboratory animals. Introduction of certain pathogenic aspects of human neurodegenerative diseases with mutant genes being wildly overexpressed into animals often creates phenotypes that only partially resemble the original human disease (**Table 2**).

Introduction of mSOD1 into mice to induce disease phenotypes that resemble those manifested in patients requires expression of multiple copies of mSOD1. In human patients, a single copy is sufficient (Bruijn et al., 1997). Moreover, the most prevalent form of mSOD1 in ALS patients, the alanine to valine substitution (A4V), does not appear to generate phenotypes in mice (Furukawa et al., 2006). Thus, differences do exist between humans and mice, with the pathogenic potential of mSOD1 just being one example, that may complicate the study of some mutations in animal models alone. Astrocytes have also been found to be very different between humans and rodents and could add a potential caveat in disease modeling (Oberheim et al., 2009). However, as illustrated in the previous chapter, glial cells such as astrocytes have a critical role in the pathogenesis of ALS and other neurodegenerative diseases. As a consequence, the utilization of model systems with a human background as means of complementing studies in animal models can greatly benefit our understanding of MNDs.

### hESCs and iPSCs as Disease-In-A-Dish Models

In order to establish human disease models, there are two main routes of approach. The first approach is to genetically alter human stem cells with the target gene(s) of interest, and the second is to derive stem cells from patients with target diseases. The genetic modification of mouse embryonic stem cells (ESCs) by targeting specific gene loci via homologous recombination has been routinely applied in many laboratories as a first step to creating transgenic animals. Likewise, this same principle is applicable when using human embryonic stem cells (hESCs; Zwaka and Thomson, 2003). Repeated cellular cloning, however, as is required by traditional approaches using homologous recombination, can often render the established hESC cell line unstable. Alternatively, random gene insertion, by lenti-viruses for instance, can reduce cloning cycles significantly, yet transgenes that are integrated during the stem cell stage may be downregulated during differentiation (Xia et al., 2007). When Du et al. (2009) screened for gene loci resistant to gene silencing using a lentiviral vector with built-in Cre-loxP cassette, they were able to establish cell lines that were resistant to gene silencing even upon differentiation of these hESCs into neurons and astrocytes. These master cell lines in turn allow for the introduction of any gene of interest, including genes provoking human diseases, via the built-in Cre-loxP cassette through Cre recombination with high efficiency. In addition, recent technological advances allowed for the use of zinc fingers to target specific gene loci with high efficiency (Hockemeyer et al., 2009), alleviating the necessity to screen large numbers of cell clones.

The second approach, obtaining human stem cells with disease traits capturing the patient genotype, has been made available ever since the hallmark discovery that human induced pluripotent stem cells (hiPSCs) can be generated from somatic cells such as fibroblasts (Takahashi et al., 2007; Yu et al., 2007; Hanna et al., 2010). Somatic cells can be transitioned into iPSCs exhibiting phenotypes very similar to hESCs by expressing pluripotency factors such as Oct3/4 and Sox2 combined with either Klf4 and c-Myc, or Lin28 and Nanog (Takahashi et al., 2007; Yu et al., 2007). With similar approaches hiPSCs were generated from somatic cells harvested from patients with MNDs including ALS (Dimos et al., 2008; **Table 4**) and SMA (Ebert et al., 2009; Sareen et al., 2012). These hiPSCs can be differentiated can be readily differentiated into neurons and astrocytes.

Many of the disease phenotypes such as MN death, or glial cell impairment occur in iPSC derived differentiated cells (**Table 5**), emphasizing that patient hiPSCs may provide a versatile model to dissect the cellular and molecular mechanisms underlying MN degeneration (**Table 4**). A recent study by Zhang and colleagues utilizing patient derived mSOD1 iPSCs (D90A and A4V) as a model for MN degeneration in ALS found mutant SOD1 to exhibit neurofilament aggregation specifically in MNs (Chen et al., 2014). This was accompanied by decreased stability of NF-L mRNA and binding of its 3 <sup>0</sup>UTR by mSOD1, yielding altered protein proportion in NF subunits. Expression of a single copy of mSOD1 could mimic this MN-selective phenotype in hESCs as well, while a genetic correction of mSOD1 in patient-derived iPSC MNs could revert the phenotype. Interestingly, conditional NF-L expression in the mSOD1 iPSC derived MNs corrected the NF-subunit proportion, mitigating NF aggregation and neurite degeneration in MNs. This suggests an mSOD1 mediated misregulation and aggregation of NF heavily impacts axonal degeneration in ALS MNs (Chen et al., 2014). Moreover, two other studies using patient derived mSOD1 iPSC models reported transcriptional and functional changes in mitochondrial and ER stress pathways as causes of perturbed electrochemical activity in ALS neurons (Kiskinis et al., 2014; Wainger et al., 2014).

Noteworthy, the studies above were also able to analyze patient iPSC-derived MNs harboring repeat expansions in C9ORF72 and retrieve a subset of the transcriptional changes found in the mSOD1 iPSCs, indicating these are being broadly conserved in ALS (Kiskinis et al., 2014; Wainger et al., 2014). Other studies revealed disease specific phenotypes,

#### TABLE 4 | Human stem cell models of ALS.


An overview of human stem cell models utilized to study the cellular pathology of MND in vitro. The left column Mendelian and atypical forms of ALS associated mutations. In the middle columns representative model terminology and disease associated genes are listed and the right column summarizes in how far ALS associated phenotypes are resembled in the designated models.

#### TABLE 5 | Phenotypes of human stem cell models of ALS.


Cellular and molecular features presenting in clinical ALS are listed in the far left column. Some of the most frequently found mutations found associated with fALS and sALS are represented at the top and the phenotypes found in iPSCs summarized for each category. +, phenotype; −, no phenotype; ±, variable phenotype; ND, Not determined (yet).

i.e., dysregulated gene expression, G4C<sup>2</sup> RNA foci, and susceptibility to excitotoxicity via a gain of function RNA toxicity (Donnelly et al., 2013; Sareen et al., 2013). These phenotypes could be mitigated with antisense oligonucleotides (ASO) targeting the C9ORF72 transcript or repeat expansions despite the presence of repeat-associated non-ATG (RAN) translation products (Donnelly et al., 2013). RAN translation can express homopolymeric expansion proteins in all three reading frames, without an AUG start codon. This noncanonical type of protein translation stands in stark contrast to the classical rules of translational initiation, is lengthand hairpin-dependent, and occurs without frameshift or RNA-editing as a result of the C9ORF72 G4C<sup>2</sup> repeat expansion (Zu et al., 2011; Cleary and Ranum, 2013). Interestingly, C9ORF72 mutant iPSC-derived MNs displayed activation of ER stress pathways and MN death when subjected to the ER specific stressor tunicamycin (Donnelly et al., 2013).

TDP-43 patient iPSC-derived mutant MNs formed de novo cytosolic aggregates similar to those observed in postmortem tissue of ALS patients (Egawa et al., 2012; Burkhardt et al., 2013). Not only was this observed consistently in independent studies, but iPSC derived mutant TDP-43 MNs were also shown to recapitulate the same degenerative phenotype as observed in postmortem tissue of the corresponding patient (Burkhardt et al., 2013). Moreover, the TDP-43 iPSC derived MNs exhibited shorter neurites and increased amount of mutant TDP-43 protein in a detergent insoluble from Egawa et al. (2012). Recently, mutations of TDP-43 in patient iPSCderived MNs associated with reduced levels of micro-RNA

FIGURE 3 | Possible causes of ALS. The most prevalent underlying reason for MN defects are genetic perturbagens, inherited or de novo mutations. Yet, the majority of ALS cases have not been linked to any mutation, suggesting that other effectors such as the environment may play into this as well. Thus MN phenotypes observed in ALS may arise from genetic and/or environmental perturbagens, as depicted in the top panel. Developmental anomalies may affect the structural integrity of neuronal cytoarchitecture as conferred by structural proteins, transport proteins, transmembrane proteins or by a disruption of RNA processing. Such defects can interfere or even prevent the formation of synapses between neurons or at the NMJ either directly within MN or indirectly by affecting neighboring glial cells. Triggering of autoimmune responses and chronic low-level inflammation may lead to MN degeneration as well and many of those developmental defects manifest in electrophysiological deficiencies such as a progressive decrease in voltage-activated Na<sup>+</sup> and K<sup>+</sup> currents correlated with a loss of functional outputs. Mitochondrial susceptibility due to ROS-induced OS in turn yields an inert vulnerability of MNs to excitotoxicity. An increased amount of mitochondrial stress in turn leads to mitochondrial fragmentation and ultimately cell death. Due to their large size and long neurite outgrowths, MNs are particularly sensitive to ion fluctuations, whether due to selective permeability for Ca2<sup>+</sup> influx or lack of messenger clearance from the synaptic cleft. The dysregulation of intracellular Ca2<sup>+</sup> levels has severe implications for MN function as well. Both, the ER and mitochondria function as buffers of cellular Ca2<sup>+</sup> homeostasis. When intracellular Ca2<sup>+</sup> levels increase, either by influx from the extracellular space or the ER and mitochondria, this triggers OS responses, their fragmentation and eventually progression of cell death signals that ultimately lead to loss of electrophysiological outputs. The generation of misfolded proteins and formation of aggregates, likewise affects the functional integrity of both organelles. Initially, misfolded proteins trigger the UPR in the ER to compensate for decreased protein translation and processing efficiency. Persistence of misfolded proteins that escape corrective degradation mechanisms cause accumulation in the cell and ultimately lead to the formation of insoluble aggregates interfering with cellular functions such as molecular transport and trafficking.

9 (miR-9) and its precursor pre-miR-9-2, suggesting miR-9 downregulation to be a potential common event in ALS and FTLD (Zhang et al., 2013). A recent study in turn reported an initial hyperexcitability followed by a loss of functional output and synaptic activity, yielding a progressive decrease in voltage activated Na<sup>+</sup> and K<sup>+</sup> currents (Devlin et al., 2015). Interestingly, this was observed in, both, mutant TDP-43 and C9ORF72 patient-derived MNs, suggesting an early defect in ion channel status as a common event contributing to potential initiation of downstream degenerative pathways in ALS.

Taken together, familial ALS patient-iPSC models recapitulate some of the disease phenotypes also observed in patient tissue and some of transgenic animal models. Human iPSC disease-in-a-dish models complement other established models to decipher the molecular pathways involved in ALS. Also, human iPSC-based models allow for greater versatility in screening of human ALS disease signatures by: (a) providing accessibility of difficult-to-obtain CNS cell types such as MNs; (b) allowing investigations on developmental effects of any ALS-associated mutation; (c) giving options to analyze single mutant cells; (d) supplying

glial cells for interaction studies in co-culture with neurons; (e) allowing prospects of ALS biomarker discovery; and (f) providing platforms for therapeutic screening strategies.

#### Identification of Molecular Markers

A crucial step on the way to treatment for ALS is the identification of an early biomarker signature. Whether it comprises markers of neural and glial pathogenesis, mitochondrial and ER stress signals (Cozzolino et al., 2012; Hetz, 2012), or early signatures before actual disease onset, such means of diagnosis could greatly facilitate determination of disease progression and the most applicable measures of treatment. To yield good biomarkers ALS models need to resemble the main characteristics of the disease phenotype in patients and allow for easy accessibility of the involved cell types. Animal models have been most frequently employed for this, with rodent animals such as mouse and rat being a major source, but even invertebrate models such as flies and worms (Carri et al., 2015).

Patient-derived iPS cells from a large cohort of wellphenotyped patients are another promising tool as they allow direct study of ALS patient-derived human cell types. For instance iPSC-derived MNs and glial cells could be screened for secreted stress molecules found in the CSF and blood of ALS patients (Atkin et al., 2008; Nardo et al., 2011; Vijayalakshmi et al., 2011). Likewise, in addition to animal models, pre-clinical models could incorporate ALS patient iPSC-derived cells in the discovery and testing of promising therapeutics. Such approaches have been put to test using patient iPSCs. In a chemical screen referring to the TDP-43 aggregate endpoint, a recent study identified FDA-approved small molecule modulators such as Digoxigenin, emphasizing the feasibility of patient-derived iPSCs-based disease modelling for drug screening (Burkhardt et al., 2013). Microarray analysis of iPSC-MNs from mutant TDP-43 patients, decreases in the expression of genes encoding cytoskeletal proteins and small increases in genes involved in RNA metabolism was observed. When treating the MNs with anacardic acid, a histone acetyltransferase inhibitor, they were able to rescue the abnormal MN phenotype (Egawa et al., 2012). Over the next 3–4 years there will be increasing accessibility of large cohorts of patient-derived iPSCs from thousands of ALS patients as a result of collaborative initiatives across the USA, in part due to funds raised by the ALS Ice Bucket Challenge, such as the TRACK/ENROLL-ALS, NeuroCollaborative, and Answer ALS among many others. These ALS iPSC panels bear great potential for screening and identification of robust biomarkers signifying various stages of ALS. In addition, iPSC-derived neurons and glia may be used to identify large sets of markers of diverse stressors thereby enabling the creation of an ALS disease signature.

There are significant limitations and challenges associated with patient iPSC-derived models. Some of these include: (a) obtaining homogenous population of neural subtypes devoid of undifferentiated progenitors; (b) prolonged time required in culture for maturation; and (c) fetal/young ''age'' of the different neural subtypes. Further development of cell sorting tools, fluorescent reporters, and better markers (surface/intracellular) for the human cell types is imperative. Recent advancements in genome editing and creation of CRISPR/Cas edited reporter iPSC lines will help solve some of these technical issues. The big question that remains and needs addressing is whether it is required to generate completely homogenous cellular models (neurons or glia) to mimic human ALS-in-a-dish. The answer is going to be context and question-dependent. Nevertheless, a motor neuron during development will not engender synaptic maturity without the presence of astrocytes in culture.

## iPSC Derived Cell Transplantation Approaches

Apart from using patient-derived iPSCs for identification of diagnostic markers, they may also bear the potential for cellbased therapies. While such approaches have been used in neurological diseases with focal pathology, i.e., PD (Björklund and Lindvall, 2000), replacement of affected or lost cells in widespread areas of the CNS, as encountered in ALS will prove to be very challenging. However, replacement of affected cells in critical parts of the brain and spinal cord, such as respiratory centers, on the other hand, may be life-saving.

The progression of ALS is heavily affected by interactions between MNs and neighboring glial cells, as outlined in the previous chapter. Thus, replacing diseased or toxic astrocytes at an early stage could potentially rescue MNs from degeneration. Replacement of astrocytes in support of MN health has been successfully conducted in a rat model of ALS. The primary astrocytes not only survived the transplant, but also contributed to increased animal life-span with grafted astrocytes migrating a certain distance along the spinal cord and offering support for MNs in the transplant site and neighboring areas (Lepore et al., 2008). Although the human spinal cord is of course substantially larger than that of rodent animals, it is nonetheless surgically feasible transplant cells at multiple sites. As dramatically enriched and even pure populations of human glial progenitors differentiated from PSCs in a chemically defined culture system can be produced in large quantities, production of clinical grade human astrocytes in a clean facility is also technically feasible. Specifically, generating astrocytes from the patient's own somatic cells through reprogramming to iPSCs will circumvent the issues of immune rejection in transplantation (Liu and Zhang, 2010). We have also shown feasibility of transplantation and integration of iPSC-derived neural progenitor cells (iNPCs) cells into the rodent spinal cord. These iNPCs could be a promising therapeutic strategy for ALS. In the host environment, they differentiate into astroglia and provide the possibility of replacing lost cells, modulating the injury environment, and creating a permissive milieu to protect and regenerate host tissue (Sareen et al., 2014).

Replacement of diseased MNs via transplantation of iPSC derived neuronal cells remains very challenging. Although MN differentiation protocols from PSCs have not achieved pure populations, recent improvements are moving this field in the right direction (Maury et al., 2015). Clinical applications require xenobiotic-free reagents such as replacement of animalderived protein growth factors with defined small molecules for the differentiation procedure. Moreover, grafted human MNs were found to have a rather poor survival rate. Mouse ESC derived MNs on the other hand survive well in transplantation paradigms involving embryonic and neonatal CNS, with their axons innervating muscles (Yohn et al., 2008). Likewise mouse ESC derived MNs can also survive when transplanted into adult mice, and their axons grow to denervated muscles (Deshpande et al., 2006). Yet, hESC-derived MNs transplanted into adult mice fail due to low survival (Lee et al., 2007). A recent study employing iPSC-derived neural progenitors for transplantation into rats, however, was able to report successful grafting and a high survival rate of transplanted cells and specification of grafted cells to MNs in the ventral horn of recipient animals (Popescu et al., 2013).

Despite mostly disappointing outcomes in approaches to graft human MNs in the CNS, the fact that mouse ESC-derived neurons, human iPSC-derived astrocytes, and recently even human neural progenitors can be grafted successfully bears hope for the transplantation of neurons. Strategies to achieve this will have to improve and promote neuron survival of either terminally differentiated neurons or committed neuronal precursors. In addition the grafted cells need to migrate long distances, extend neurites and axons to innervate their appropriate target muscles and make functional connections.

### 'OMICS—UTILIZING BIG DATA TO GENERATE DISEASE SIGNATURES

With a demographically growing number of documented cases of neurodegenerative diseases, the urgency to define the state and behavior of various CNS cells under homeostatic and diseased conditions is apparent. Nonetheless, our understanding of the CNS under diseased conditions such as neurodegeneration observed in ALS, Huntington's disease and PD, despite dramatic progress over the past 3 decades, still is rather limited. Particularly with respect to ALS, while research using animal models has brought significant insight into some of the major events accompanying ALS, yet, not a single significant disease modifying therapy exists. With patient-derived iPSCs technology enabling the generation of CNS specific cell types, new avenues of studying disease progression and signatures in human cells can be explored.

The Library of Integrated Network-Based Cellular Signatures (LINCS) is a collaborative initiative with the aim to catalog cellular processes under conditions such as cell stress and disease. Funded by the National Institute of Health (NIH) the LINCS centers set out to collect large data sets of cellular events in human diseases compared to healthy control samples. Utilizing computational tools to integrate such diverse information into a comprehensive view of molecular disease events and the development of new biomarkers and therapeutics (Vidovic et al., 2014; Liu et al., 2015). That this approach is likely to reveal a wealth of new findings is supported by a line of previous LINCS studies in areas of cell lineage identity, genome modification and cancer research (Duan et al., 2014; Gujral et al., 2014; Ma'ayan and Duan, 2014; Olson et al., 2014; Fallahi-Sichani et al., 2015).

Specifically for neurodegenerative diseases the NeuroLINCS consortium has been established and plans to combine expertise in the fields of iPSC technology, disease modeling, whole genome sequencing, epigenomics, transcriptomics, proteomics, and metabolomics to integrate cell-based assays and highthroughput single-cell analysis with statistics, bioinformatics and computational biology. With regard to ALS this NeuroLINCS consortium would develop signatures of diseased motor neurons and glial cells under various ALS genotypes and baseline conditions. Eventually, we will determine how these patient genotypes in the diseased neurons interact with environmental perturbagens (glutamate and tunicamycin) to elicit responses related to mitochondrial and ER stress. Our understanding of the molecular circumstances of disease progression could be dramatically improved by the generation of quantitative molecular phenotypes and cell signatures of human neurons and glia, providing rational intervention points.

### CONCLUDING REMARKS

More than a century ago, Jean-Martin Charcot first described a disease entity of broad neurodegeneration known as ALS. Since then multiple molecular events associated with ALS disease progression have been identified, each of these working in concert leading to disease onset: (1) Developmental defects caused by genetic mutations making certain MN subtypes susceptible to aberrant maturation during CNS development could lead to early degeneration in adult life; (2) An inert susceptibility of mitochondria, provoked by an inability to balance ROS, making MNs more prone to cell stress and degeneration; (3) Misfolded proteins and protein-RNA aggregates triggering the ER UPR stress responses leading to initiation of MN apoptosis; and (4) The dysregulation of Ca2<sup>+</sup> levels observed during disease progression is a common link to both mitochondrial and ER stress, as both organelles finely tune Ca2<sup>+</sup> homeostasis. Particularly, the latter one provides putative insights into MN susceptibility through mitochondrial stress leading to increased local Ca2<sup>+</sup> levels and ultimately functional alteration of ALS-critical proteins; given that Ca2<sup>+</sup> itself is typically found to affect ALS-associated proteins and processes.

The post-mitotic MNs, however, are not the only cell types affected by the pathological processes and mutant genes involved in ALS. Glial cells, in fact, are functionally impaired, as shown in animal and stem cell models alike, and these ultimately impact MNs as well. Animal models and multidimensional ''organson-a-chip'' will be crucial when future studies are required to dissect CNS cell-cell interactions. However, animal models alone will not enable us to decipher the full complexity of an ALS patient's phenotype. Years of research with animal models have led to mostly ineffective treatments for ALS. In this regard human models utilizing patient iPS cell technology continue to provide a wealth of additional insight into the relevant disease phenotype in pathophysiologically affected human cells. The logical next step will be to use both animal and cell models synergistically and coordinate collaborative efforts of large consortium studies such as the NeuroLINCS initiative. By utilizing 'OMICS, longitudinal imaging, big data and machine-learning approaches on large cohort of ALS patient iPSC-derivatives harboring both familial and sporadic forms of the disease we may be able to discover central ALS-specific mechanistic signatures beyond the phenotype of an individual ALS mutation, and rather also across sporadic disease that affects almost 90% of the patients. Such an approach promises discovery of key biomarkers and pathways that shift subtype-specific motor neurons and glial cells out of balance. Whether iPS cell technology can fulfil such commitments remains to be determined.

Short-term approaches must develop early ALS biomarkers enabling faster patient evaluation and diagnosis. Mid-term goals are to improve MN survival and delay disease progression significantly, either by targeting MNs directly or indirectly via glial cells, or employing small molecules and neurotrophic factors to interfere with ROS production, Ca2<sup>+</sup> deregulation and apoptotic signaling. Long-term strategies will require a detailed understanding of early signs of ALS initiation and the ability to repair and regenerate at least MN sets crucial for survival and extending life span. Recent reports (Popescu et al., 2013), provide a glimpse of hope to utilize patient-derived iPSCs for large scale screening of new diagnostic makers, and the perspective of regenerative cellular therapies to improve patients quality of life.

### FUNDING

This work was supported by Cedars-Sinai institutional startup funds (DS), National Center for Advancing Translational Sciences (NCATS), Grant UL1TR000124 (DS). DS is also supported by funds from National Institute of Neurological Disorders and Stroke (NINDS) grant U54NS091046. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

### ACKNOWLEDGMENTS

We thank Dr. Uthra Rajamani, Vivian Chiu, Caitlin Clapacs, Camille Ocampo and Andrew Gross for critical reading of the manuscript and helpful discussions.

## REFERENCES


neurodegeneration. J. Neurochem. 63, 584–591. doi: 10.1046/j.1471-4159.1994. 63020584.x


in living cells. Proc. Natl. Acad. Sci. U S A 108, 16398–16403. doi: 10.1073/pnas. 1102645108


and induces motor neuron degeneration. PLoS One 8:e54511. doi: 10. 1371/journal.pone.0054511


assessment of mitochondrial function in a neurodegenerative disease: monitoring the primary motor cortex of G93A mice, an amyotrophic lateral sclerosis model. Mitochondrion 10, 108–114. doi: 10.1016/j.mito.2009. 11.001


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Kaus and Sareen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Dysregulated expression of death, stress and mitochondrion related genes in the sciatic nerve of presymptomatic SOD1G93A mouse model of Amyotrophic Lateral Sclerosis**

#### *Chrystian J. Alves, Jessica R. Maximino and Gerson Chadi\**

*Department of Neurology, Neuroregeneration Center, University of São Paulo School of Medicine, São Paulo, Brazil*

#### *Edited by:*

*Manoj Kumar Jaiswal, Center for Neuroscience and Regenerative Medicine, USA*

#### *Reviewed by:*

*Robert Weissert, University of Regensburg, Germany Lei Liu, University of Florida, USA Hakan Muyderman, Flinders University, Australia*

#### *\*Correspondence:*

*Gerson Chadi, Department of Neurology, Neuroregeneration Center, University of São Paulo School of Medicine, Av. Dr. Arnaldo, 455, 2nd Floor, Room 2119, 01246-903 São Paulo, Brazil gerchadi@usp.br*

> *Received: 05 June 2015 Accepted: 10 August 2015 Published: 01 September 2015*

#### *Citation:*

*Alves CJ, Maximino JR and Chadi G (2015) Dysregulated expression of death, stress and mitochondrion related genes in the sciatic nerve of presymptomatic SOD1G93A mouse model of Amyotrophic Lateral Sclerosis. Front. Cell. Neurosci. 9:332. doi: 10.3389/fncel.2015.00332* Schwann cells are the main source of paracrine support to motor neurons. Oxidative stress and mitochondrial dysfunction have been correlated to motor neuron death in Amyotrophic Lateral Sclerosis (ALS). Despite the involvement of Schwann cells in early neuromuscular disruption in ALS, detailed molecular events of a dying-back triggering are unknown. Sciatic nerves of presymptomatic (60-day-old) SOD1G93A mice were submitted to a high-density oligonucleotide microarray analysis. DAVID demonstrated the deregulated genes related to death, stress and mitochondrion, which allowed the identification of Cell cycle, ErbB signaling, Tryptophan metabolism and Rig-I-like receptor signaling as the most representative KEGG pathways. The protein-protein interaction networks based upon deregulated genes have identified the top hubs (TRAF2, H2AFX, E2F1, FOXO3, MSH2, NGFR, TGFBR1) and bottlenecks (TRAF2, E2F1, CDKN1B, TWIST1, FOXO3). Schwann cells were enriched from the sciatic nerve of presymptomatic mice using flow cytometry cell sorting. qPCR showed the up regulated (*Ngfr, Cdnkn1b, E2f1, Traf2 and Erbb3, H2afx*, *Cdkn1a*, *Hspa1*, *Prdx*, *Mapk10*) and down-regulated (*Foxo3, Mtor*) genes in the enriched Schwann cells. In conclusion, molecular analyses in the presymptomatic sciatic nerve demonstrated the involvement of death, oxidative stress, and mitochondrial pathways in the Schwann cell non-autonomous mechanisms in the early stages of ALS.

**Keywords: ALS, SOD1G93A, pre-symptomatic, sciatic nerve, microarray, flow cytometry sorting, Schwann cells, network analysis**

### **Introduction**

Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease characterized by a selective death of motor neurons of the spinal cord, brainstem and cerebral cortex, leading to progressive paralysis. The patient's death is often due to a respiratory failure, usually within 3–5 years from diagnosis (Kiernan et al., 2011).

ALS pathogenesis is poorly understood and the proposed mechanisms related to neurodegeneration seem to be multifactorial and include mitochondrial dysfunction, oxidative stress, protein aggregation and axonal transport impairment (Boillée et al., 2006a; de Vos et al., 2007). Additionally, mechanisms associated with motor neuron degeneration involve non-neuronal cells (Boillée et al., 2006a,b; Yamanaka et al., 2008; Wang et al., 2011), as seen by toxicity in neuron-glial co-culture experiments and also by the activation of molecular pathways related to paracrine stress signaling (Boillée et al., 2006a,b; Ferraiuolo et al., 2007; Kiernan et al., 2011).

Evidence obtained from studies on early axon and neuromuscular junction events in ALS patients and a mouse model have indicated the presence of peripheral motor neuron dysfunction before the activation of death pathways and the onset of clinical symptoms (Fischer et al., 2004; Gould et al., 2006; Moloney et al., 2014). Indeed, the peripheral events seem to proceed toward soma in a retrograde dying-back manner (Coleman and Perry, 2002; Fischer et al., 2004; Saxena and Caroni, 2007; Rocha et al., 2013). Oxidative stress and compromised mitochondria represent one set of proposed mechanisms underlying peripheral ALS dysfunction (Barber et al., 2006; Cozzolino and Carrì, 2012; Cozzolino et al., 2013).

The Schwann cell is the major functional cell type supporting axonal integrity (Mirsky and Jessen, 1999). Recent evidence suggests that Schwann cells may contribute to ALS distal axonopathy (Fischer et al., 2004; de Winter et al., 2006; Gorlewicz et al., 2009; Keller et al., 2009; Lobsiger et al., 2009; Chen et al., 2010; Verheijen et al., 2014). For instance, the up-regulation of inducible nitric oxide synthase and semaphorin 3A in the Schwann cells close to terminal fibers of the sciatic nerve has been associated with dying-back mechanisms in presymptomatic ALS mice (de Winter et al., 2006; Keller et al., 2009; Chen et al., 2010; Malaspina et al., 2010; Venkova et al., 2014). Furthermore, accumulation of axonal ribosomes in Schwann cells bearing mutant hSOD1 in a presymptomatic phase of ALS mouse model further suggests an early involvment of this glial cell type in the pathogenesis of the disease (Verheijen et al., 2014).

Microarray analyses have been useful in identifying new molecular cues potentially involved in ALS pathogenesis both in *postmortem* human tissue and also in several clinical stages of experimental animal models of ALS (Olsen et al., 2001; Hensley et al., 2002; Yoshihara et al., 2002; Dangond et al., 2004; Perrin et al., 2005; Ferraiuolo et al., 2007, 2009; Fukada et al., 2007; Vargas et al., 2008; Kudo et al., 2010; Boutahar et al., 2011; Cooper-Knock et al., 2012; de Oliveira et al., 2013, 2014; Maximino et al., 2014). However, there is a lack of information on gene expression in peripheral motor nerves in ALS despite the importance of recently described dying-back events in this disorder. Furthermore, an evaluation of dysregulated genes in specific, enriched cell populations obtained by cell sorting might extend these molecular analyses at cellular level.

By means of a high-density oligonucleotide microarray analysis linked to specific tools capable of identifying distinct cellular components and biological processes, the aim of this work was to determine whether the expression of genes involved in the regulation of death, stress and mitochondrial function was dysregulated in the sciatic nerve of the SOD1G93A mouse model during the presymptomatic stage of ALS. This work has also evaluated the modulation of selected molecules in enriched sciatic nerve-derived Schwann cells, thus detailing the role of these glial cells in the early phase of this disease.

### **Materials and Methods**

#### **Animal and Tissue Sample**

Transgenic SOD1G93A mice (The Jackson Laboratory, Bar Harbor, ME, USA) were crossbred and the colony was maintained in a specific pathogen-free environment within the animal facility of the University of São Paulo Medical School (São Paulo, Brazil) as described previously (Gurney, 1994; Scorisa et al., 2010; Alves et al., 2011). Animals were kept under controlled temperature and humidity conditions with a standardized light–dark cycle (lights on at 7:00 a.m. and off at 7:00 p.m.) and free access to food pellets and tap water. Mice were genotyped by PCR amplification of tail extracted DNA which identified the presence of the human SOD1 mutated gene (mSOD1) (Gurney, 1994; Scorisa et al., 2010; Alves et al., 2011). The Transgenic SOD1G93A mice express high number of mutant human SOD1 copies (Gurney, 1994; Verheijen et al., 2014). The study was conducted under protocols approved by the Ethical Animal Care and Use Committee at the University of São Paulo and in accordance with the Guide for the Care and Use of Laboratory Animals adopted by the National Institutes of Health.

Sixty-day-old presymptomatic male SOD1G93A mice and their age-paired wild-type controls (∼20–25 g body weight) were used in the experiments. No motor neuron death was seen in any animal at this age (Alves et al., 2011), therefore the animals were chosen for the present presymptomatic analyses. Animals were killed by decapitation and sciatic nerves were removed, frozen and stored at −80◦C for further use. Four mice were used in each group for the microarray experiments. The quantitative polymerase chain reaction (qPCR) analyses of sciatic nerves were performed using samples from six different mice from each transgene and wild-type groups.

#### **Immunofluorescence Labelings and Histological Sections of Sciatic Nerve**

Four animals per genotype were used for immunofluorescence labelings, according to previous publications (Guzen et al., 2009; Batista et al., 2014). Mice were anesthetized with sodium pentobarbital and euthanized by a transcardiac perfusion with 7 ml isotonic saline at room temperature followed by 35 ml fixation fluid (4◦C) over a period of 6 min. The fixative consisted of 4% paraformaldehyde (w/v) in 0.1 M phosphate buffer (pH 6.9). The sciatic nerves were removed, kept in fixative at 4◦C for 90 min and rinsed for 24 h in 10% sucrose dissolved in PBS. Sciatic nerves were then frozen in dry ice-cooled (−40◦C) isopentane (Sigma) and stored at a −80◦C freezer until use. Longitudinal sections (5μm thick) were obtained using a cryostat (Leica, CM3000, Germany). The sections were initially washed for 2 × 10 min in PBS and then were incubated overnight in PBS containing 0.5% Triton X-100 (Sigma) and 1% bovine albumin serum (BSA, Sigma) with a polyclonal rabbit antibody against microtubule associated protein 2 (MAP2; diluted 1:200; Sigma), growth associated protein 43 (GAP-43; diluted 1:200; Sigma), S100 (diluted 1:200; Abcam) and p75NGF neurotrophin receptor (p75; diluted 1:200; Abcam). After the incubation of the primary antibodies, sections were washed for 2 × 10 min in PBS and incubated for 1 h in the dark at 37◦C with a dilution of Alexa Fluor<sup>R</sup> 488 or 594-conjugated secondary antibodies specific for rabbit (1:200, all from Invitrogen, USA). Preparations were mounted on microscope slides and counterstained with nuclear 4 ,6-diamidino-2-phenylindole dihydrochloride (DAPI; Vector, USA). Digital images were obtained by means of an Olympus BX-51 microscope (Olympus, USA).

In addition, four animals per genotype were processed to histological staining using methylene blue. Briefly, sciatic nerves were fixed in 2.5% glutaraldehyde, pH 7.4 (24 h). After extensive wash, tissues were embedded in araldite and transverse semithin sections (0.5μm thickness) were obtained. The sections were stained with methylene blue and photomicrographed using an Olympus BX-51 microscope (Olympus).

#### **RNA Isolation and Microarray Experiments**

The procedures for RNA isolation and microarray experiments with sciatic nerve were described in our previous publication (Maximino et al., 2014). Briefly, RNAs from samples (25 ng) and reference (100 ng) were reverse transcribed by the Lowinput RNA Linear Amplification kit and then transcribed to Cy3 labeled (samples) or Cy5-labeled (reference) RNAs according to the manufacturer's instructions (Agilent Technologies, USA) and to our previously described protocols (de Oliveira et al., 2013).

#### **Microarray Analysis**

Raw image data were converted to numerical data using the Agilent Feature Extraction Software, version 11.0.1.1, as described in our previous study (Maximino et al., 2014; Alves et al., 2015). Raw signal intensities were normalized using the GeneSpring GX v12.6 software package (Agilent Technologies, USA). After normalization, the probes were tested for differential expression. GeneSpring GX was also used in the statistical analyses of gene expressions between genotypes (SOD1G93A <sup>×</sup> wild-type), according to previous publications (Smyth, 2004; Fu et al., 2014; Yang et al., 2014; Ryan et al., 2015; Wang et al., 2015). Genes with *p* < 0.05 were considered differentially expressed. The raw data from hybridizations are available on the Gene Expression Omnibus Database, and the GEO accession number is GSE69450.

#### **Bioinformatic Analysis**

The dysregulated genes were submitted for the following analyses to provide information regarding their involvement with specific cellular/molecular pathways related to ALS:

#### **Functional Enrichment Analysis**

The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7b functional tool (https://david.ncifcrf. gov/) (Huang Da et al., 2008) was used to identify genes related to death, stress and mitochondrial function through the Gene Ontology (GO) annotation database. The DAVID analysis focused on the category: Biological Process and Cellular Component. Stringency (EASE score set to 0.05) parameters were selected to improve confidence in the terms designated as enriched. The Biological Process and Cellular Component terms related to death, stress and mitochondrial function were further organized using the BioVenn tool (http://www.cmbi.ru.nl/cdd/biovenn/) (Hulsen et al., 2008) which identifies common and exclusively expressed genes between lists.

In order to identify over-represented pathways, the GO terms list containing genes related to death, stress and mitochondrion was submitted to the DAVID tool using the Kyoto Encyclopedia of Genes and Genomes (KEGG).

#### **Protein Interaction Network Analysis**

The Cytoscape plugin GeneMANIA (Warde-Farley et al., 2010) was used to predict protein interactions from the list of differentially expressed genes in the sciatic nerve of presymptomatic SOD1G93A mice related to death, stress and mitochondrion. The network was generated using only information derived from the pathway and physical interactions categories in GeneMANIA. The connectivity of the nodes contained in the network was achieved by means of the node centrality parameters "degree" and "betweenness," using the Cytoscape plug-in CentiScaPe (Scardoni et al., 2009). Node degree is a local structure measure in networks that determines the number of edges in each node. Conversely, betweenness centrality is a global structure measure in networks that determines the number of shortest paths passing through a specific node while connecting, directly or indirectly, pairs of nodes (Scardoni et al., 2009). A scatter plot was constructed by inputting node degree and betweenness values for each node in GraphPad Prism 5. The combination of such attributes in the scatter plot allowed the visualization of hubs (nodes with high node degree) and bottlenecks (nodes with high node betweenness). Subnetworks were built using list of genes related to death, stress and mitochondrion separately; scatter plots were also constructed by inputting node degree and betweenness values for each node in GraphPad Prism 5.

#### **Flow Cytometry Sorting of Isolated Schwann Cells and Fibroblast**

Schwann cells and fibroblasts were isolated by means of flow cytometry sorting from sciatic nerve explants of presymptomatic SOD1G93A mice and their age-paired wild-type controls as described in our previous publication (Maximino et al., 2014). The sciatic nerve-derived cell suspension was submitted to a double immunolabeling to identify Schwann cells and fibroblasts by means of a fluorescein isothiocyanate (FITC)-conjugated mouse p75NGF Receptor antibody (Abcam, USA) and a fluorescein phychoerythrin (PE-Cy5)-conjugated monoclonal antibody against Thy-1 (Abcam, USA), respectively. The p75NGF Receptor labeling was employed in the cell sorting experiments because it is a well-characterized surface marker for Schwann cells (Niapour et al., 2010). Cells were then analyzed for type and specificity as well as separated on a FACSAria III Cell Sorter (BD Biosciences, USA). A maximum of 106 cells were ressuspended in 500μl of buffer. Flow cytometry dot plot Schwann cell and fibroblast profiles are shown in **Figures 1A–D**. Of note, the flow cytometry sorting Schwann cells of ALS mice did not show morphological differences (cell size and cytoplasmic granules) compared to control mice (**Figures 1C,D**). Also, the Schwann cells of ALS mice expressed high levels of mutant human SOD1, while no signal was seen in the Schwann cells of wild-type mice, as evidenced by PCR (**Figure 1E**).

Total RNA from enriched Schwann cells was extracted using Trizol (Life Technologies, USA) according to the manufacturer's protocol. The quantity (NanoDrop 1000 Spectrophotometer) and quality (Agilent 2100 bioanalyzer, RNA 6000 Pico LabChip) of RNAs were analyzed as described in our previous publication (Maximino et al., 2014). Also, the Schwann cell samples were submitted to PCR analyses in order to assess contamination from other cell types.

#### **Schwann Cell Enrichment and hSOD1G93A Verification by PCR**

Schwann cells and fibroblast from sciatic nerve of 60-day-old presymptomatic SOD1G93A and their wild-type controls were obtained by fluorescence activated cell sorting and submitted to PCR for sample purity verification. Total RNA from enriched Schwann cells and fibroblast was extracted using Trizol and synthesized in cDNA as described above. Primers to evaluate the presence of Schwann cells *(S100)* and fibroblast *(Thy1)* used in the PCR reactions are shown in **Table 1**, as well as the primer for the internal control *(Actb)*. The reactions were performed to 20μl final volume, using GoTaq Flexi DNA Polymerase (Promega), according to the manufacturer, and 500 nM of each primer. The protocol for PCRs consisted in 95◦C during 5 min, followed by 35 cycles of 95◦C during 30 s, 60◦C during 30 s, 72◦C during 45 s, ending with 72◦C in 7 min. Whole sciatic nerve sample was used as a positive control. PCR products were submitted to electrophoresis in 2% agarosis gel containing ethidium bromide for 60 min at 100 V, and then visualized under UV exposure.

Sciatic nerve and Schwann cells of 60-day-old presymptomatic SOD1G93A and their wild-type controls were submitted to PCR for verification the presence of human SOD1G93A (hSOD1G93A). The procedures were the same as described above. Primers to evaluate the presence of *hSOD1*G93A and *Actb* (internal control) used in the PCR reactions are shown in **Table 1**.

#### **Quantitative PCR**

A subset of genes was chosen for verification of expression patterns based on their possible involvement in ALS/neurodegeneration-related death, stress and mitochondrion or based on their higher level of connectivity (betweenness/degree). qPCR was performed on samples from mouse sciatic nerves and also enriched Schwann cells as previously described (Maximino et al., 2014). Briefly, cDNA was synthesized from 100 ng of total RNA using the Maxima First Strand cDNA Synthesis Kit (Thermo Scientific, USA), according

to manufacturer's instructions. qPCR reactions were carried out in duplicate with 10 ng cDNA, using the DyNAmo ColorFlash SYBR Green qPCR kit (Thermo Scientific, USA) and 400 nM of each primer in a final reaction volume of 20μl. Reactions were run with the Applied Biosystems 7500 Real-Time PCR System (Applied Biosystems). Sequence information regarding the SYBR primers can be found in **Table 1**. qPCR of *Foxo3*

samples obtained by flow cytometry sorting of SOD1G93A mice. Mouse whole

sciatic nerve sample was used as a positive control **(E)**.


**TABLE 1 | Information of primers used to evaluate Schwann cell enrichment, demonstration of hSOD1G93A in sciatic nerve and Schwann cell samples by PCR and to SYBR qPCR experiments in the sciatic nerve and Schwann cells isolated by means of flow cytometry sorting of 60-day-old pre-symptomatic SOD1G93A mice.**

*S100 (S100 calcium binding protein), Thy1 (THYmocyte differentiation antigen 1), hSOD1G93A (mutated human superoxide dismutase 1), Ngfr (Nerve growth factor receptor), H2afx (H2A histone family, member X), Mapk10 (mitogen-activated protein kinase 10), Cdkn1a (cyclin-dependent kinase inhibitor 1A (P21)), Cdkn1b (cyclin-dependent kinase inhibitor 1B), E2f1 (E2F transcription factor 1), Traf2 (TNF receptor-associated factor 2), Foxo3 (forkhead box O3), Mtor (mechanistic target of rapamycin-serine/threonine kinase), Erbb3 (v-erb-b2 erythroblastic leukemia viral oncogene homolog 3-avian), Hspa1a (heat shock protein 1A), Prdx2 (peroxiredoxin 2) and Actb (Actin, Beta).*

was also performed on enriched fibroblasts obtained from the mouse sciatic nerves by means of flow cytometry cell sorting (Figure S1).

Thermocycling conditions for SYBR reactions included an initial denaturation at 95◦C for 10 min. Templates were amplified for 40 cycles at 95◦C for 15 s, and for additional 40 cycles for at 60◦C for 30 s. A dissociation curve was then generated to ensure amplification of a single product, and the verification of no primer dimer formation. A standard curve was generated for each primer pair in order to determine the efficiency of the PCR reaction over a range of template concentrations from 0.032 to 20 ng/μl, using cDNA synthesized from reference mouse RNA. The efficiency for each set of primers was 100 ± 5%. Gene expression was normalized to the expression of *Actb* and determined using the --Ct mathematical model (Livak and Schmittgen, 2001). *Actb* was chosen as a housekeeping gene to normalize the qPCR values because the microarray analysis showed no alteration in expression of this gene across samples.

#### **Statistical Analysis**

The statistical method employed for the microarray analysis is described in details in the microarray analysis section. Furthermore, a two-tailed unpaired *t*-test was used to evaluate the level of significance of gene expression independently between the two genotypes (SOD1G93A <sup>×</sup> wild-type) in the qPCR analyses. Analyses were performed using the GraphPad Prism 5 (San Diego, CA). Data were presented as Means ± Standard Error of Mean (SEM) and significance level was set to *p* < 0.05.

### **Results**

#### **Qualitative Analyses of Histological Sections and Enriched Schwann Cells of the Sciatic Nerves**

No qualitative changes were found regarding morphology of sciatic nerves of presymptomatic ALS mice compared to control at histopathological examination (**Figures 2A–J**). PCR analysis of sciatic nerve and Schwann cells enriched by flow cytometry showed the presence of hSOD1G93A in the SOD1G93A mice, but not in the wild-type controls (**Figures 2K,L**). Also, the flow cytometry sorting Schwann cells of ALS mice did not show morphological differences (cell size and cytoplasmic granules) compared to control mice (**Figures 1C,D**).

#### **Verification of Microarray Results by Quantitative PCR**

The results of qPCR verification of 10 representative genes in the sciatic nerve of presymptomatic SOD1G93A mice are shown in **Table 2**. The up and down-regulations of the verified genes in the sciatic nerve of 60-day-old SOD1G93A mice by means of qPCR were coincident and supported the microarray findings of correspondent animal ages (Table S1).

## **Bioinformatics Analysis**

#### **Functional Enrichment Analysis**

The DAVID analysis of differentially expressed genes in the sciatic nerve of 60-day-old SOD1G93A mice revealed 19 GO terms of Biological Process related to cell death and apoptosis (genes related to Death in **Table 3**). These GO terms of genes related

**FIGURE 2 | Histopathological analysis (A–J) and molecular evaluation of** *hSOD1***G93A signal (B) in the sciatic nerve and Schwann cell samples of 60-day-old presymptomatic SOD1G93A and aged paired wild-type mice.** Immunofluorescence staining of MAP2 (**A,B**; red), GAP-43 (**C,D**; green), S100 (**E,F**; green) and p75NGF (**G,H**; green) in the sciatic nerve of 60-day-old presymptomatic SOD1G93A mice **(B,D,F,H)** and their wild-type controls **(A,C,E,G)**. MAP2 and GAP-43 are markers of neuronal fibers; S100 and p75NGF are markers of Schwann cells. Cell nuclei were stained with DAPI (blue). The insert boxes in the bottom left of images show a higher magnification of the cell profiles. Methylene blue staining of Schwann cell myelin sheets of sciatic nerve of 60-day-old presymptomatic SOD1G93A mice **(J)** and their wild-type controls **(I)** are also seen. Scale bars: 10 μm. Of note, the same staining pattern was observed for both genotypes (SOD1G93A and wild-type controls) for all cell markers and for the histological sections. Representative bands of PCR for specific gene markers of human SOD1G93A (*hSOD1<sup>G</sup>*93A) and actin b (*Actb*) in sciatic nerve **(K)** and Schwann cells enriched samples **(L)** obtained by flow cytometry sorting of SOD1G93A and wild-type control mice.



*qPCR of differentially expressed genes in sciatic nerves of SOD1G93A mice compared to the wild-type controls selected from microarray data (p* < *0.05; see text for details). The genes are related to death, stress and mitochondrion. The means of fold change and p-values are shown, according unpaired two-tailed t-test. The regulation of studied genes for verification showed the same direction as that found in the microarray experiments (Table 3).*

to Death showed 112 dysregulated genes (46 down and 66 upregulated genes). Furthermore, DAVID also identified two GO terms of Biological Process related to Stress (**Table 3**). These GO terms related to Stress showed 66 dysregulated genes (31 down and 35 up-regulated genes). Finally, DAVID identified two GO terms of Cellular Components related to mitochondrial function (**Table 3**). These GO terms of Mitochondrion showed 143 dysregulated genes (91 down and 52 up-regulated genes). **Table 3** shows the down regulated and up regulated genes of Death, Stress, Mitochondrion categories with fold change equal or higher than 1.5. The deregulated genes of these categories with fold change smaller than that are shown in the Table S2 of the Supplementary Material. We have not expected to find gene regulation with high degree of fold change in this stage of presymptomatic events, a period in which dramatic occurrences related to inflammation, neurodegeneration and necrosis are not taking place. The above-mentioned cut-off in the **Table 3** is just to facilitate the demonstration of the relatively higher deregulated genes. All analyses of this study were performed without a cut-off.

A VEEN diagram of those dysregulated genes related to Death and Stress and Mitochondrion showed 17 genes belonging to both Death and Stress groups of genes, 9 genes belonging to both Death and Mitochondrion groups of genes, 5 genes belonging to both Stress and Mitochondrion groups of genes, and finally 3 genes belonging to the three, Death, Stress, and Mitochondrion groups of genes (**Figure 3**).

The DAVID analysis of dysregulated genes related to Death, Stress and Mitochondrion shown in **Table 3** are identified with 16 KEGG pathways (**Table 4**). Furthermore, those identified KEGG pathways that have been previously described in the literature as being related to ALS include RIG-I-like receptor signaling (2 down and 3 up-regulated genes; **Figure 4**), tryptophan metabolism (5 down-regulated genes; **Figure 4**), ErbB signaling (1 down and 5 up-regulated genes; **Figure 4**) and cell cycle (2 down and 5 up-regulated genes; **Figure 4**). Of note, Cancer,


*(Continued)*


*(Continued)*

**318**

**TABLE 3 |** 


Small cell lung cancer, the Adipocytokine signaling pathway and Chronic myeloid leukemia pathways are likely not related to ALS and were not included in **Table 4**.

#### **Protein Interaction Network Analysis from Dysregulated Genes**

Protein interaction network analysis using dysregulated genes related to Death, Stress and Mitochondrion showed the hubs (**Figure 5A**) TRAF2 (16 connectors), H2AFX (8 connectors) and E2F1, FOXO3, MSH2, NGFR, TGFBR1 (7 connectors). Furthermore, the network analysis showed five bottlenecks (TRAF2, E2F1, CDKN1B, TWIST1, and FOXO3). The scatter plot of the values of the hubs node degree vs. the values of node betweenness is shown in **Figure 5B**. Of note, TRAF2, E2F1, CDKN1B, FOXO3, and H2AFX occupied the highest positions in the scatter plot (**Figure 5B**). Networks and scatter plots produced from the analyses of the specific genes of Death, Stress and Mitochondrion categories were shown in Figures S2–S5 of Supplementary Material.

### **Schwann Cell and qPCR Experiments**

The results of selected genes for verification in the mouse sciatic nerves by qPCR are shown in the **Table 2**. qPCR analyses of gene expression from the enriched Schwann cells isolated from sciatic nerves of presymptomatic 60-day-old SOD1G93A mice identified a number of up-regulated genes including *Ngfr* (6.85 fold)*, Cdkn1b* (1.51-fold)*, E2f1* (1.89-fold)*, Traf2* (1.31-fold) and *Erbb3* (1.38-fold) related to Death (**Figure 6A**). The gene *H2afx*

*15, Oliveira et al., 1994; 16, Cova et al., 2010; 17, Kudo et al., 2010; 18, Beers et al., 2011;19,* *et al., 2012; 25, Morimoto et al., 2007; 26, Yang et al., 2001; 27, Zhang et al., 2011; 28, Körner et al., 2013; 29, Reinholz et al., 1999; 30,*

*Lopez-Lopez*

 *et al., 2014; 20, Gorlewicz et al., 2009; 21, Ma et al., 2013; 22, Manabe et al., 2003; 23, Morimoto et al., 2012; 24, Thau*

*Nikolic-Kokic*

 *et al., 2006.*


#### **TABLE 4 | KEGG pathways obtained from gene lists related to Death, Stress and Mitochondrion from 60-day-old SOD1G93A sciatic nerve microarray analyses.**

*KEGG pathways were obtained from gene lists related to Death, Stress and Mitochondrion from 60-day-old SOD1G93A sciatic nerve microarray analyses and genes included in these pathways and categories are shown in the table. EASE score was set to 0.05.*

(1.42-fold) of the Stress group was up-regulated. The genes *Cdkn1a* (1.95-fold) and *Foxo3* (-1.67-fold) from the Death and Stress groups were up- and down regulated, respectively. The genes *Hspa1a* (4.48-fold) and *Prdx2* (1.48-fold) of the Death, Stress and Mitochondrion groups were up-regulated. *Mapk10* (1.42-fold) of the Stress and Mitochondrion groups was upregulated and *Mtor* (−1.86-fold), related to mitochondrion, was down regulated (**Figure 6B**). Interestingly, the expression of *Foxo3* was not altered in the enriched fibroblasts of the mouse sciatic nerve (Figure S1).

### **Discussion**

#### **Triggering Mechanisms of Motor Neuron Death in ALS**

Several mechanisms have been proposed as triggers for both autonomous and non-autonomous events related to motor neuron death in ALS. Events of oxidative stress, neuroimmune reactions, protein aggregation, glutamate excitotoxicity, mitochondrial dysfunction and impaired axonal transport in ALS are currently under investigation (Bruijn et al., 2004; Boillée et al., 2006a; de Vos et al., 2007; Jaiswal and Keller, 2009; Redler and Dokholyan, 2012).

The initial triggering and the secondary reactive events associated with motor neurons and their neighboring glial cells are unknown (Tapia, 2014). For instance, early findings related to morpho/physiological changes have been described in presymptomatic phases of the ALS mouse model (Boillée et al., 2006a; Ferraiuolo et al., 2007; Alves et al., 2011; de Oliveira et al., 2013, 2014; Maximino et al., 2014). Of note, fragmentation of the Golgi apparatus (Mourelatos et al., 1996), vacuolization of mitochondria (Bendotti et al., 2001), deficits in axonal transport (Ikenaka et al., 2012), endoplasmic reticulum stress (Tadic et al., 2014), the activation of glial cells (microglia and astrocytes) (Nagai et al., 2007; Graber et al., 2010) and electrophysiological changes (Quinlan, 2011) have all been described within motor neurons and neighboring cells at early postnatal ages. Technological advances have facilitated the discovery of molecular events underlying both autonomous (Wada et al., 2012) and non-autonomous (Ferraiuolo, 2014) mechanisms possibly related to early presymptomatic neuronal toxicity in ALS (Arbour et al., 2015; Saba et al., 2015).

Evidence indicates the existence of very early events, possibly anticipating the motor neuron death, that take place peripherally, i.e., in the motor nerve, motor nerve terminals, neuromuscular junction and muscle (Rocha et al., 2013; de Oliveira et al., 2014; Moloney et al., 2014). In fact, very early electrophysiological

events in the motor nerve precede cell body disappearance in the spinal cord as well as the onset of clinical symptoms in ALS (Alves et al., 2011). Absence of neuronal death or morphological alterations in the motor neuron cell bodies has been described in the 60-days-old presymptomatic SOD1 mice (Fischer et al., 2004; Gould et al., 2006; Casas et al., 2013).

Remarkably, early morphological, biochemical and molecular changes in peripheral non-neuronal cells, i.e., Schwann cells and skeletal muscles, are emerging as dying-back mechanisms of motor neuron degeneration in ALS (de Winter et al., 2006; Dupuis and Loeffler, 2009; Dupuis et al., 2009; Keller et al., 2009; Narai et al., 2009; Chen et al., 2010; Dadon-Nachum et al., 2011; Venkova et al., 2014). In fact, the involvement of Schwann cells, which maintain close morphological/physiological relationships with motor axons, has gained more attention in ALS (de Winter et al., 2006; Keller et al., 2009; Chen et al., 2010; Maximino et al., 2014; Venkova et al., 2014). Our work has contributed to a more detailed understanding of the mechanisms involved in the dyingback events associated with motor nerves in ALS by performing a large-scale gene profiling analysis utilizing a microarray analyses in the sciatic nerve of presymptomatic ALS mice.

A deficit in the paracrine trophic interactions between Schwann cells and motor neurons in presymptomatic ALS is a logical choice for further study since Schwann cells offer the main source of trophic stimuli for the maintenance of mature motor neurons (Bhatheja and Field, 2006) and for the regeneration of their fibers after injury (Gupta et al., 2005). In support of this hypothesis, mSOD1 gene expression in distal Schwann cells has been suggested to interfere with the trophic maintenance of motor axon projections (Inoue et al., 2003). Also, reduction of Schwann cell-derived insulin growth factor-1 and cilliary neurotrophic factor in sciatic nerves of both ALS mice and

shown in **(B)**, as described in the text. Up and down-regulated genes are represented respectively as red and green diamonds. Nodes with the highest values for node degree (number of connections) and node betweenness (number of shortest paths) are represented with a yellow border. Of note from this analysis, the genes *E2f1, Foxo3, Gli3, Ngfr, Cdkn1a* or their related products were already described in the context of ALS.

patients have been correlated to disease worsening (Lee et al., 1996; Lobsiger et al., 2009).

In addition to a deficit in paracrine trophic factor maintenance, the presence of toxic factors from Schwann cells affecting motor neurons at early presymptomatic stages in ALS is an additional matter for investigation. Indeed, our DAVID analyses of dysregulated gene expression in sciatic nerves of presymptomatic SOD1G93A mouse model highlighted the GO categories death, stress and mitochondrion, which may be related to the above mentioned possibilities. Of note, the findings of a high number of dysregulated genes related to death, with many also included in the stress and mitochondrial function categories, indicated a complex regulation of these events before motor neuron death, as described for other

neurodegenerative disorders (Friedman et al., 1996; Gatzinsky et al., 2003; Pun et al., 2006; Campana, 2007; Lobsiger et al., 2009; Nobbio et al., 2009a,b). Conversely, negative and positive gene regulation related to death and anti-apoptosis mechanisms were also identified in several categories of dysregulated genes, indicating a concomitant regulation of cell toxicity/maintenance before neurodegeneration during the presymptomatic stages in ALS. It should be emphasized that the present description of gene profiling in the sciatic nerve in the presymptomatic ALS mouse model may represent predominantly the regulation of Schwann cells transcripts. In fact, as discussed below, the qPCR verification of selected genes in the mouse sciatic nerve tissue has confirmed the microarray results. qPCR results of selected genes in the enriched Schwann cells were also coincident to microarray. Conversely, the regulation of *Foxo3* in the enriched fibroblasts of the sciatic nerve was in the opposite direction.

#### **Mitochondrial/Oxidative Stress Mechanisms Related to Presymptomatic Motor Neuron Impairment**

Motor neurons are susceptible to reactive oxygen species and oxidative stress due to their high metabolic rates and decreased ability to buffer calcium (Shaw and Eggett, 2000). Long axons and increased functional activity also confer a high susceptibility to mitochondrial impairments in motor neurons (Barber et al., 2006; Pizzuti and Petrucci, 2011; Cozzolino et al., 2013).

Altered morphology of motor neuron mitochondria has been described in an ALS mouse model (Parone et al., 2013) as well as in humans (Hirano et al., 1984b) at both histological (Hirano et al., 1984a,b) and ultrastructural (Sasaki and Iwata, 1996) levels. Of note, swollen and vacuolated mitochondria were found in motor neurons, glial cells and muscles in ALS (Afifi et al., 1966; Atsumi, 1981; Siklós et al., 1996; Sasaki et al., 2007; Cassina et al., 2008).

Furthermore, oxidative stress has been described in several regions of the brain (Ferrante et al., 1997; Bogdanov et al., 1998, 2000), in the spinal cord (Shaw et al., 1995; Andrus et al., 1998; Liu et al., 1998, 2004; Shibata et al., 2002) and in the skeletal muscle (Mahoney et al., 2006) of both ALS mice and humans. Altered levels of reactive oxygen species within the spinal cord mitochondria of ALS mice (Jung et al., 2002) and patients (Bogdanov et al., 2000) support a possible impairment of the electron transport chain and energy defects in this disorder. Indeed, mitochondrion-induced damage to motor neurons has previously been mentioned as an early ALS signal (Wong et al., 1995; Mattiazzi et al., 2002; Kirkinezos et al., 2005; Cassina et al., 2008; Loizzo et al., 2010). Furthermore, peroxynitrite and superoxide overload in reactive astrocytes and microglia in *in vitro* models of ALS lead to protein impairment-induced motor neuron damage (Hensley et al., 2006; Dadon-Nachum et al., 2011). These studies all underline the critical involvement of mitochondrial dysfunction in early autonomous and nonautonomous pathogenesis of ALS.

Interestingly, increases in inducible nitric oxide synthase and peroxynitrite in Schwann cells and motor axons of paranodal regions in presymptomatic ALS mice were associated with local mitochondrial reactive oxygen species formation (Chen et al., 2010). Furthermore, Schwann cell-induced trophic support failure and Schwann cell-induced mitochondrial toxicity to motor neurons have both been correlated with high levels of mSOD1 in those peripheral glia of presymptomatic SOD1G93A ALS mice (Gould et al., 2006). Thus, Schwann cell mitochondrion/oxidative stress mechanisms seem to play a key role during the early stages of ALS.

#### **KEGG Pathways Related to Death, Stress and Mitochondrion in ALS**

Our DAVID analyses identified genes associated with death, stress and mitochondrial function from the complete list of dysregulated genes of the sciatic nerve from 60-day-old presymptomatic ALS mice before the onset of neurological impairment.

Dysregulated genes of death signaling (Olsen et al., 2001; Hensley et al., 2002; Yoshihara et al., 2002; Dangond et al., 2004;

respectively, according to unpaired two-tailed *t*-test.

Ferraiuolo et al., 2007; Guipponi et al., 2010) and also those of stress and mitochondrion-related signaling (Dangond et al., 2004; D'arrigo et al., 2010; Guipponi et al., 2010; Bernardini et al., 2013) have been described in ALS models (spinal cord and motor neurons) and in patients (e.g., from *post mortem* spinal cord and skeletal muscle biopsy) by means of microarray technology. Thus, our work supplements those cited by providing a largescale profile of genes related to death, stress and mitochondrial function in the peripheral motor nerves of presymptomatic ALS mice.

KEGG pathways from the dysregulated genes associated with death, stress and mitochondrial function of sciatic nerves of presymptomatic ALS mice identified valine, leucine and isoleucine degradation and propionate metabolism as pathways with the highest number of dysregulated genes. Of note, impairments in valine, leucine and isoleucine metabolism (Ilzecka et al., 2003) but not in propionate metabolism, have been described in the context of ALS. Furthermore, fatty acid metabolism, citrate cycle, gluthatione metabolism, arginine and proline metabolism, pyruvate metabolism, sphingolipid metabolism, synthesis and degradation of ketone bodies, and beta-alanine metabolism pathways have been correlated with the onset of ALS (Ariga et al., 1998; Schulz et al., 2000; Cutler et al., 2002; Zhao et al., 2006; Nunn and Ponnusamy, 2009; Panov et al., 2011; Dormann et al., 2012; Zhang and Chook, 2012; de Munck et al., 2013; Yip et al., 2013; Allen et al., 2014). Additionally, dysregulated genes related to RIG-I-like receptor signaling (Day et al., 2005; Katsuno et al., 2011; Zhu et al., 2014), tryptophan metabolism (Sandyk, 2006), ErbB signaling (Gorlewicz et al., 2009), and cell cycle (Manzano et al., 2013) have been reported in ALS (Gorlewicz et al., 2009; Manzano et al., 2013).

Interestingly, the majority of dysregulated genes of the KEGG pathways RIG-I-like receptor signaling, ErbB signaling and cell cycle were found to be up-regulated. Conversely, all genes of tryptophan metabolism were down regulated in the sciatic nerve of presymptomatic ALS mice. The understanding on how such a complex regulation of important singling pathways may represent early triggering or reactive responses in the dyingback mechanisms in ALS is a matter for further analyses. RIG-I-like receptor genes are regulated in ALS, however, the exact mechanisms by which innate immune activation may drive neuronal death in neurodegenerative disorders are far from elucidated. In particular, the role of RIG-I-like receptor gene products in those mechanisms underlying glial cell activation, misfolded proteins/aberrantly localized nucleic acids and mitochondrion signaling-induced autophagy remain unclear (Kawai and Akira, 2008; Tal and Iwasaki, 2009; Heneka et al., 2014; Ying et al., 2015). Furthermore, the involvement of ErbB signaling in ALS was raised with the description of disrupted neuregulin-ErbB4 pathway signaling in presynaptic synapses of less resistant spinal cord motor neurons (Takahashi et al., 2013), but not in the resistant oculomotor neurons in human ALS (Gallart-Palau et al., 2014). Furthermore, ErbB signaling may influence non-autonomous microglial ALS mechanisms (Falls, 2003; Esper et al., 2006; Song et al., 2012) as well as Schwann cellinduced motor axon terminal changes in ALS (Gorlewicz et al., 2009).

The large-scale down-regulation of genes associated with the tryptophan metabolism pathway may be related to serotonin's ability to modulate glutamate motor neuron transmission (Palchaudhuri and Flügge, 2005), and thus excitotoxicity in ALS. Indeed, evidence of serotonin deregulation has been obtained from studies on ALS patients (Sandyk, 2006).

As discussed below, cell cycle gene regulation in the sciatic nerve in presymptomatic phases of ALS might disrupt the normal interactions between reactive Schwann cells and motor axons. In fact, cell cycle impairments have been described in neurodegenerative diseases (Van Leeuwen and Hoozemans, 2015; Wojsiat et al., 2015) and the interaction between mSOD1 and cyclin regulators seems to contribute to autonomous and non-autonomous mechanisms in ALS (Nguyen et al., 2003; Ranganathan and Bowser, 2003; Cova et al., 2010).

#### **Schwann Cell Genes Related to Death, Stress and Mitochondrial Function in the Sciatic Nerve of Presymptomatic ALS Mouse Model**

Our study demonstrated the existence of highly interconnected genes from the lists of dysregulated genes involved in death and stress signaling as well as mitochondrial function in the sciatic nerve of presymptomatic ALS mice. The regulation of these genes was further evaluated in the enriched Schwann cells from the ALS mice using qPCR.

*Traf2*, *E2f1*, and *Cdkn1b* from the list of genes of GO biological process related to death were identified as nodes with highest values for node degree (number of connections) and node betweenness (number of shortest paths) in the Protein interaction network analysis.

Mutant SOD1 led to TNF-α pathway activation in the absence of inflammation in ALS (Carter et al., 2009), an event that may also occur in Schwann cells (Au and Yeh, 2007) based on the results of *Traf2* up-regulation in ALS Schwann cells described in our work.

TNFα is a potent regulator of Schwann cell division (Chandross et al., 1996) and activation (Bonetti et al., 2000). Moreover, despite a lack of histopathological descriptions of peripheral nerves from ALS subjects and animal models (Fischer et al., 2004; Kano et al., 2012), there are no reports describing local Schwann cell division or death. However, up-regulation of markers of glial activation indicated an early process of Schwann cell reaction in ALS peripheral nerves (Keller et al., 2009; Malaspina et al., 2010). Cell division is a common feature of activated central glia (Pekny and Nilsson, 2005; Dheen et al., 2007). Therefore, it might be possible that the absence of cell division would disrupt normal functions of activated Schwann cells in presymptomatic stages of ALS.

It is likely that the up-regulation *Cdkn1b* and *E2f-1* may lead to intracellular mobilization of cell-cycle proteins and transcriptional regulators in ALS glia (Weinberg, 1995; Ranganathan and Bowser, 2003) and also in ALS activated Schwann cells. These events may interfere with the balance between death/survival signaling pathways in TNFα-activated Schwann cells (Tang et al., 2013), possibly altering the normal functions of activated Schwann cells and leading to a failure of trophic surveillance and/or to toxicity signaling in ALS (Ranganathan and Bowser, 2003; Cova et al., 2010). Therefore, the absence of hyperplasia may impair substantially the paracrine trophic mechanisms of TNFα-activated Schwann cells with motor neurons in ALS.

Other nodes with high values for connections and betweenness described in the Protein interaction network analysis support a non-autonomous mechanism of activated Schwann cells in ALS. The nodes *Ngfr*, *Erbb3*, and *H2afx* of dysregulated genes from the death list further indicated the presence of impaired paracrine trophic functions of Schwann cells in ALS. The up-regulation of *Ngfr* and *Erbb3*, which encode the high-affinity neurotrophin receptor TRKA and the neuregulin-associated tyrosine kinase receptor, respectively, could be involved in Schwann cell paracrine functions (Wang et al., 1996; Lyons et al., 2005; Adilakshmi et al., 2011) as well as in early, complex mechanisms of axonal retraction and neuromuscular junction alterations in ALS prior to motor neuron degeneration (Kerkhoff et al., 1991; Gorlewicz et al., 2009). Furthermore, the up-regulation of *H2afx* in Schwann cells might correlate with the described Schwann cell participation in the pathogenesis of ALS since H2AFX-induced DNA damage in reactive astrocytes has been associated with impaired paracrine glial functions in other neurodegenerative disorder (Simpson et al., 2010).

*Foxo3* and *Cdkn1a* are present in the death and stress lists of genes and were also found to be dysregulated in the enriched Schwann cells from the sciatic nerve of 60-day-old presymptomatic ALS mice. Indeed, the encoded products of these genes have been investigated in ALS skeletal muscles but not in the ALS motor nerve (Léger et al., 2006; Gonzalez de Aguilar et al., 2008; Manzano et al., 2013). *Foxo3* showed high levels of betweenness and degree and FOXO3 has been studied as a catabolic target in ALS skeletal muscles (Léger et al., 2006). Conversely, *Cdkn1a* dysregulation was described in muscles of ALS mice (Gonzalez de Aguilar et al., 2008) and CDKN1A, a cyclin-dependent kinase inhibitor, interferes with satellite cell-induced myoplasticity in ALS skeletal muscles (Manzano et al., 2013). The above findings provide support for the involvement of *Foxo3* and *Cdkn1a* in the dying-back Schwann cell mechanisms in ALS, a matter that should be further investigated.

*Hspa1a*, *Prdx2*, and *Rrm2b* were present in three GO lists, the death, stress, and mitochondrial function categories. *Hspa1a* and *Prdx2* were also found to be dysregulated in the enriched Schwann cells from the sciatic nerves of presymptomatic ALS mice. Those three genes and their encoded products have not been described in the context of ALS. Moreover, up-regulation of *Hspa1a* and *Prdx2* were found in other neurodegenerative disorders and their encoded products have been studied in the context of therapeutic strategies (Muchowski and Wacker, 2005; Ali et al., 2010; Gestwicki and Garza, 2012). Furthermore, RRM2B-related mitochondrial diseases are frequently inherited and associated with neurological symptoms (Horga et al., 2014). Conversely, non-inherited mitochondrial dysfunctions have been suggested as possible mechanisms underlying the pathogenesis of ALS (discussed above). Thus, mitochondria/oxidative stress impairing Schwann cell paracrine regulation in the presymptomatic phases of ALS is a matter for further investigation.

*Mtor,* identified in the mitochondrion list of dysregulated genes, was down-regulated in the enriched Schwann cells from presymptomatic SOD1 mice, a change which is in line with descriptions of MTOR reduction in spinal cords of ALS rodents (Morimoto et al., 2007). MTOR reduction induced by PI3-K and Akt/PKB signaling (Nagano et al., 2002) worsened ALS pathology in ALS transgenic mice (Zhang et al., 2011). Because MTOR activation leads to neuroprotection in ALS (Saxena et al., 2013), it is a matter of further investigation whether modulation of MTOR signaling in Schwann cells could counteract the degenerative processes associated with this disease.

Finally, *Mapk10* of mitochondrion and stress categories was found to be up-regulated in the enriched Schwann cells from sciatic nerves of presymptomatic ALS mice. A MAPK10/JNK3 truncation mutation has previously only been associated with cognitive disorders (Kunde et al., 2013). Furthermore, *Mapk10* itself has not been investigated as a factor in the pathogenesis of ALS, despite the general consensus that deregulated MAPK signaling contributes to ALS (Kim and Choi, 2010; Xia et al., 2015). Nevertheless, the involvement of p38MAPK and TNFα receptors in the non-autonomous microglial toxicity in ALS (Veglianese et al., 2006) raised the possibility of a similar mechanism in ALS Schwann cells, an issue that should be explored further.

In conclusion, our large-scale gene profiling reveled the presence of death, stress (with emphasis on oxidative stress), and mitochondrial function pathway signaling taking place in the sciatic nerve of presymptomatic ALS mice. The regulation of highly interconnected genes in ALS Schwann cells indicated the involvement of these pathways in the non-autonomous mechanisms of that peripheral glial cell type in the early stages of the disorder.

### **Author Contributions**

JM and GC designed the study. CA and JM performed the experiments. GC interpreted the results. All authors wrote, read and approved the final manuscript. GC is responsible for the ALS Brazil Project of the University of São Paulo School of Medicine.

### **Acknowledgments**

We thank professor Robert Carlone of Brock University for reading the manuscript and for his valuable suggestions and we also thank Hatylas Azevedo for helping us with the Protein interaction network analysis. This work was supported by grants from the São Paulo Research Foundation (FAPESP; #2010/20457-7) and National Council for Scientific and Technological Development (CNPq).

### **Supplementary Material**

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2015.00332

### **References**


axonopathy: evidence in mice and man. *Exp. Neurol.* 185, 232–240. doi: 10.1016/j.expneurol.2003.10.004


PGC-1alpha-regulated factors in the SOD1G93A ALS mouse model and in human sporadic ALS. *J. Neuropathol. Exp. Neurol.* 71, 1064–1074. doi: 10.1097/NEN.0b013e318275df4b


of the Fused in Sarcoma protein (FUS). *Proc. Natl. Acad. Sci. U.S.A.* 109, 12017–12021. doi: 10.1073/pnas.1207247109


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Alves, Maximino and Chadi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Promise and Pitfalls of Mitochondrial Replacement for Prevention and Cure of Heritable Neurodegenerative Diseases Caused by Deleterious Mutations in Mitochondrial DNA

#### Ananta Paine<sup>1</sup> and Manoj Kumar Jaiswal 2, 3 \*

*<sup>1</sup> Division of Allergy/Immunology and Rheumatology, University of Rochester Medical Center, Rochester, NY, USA, <sup>2</sup> Molecular Imaging and Neuropathology Division, New York State Psychiatry Institute, Columbia University, New York, NY, USA, <sup>3</sup> Department of Psychiatry, Columbia University, New York, NY, USA*

Keywords: neurodegenerative disease, mitochondria, mitochondrial gene transfer (MGT), mitochondrial replacement techniques (MRT), mitochondrial DNA (mtDNA), mitotherapy

Mitochondria are cytoplasmic organelles present in eukaryotic cells that serve as major source of cellular energy produced through oxidative phosphorylation and thus also known as the power plants of eukaryotic cells. A distinct feature of mitochondria is its capacity to regenerate owing to own set of genomic material containing 37 genes. It is now well-known that mutations in mitochondrial DNA can cause many inherited diseases including ones that affects neurons and nervous systems. Importantly, in contrast to other cells, neuron rely heavily upon mitochondria due to their inability to derive sufficient energy though glycolysis (Wallace et al., 2010). As a results mitochondrial dysfunction severely affects neuronal cells and proved to be central in the pathogenesis of many neurodegenerative diseases such as Amyotrophic lateral sclerosis (ALS), Parkinson's disease (PD), Huntington's disease (HD), Alzheimer's disease (AD) and many others (Johri and Beal, 2012). According to Friedrich Max Müller, Mitochondria's are "Advaya" ( ) which in Sanskrit means unique and in Upanishads it means ultimate (Max Müller, 1823–1990). Mitochondria are cell's energy factories, divide, multiply, manufacture ATP to fuel all of life's activities. Mitochondrial DNA replacement has been successful in mice and primates and with the refinement of MRT, we hope that it becomes a reality in human. There are more than ∼700 known disease-associated mitochondrial DNA (mtDNA) mutations (mitomap.org). Up to 4000 children per year in the US are born with inherited mtDNA disorders (Schaefer et al., 2008).

In recent years, major advances in the field of nuclear transfer techniques offer the possibility to transfer the nuclear material from one cells with damaged and dysfunctional mitochondria to into another cell only containing cytoplasmic material resulted from careful removal of the nuclear material before the transfer. Such transfer gives rise to cells where damaged and dysfunctional mitochondria gets replaced by healthy mitochondria from the donor cells keeping

#### Edited by:

*Rosanna Parlato, University of Ulm, Germany*

Reviewed by: *Jari Koistinaho, University of Eastern Finland, Finland*

\*Correspondence:

*Manoj Kumar Jaiswal mj2750@cumc.columbia.edu*

Received: *10 June 2016* Accepted: *07 September 2016* Published: *23 September 2016*

#### Citation:

*Paine A and Jaiswal MK (2016) Promise and Pitfalls of Mitochondrial Replacement for Prevention and Cure of Heritable Neurodegenerative Diseases Caused by Deleterious Mutations in Mitochondrial DNA. Front. Cell. Neurosci. 10:219. doi: 10.3389/fncel.2016.00219*

**Abbreviations:** AD, Alzheimer's disease; ALS, amyotrophic lateral sclerosis; CRISPR/Cas, clustered regularly interspaced short palindromic repeats and CRISPR-associated; fALS, familial amyotrophic lateral sclerosis; FDA, food and drug administration; HD, Huntington's disease; HFEA, human fertilization and embryology authority; MRT, mitochondrial replacement technique; mHTT, mhuntingtin; mtDNA, mitochondrial DNA; mSOD1, mutant superoxide dismutase 1; MNs, motor neurons; MND, motor neuron disease; MQC, mitochondrial quality control; PD, Parkinson's disease; PINK1, PTENinduced putative kinase 1; PBT, polar body transfer; PNT, pronuclear transfer; ROS, reactive oxygen species; ST, spindle transfer; SOD1, superoxide dismutase 1; TALENs, Transcription Activators like effector nucleases; [Ca2+]<sup>i</sup> , cytosolic calcium.

the nuclear genomic material unaltered (Falk et al., 2016). This is now known as mitochondrial replacement and the associated techniques are known as mitochondrial replacement techniques (MRT). Mitochondrial gene replacement in oocytes leads to complete replacement of entire mtDNA, applicable to any mtDNA mutation type and eliminates entire spectrum of mtDNA disease. Since genetic corrections will be heritable and passed on to later generations, MRT prevents the need for repeated therapy generation after generation. This significant progress raises the hope that replacement of affected mitochondria in patient's cells can provide curative measures for several mitochondrial diseases. In spite of various ethical and technical concerns, clinicians see huge potential of MRT for cures of several devastating mitochondrial diseases. In this opinion article we highlights an update on new advances and implications of mitochondrial therapy in neurodegenerative disorders and provide insights into studies, suggesting limitations of this advance technology and its future use in clinics. Our hope is that this article will provide a platform for further critical discussion of this pertinent issue.

### MITOCHONDRIAL DYSFUNCTION AND NEURODEGENERATIVE DISEASES

Neurological diseases are heterogeneous in nature and affect millions of people around the world. Genetic association studies pinpointed ∼150 genetic diseases in which abnormalities in a gene encoding a protein involve in regulation of mitochondria (Calvo et al., 2016). For example, abnormalities in mitochondrial functions have been attributed to be either a root cause or a major driving mechanism involved in ALS disease (Jaiswal, 2013, 2014). In ALS, superoxide dismutase 1 (SOD1) has been proven to be a key gene involved in the disease pathogenesis (Jaiswal, 2012). SOD1 has been found to be mutated in at least in ∼20% familial ALS (fALS) cases. Most importantly, studies over the years have identified mitochondrial dysfunction and associated oxidative stress, reactive oxygen species (ROS) formation and calcium dysregulation to be the critical effecter mechanism through which mutated SOD1 affects neuronal function and survival (Jaiswal and Keller, 2009; Jaiswal et al., 2009; Grosskreutz et al., 2010). Similarly, in case of PD, scientists have focused on the function of Parkin and PTEN-induced putative kinase 1 (PINK1), two proteins mutated in familial, early-onset Parkinson's disease and recommended drugs for therapies that boost mitophagy or stimulate activity of Parkin/PINK1 (Narendra and Youle, 2011). Recently it was shown that in autosomal dominant HD where mhuntingtin (mHTT) mutated protein localizes to the outer mitochondrial membrane, where it wields harmful effects on mitochondria by diminishing mitochondrial motility, alters mitochondrial morphology, fusion and fission, causes calcium dysregulation, reduces oxidative phosphorylation, and depolarizes the mitochondrial membrane potential in HD patients (Reddy et al., 2009). Some of the therapeutic approaches recently tested to target mitochondrial fission protein to which mHTT binds, in mice and cells from humans with HD not only curtailed the toxicity of HTT, but also restored normal motor function in symptomatic HD mice (Song et al., 2011; Di Pardo et al., 2012). Douglas Wallace discovered that mutation in one of the mitochondrial tRNA genes precedes the pathological change, which leads to mitochondrial dysfunction, are the hallmarks of AD (Coskun et al., 2010). So, in general, dysfunctional mitochondria are the hallmarks in the major neurodegenerative diseases.

### MITOCHONDRIAL DYNAMICS, TURN OVER AND ITS ROLE IN NEURODEGENERATIVE DISEASES

Mitochondria have turn over with a half-life of ∼25 days in neurons (Menzies and Gold, 1971). Neurons, like other cells critically depend upon mitochondrial functions for longdistance transport of mitochondria to the synapse, isolation and removal of faulty mitochondria from synaptic sites and metabolic demands that require high bioenergetic outputs and often associated with enhanced production of ROS. Continuous buildup of ROS leads to oxidative damage and impaired protein homeostasis within mitochondrial micro-domains (Lin and Beal, 2006; Jaiswal, 2013, 2014). Therefore, mitochondria have internal "mitochondrial quality control (MQC)" machinery to maintain a vigorously functional, and healthy mitochondrial number and mtDNA mutations by versatile proteolytic enzymes and exchange of components during fusion/fission to maintain normal function (Rugarli and Langer, 2012). Asymmetric fission mechanism segregate damaged components into one daughter mitochondrion, which is removed by autophagy (Twig et al., 2008). Moreover, during fusion/fission mitochondria form a tubular network via a dynamic process and dictated by the fine balance between mitochondrial fusion and fission. Altered balance of mitochondrial dynamics leads to pathogenesis of complex neurodegenerative disorders such ALS, AD, PD, and HD. For details see Burté et al. (2015). Selective mitochondrial mitophagy removes damaged mitochondria and closely associated to biogenesis, which permits replacement of mitochondria and assembly of multiple mitochondrial proteins (Kim and Lemasters, 2011). Thus, maintenance of protein homeostasis through MQC is one of the key aspects in preservation of functional integrity of neuronal mitochondria (Bohovych et al., 2016). MQC involved in removal of impaired misfolded proteins and regulation of assembly, maturation and cleavage of proteins.

### MITOCHONDRIAL RESUSCITATION/REPLACEMENT IN NEURODEGENERATIVE DISORDERS

Although many new candidate drugs or treatment strategies are well on the way, unfortunately the fact remains that there is still not a single broadly applicable effective cure available for ALS, MS, PD, HD, and AD. Moreover, those suffering from familial neurodegenerative diseases and carries the harmful mutations in their nuclear and mitochondrial genetic materials face significant risk of passing over these fatal diseases to their children. Thus, many individuals with genetic predisposition and mutated mtDNA combined with other contributing factors are likely to be affected ruining their life without any hope for cure. Therefore, there are eminent needs to find cure for such neurodegenerative diseases. While approaches such as gene therapy or other techniques will be needed for correcting the harmful errors in nuclear genome, recent advancements now provides the opportunity to correct the errors in the mitochondrial genetic material applying different versions of recently developed MRT procedures. We can think of two different modes of applying the related therapeutic approaches, (a) preventive and (b) curative (**Figure 1**).

In case of different neurodegenerative diseases with known mutation in mtDNA, corrective measures such as mitochondrial replacement by in vitro fertilization can serve as preventive cure.

As a preventive measure, MRT can be applied to replace the dysfunctional mitochondria in the germ line cells thereby avoiding harmful presence of the faulty mitochondrial genes in the children. This can be achieved by harvesting and transferring the nucleus of a woman carrying mtDNA mutations to the eggs of a woman donor with healthy mtDNA to be fertilized. This allows the "original" mother to avoid transmitting faulty mitochondrial genes to the fetus, which inherit all mtDNA from the egg. An alternative method involves fertilizing eggs from both women in-vitro. Then, physicians remove and transfer chromosomes

known as polar bodies (PB)s which eventually disintegrates. The first PB (PB1) contains a diploid set of chromosome whereby the second polar body (PB2) contains a haploid set. In PBT techniques, transfer of PB1 and PB2 into appropriate oocyte or zygotic cytoplasm supports the normal completion of meiosis giving rise to the viable offspring after completion of its full term development. Advance genomics tools such as Transcription Activators like effector nucleases (TALENs), and CRISPR/Cas edited iPS cells with corrected mutations in mitochondrial DNA possibly can overcome some of technical challenges MRT facing right now. MRT can be achieved by germ cell (preventive) and stem cell (curative) modifications. (C) Major challenges of MRT are: (a) heteroplasmy and reversal, (b) incompatibility and alloreactivity, (c) inapplicability of some techniques in human cells, and (d) accumulated mutations in mitochondria.

from the nucleus of one egg to the egg from the woman with healthy mitochondria. This technique is dubbed by the media as "three-parent in-vitro-fertilization."

Currently, three established techniques for mitochondrial replacement exist for germ line therapies to circumvent mtDNA based disease transmission: (1) pronuclear transfer (PNT, McGrath and Solter, 1983; Craven et al., 2010), (2) spindle transfer (ST, Tachibana et al., 2009; Cohen and Alikani, 2013; Tachibana et al., 2013), and (3) polar body transfer (PBT; Wang et al., 2014; Wolf et al., 2015). In PNT, pronuclei are removed from single-cell stage embryo at day one and transferred to a second embryo created using a donor egg with healthy mitochondria. In ST technique, first spindle from donor and mtDNA-carrying unfertilized oocytes are extracted into karyoplasts and thereafter karyoplasts from patient oocytes are fused to donor cytoplasts and fertilized with the partner's sperm to create mtDNA-mutation free embryos. Whereas in PBT techniques, first or second polar bodies are transfer into appropriate oocyte or zygotic cytoplasm supports the normal completion of meiosis giving rise to the viable offspring after completion of its full term development.

These techniques have distinct advantages and disadvantages. In the recent years, these techniques have matured and have been tested in various model organisms with some degree of success. Recent PNT results between zygotes and nuclear spindles among oocytes suggest the possibility of its use in humans (Tachibana et al., 2009; Paull et al., 2013). Apart from the abovementioned preventive application, mitochondrial replacement can also serve as a curative strategy. For this, MRT can be applied to manipulate harvested stem cells collected from patients and thereafter doing the homologous stem cell transplantation after correcting the mitochondrial defects. It is important to mention that recently scientist successfully demonstrated that genetically corrected induced pluripotent stem cells (iPSCs) can be created from fibroblasts from patients with mitochondrial mutations (Ma et al., 2015). A similar approach has been tested earlier on macaques produced normal offspring (Tachibana et al., 2009). These discoveries indicate the feasibility of such techniques for future therapeutic approaches.

## TECHNICAL CHALLENGES, CONCERNS AND RECENT ADVANCEMENTS

While recent developments in the field of MRT raising high hope for cure, there remain major technological challenges and concerns as outlined below.

### Heteroplasmy and Reversal

Mitochondria constantly undergo fusion and fission (Karbowski and Youle, 2003; Youle and van der Bliek, 2012). Moreover, during mitochondrial transfer chances of those residual damaged mitochondria remains in the cells can compromise the effectiveness and long-term benefit of the MRT. As a result, a prerequisite for ensuring successful and effective replacement of damaged mitochondria is avoiding or critically minimizing the carryover of damaged mitochondria. Most effective replacements of damage mitochondria are thus needed to circumvent this problem and undesired effect. Prior work already showed that different techniques has different efficacy in minimizing such carryover. While in case of techniques such as ST, first polar body transfer (PBT1) or PNT, chances of carry over and heteroplasmy remains high, in case of second polar body transfer (PBT2) such carry over remains below detection level and ∼1% (Yamada et al., 2016). Thus, in future PBT2 or other effective novel techniques might serve as a technique of choice for MRT. A recently published article reported improved PNT technique which solve the problem of mitochondrial carry over and reduce the number of defective mitochondria transferred along the nuclear DNA to the donor cell (Hyslop et al., 2016).

### Incompatibility and Alloreactivity

Since the conception of mitochondrial replacement as a strategy, a major concern has been the possible incompatibilities of donor's mitochondria with the nuclear from unrelated individuals. However, there are no strong evidence available strongly supporting this concern especially in case of human. Moreover, current studies indicate persistent compatibility between nuclear genome and donor's mitochondria (Mitalipov and Wolf, 2014). Apart from nuclear-mitochondrial compatibility, alloreactivity also emerges as serious concern if adequate attention not paid to avoid it. This aspect is especially relevant for future curative applications to treat patients with modified stem cells after mitochondrial replacement. However, established approaches such as immune suppression might be able to provide us viable solution to circumvent such undesired side effects if needed. Moreover, inexpensive genetic testing and matched donors also can provide us ways to prevent such mismatches as mean to eliminate or reduce chances for such undesired immune response.

### Inapplicability of Certain Techniques in Human Cells

A major concern that effectiveness of many of these associated procedure has been mainly shown successfully in animal studies and applicability of such procedure still remained to validated in human. While future studies will be needed to achieve such validation to confirm the applicability of such techniques, recent studies using human cells shows the initial evidences for the effectiveness of MRT in human (Yamada et al., 2016).

### Accumulated Mutations in Mitochondria from Aged Persons

A recent discovery on stem cell mitochondria points that iPS cells clones derived from elderly adults show accumulation of DNA mutations and therefore screening cell lines for mitochondrial mutation is important for clinical use (Kang et al., 2016). Findings came with major conclusions that iPS cells derived from older patients tend to show DNA mutations which can impact metabolic function in iPS cells. However, recent improvement of PNT which solve the problem of mitochondrial carry over reduce the risk of mutant mtDNA (Hyslop et al., 2016). It is also important to mention that currently MRT experts think that physiological behavior of mitochondria is very different in embryonic stem cells compared to normal human development and mutations of iPS mitochondria might not be true for human and therefore procedure is safe.

### CONCLUSION AND FUTURE DIRECTIONS

In spite of all of the concerns, it seems MRT has considerable potential for preventing and possibly curing certain forms of neurodegenerative diseases caused due to damaged or dysfunctional mitochondria. Bench-to-bedside implementation of above-mentioned approaches will be based on information from a variety of scientific data and clinical trials of safety and efficacy. Mitochondrial replacement combined with stem cells might provide us future cures at least to a subset of patients where structural and functional mitochondria-derived abnormalities dictate disease pathology and progression. One key factor for success of this technology will be the cost effectiveness and viability in clinical set ups. Technology is still maturing and there are several concerns that need to be addressed regarding the possible differences between the animal models, preclinical studies and human clinical conditions. While this approach can serve as a good alternative approach, it might have limited scope considering the disease heterogeneity and multiple mode of the

### REFERENCES


disease pathogenesis. Nonetheless, even if the techniques can be successfully implemented in few of the well-equipped centers and clinical facilities, that still be extremely beneficial to patients who are currently suffering without hope for a cure. Only longterm studies can tell us for certain if the technique is mature enough to be applied at broad scale in clinics. We also hope there will be a consensus built across our society and political arena on the use of MRT. Furthermore, we hope that the recent and future developments of MRT research will further improve the efficacy and safety of the related techniques and will provide the therapeutic benefit to the affected individuals and to avoid the transmission of such life threatening diseases in their future generations.

### AUTHOR CONTRIBUTIONS

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

### ACKNOWLEDGMENTS

We would like to thank A. Morgan and A. Chemiakine for comments on the manuscript.

gained from genetically modified animals and cell culture models," in Motor Neuron Diseases: Causes, Classification and Treatments, 1st Edn. eds B. J. Turner and J. B. Atkin (New York, NY: Nova Publishers), 87–114.


**Disclaimer:** The views expressed in this article are those of the authors and do not reflect the official policy or position of the University of Rochester Medical Center, or NYSPI/Department of Psychiatry, Columbia University Medical center.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Paine and Jaiswal. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.