**Ca2+ AND Ca2+-INTERLOCKED MEMBRANE GUANYLATE CYCLASE MODULATION OF NEURONAL AND CARDIOVASCULAR SIGNAL TRANSDUCTION**

**Topic Editors Rameshwar K. Sharma, Wolfgang Baehr, Clint L. Makino and Teresa Duda**

#### *FRONTIERS COPYRIGHT STATEMENT*

© Copyright 2007-2015 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** 978-2-88919-506-0 **DOI** 10.3389/978-2-88919-506-0

## *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

# **Ca2+ AND Ca2+-INTERLOCKED MEMBRANE GUANYLATE CYCLASE MODULATION OF NEURONAL AND CARDIOVASCULAR SIGNAL TRANSDUCTION**

Topic Editors:

**Rameshwar K. Sharma,** Salus University, USA **Wolfgang Baehr,** University of Utah, USA **Clint L. Makino,** Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA **Teresa Duda,** Salus University, USA

Image property of the Sharma lab

The tale of cyclic GMP has been astonishing. Having overcome an initial disbelief, cyclic GMP has risen to its present eminence as a premium cellular signal transduction messenger of not only hormonal extracellular but also of the intracellular signals. This research topic focuses on the pathways and functions of membrane guanylate cyclases in different tissues of the body and their interplay with intracellular sensory signals where in many cases, cyclic GMP along with Ca2+ have taken on roles as synarchic co-messengers.


**Table 1. Nomenclature: Membrane Guanylate cyclases;** Extracellular Ligands; Ca2+ Binding Protein Subunit/s

Hippocalcin and GCIP are not specifically covered in this Topic. Future research is needed to revise this Table and eliminate any errors of omission. ANP, atrial natriuretic peptide (also known as ANF, atrial natriuretic factor); ANF-RGC, atrial natriuretic factor receptor guanylate cyclase; BNP, B-type natriuretic peptide; CNP, C-type natriuretic peptide; CNP-RGC, CNP receptor guanylate cyclase; GCAP, guanylate cyclase activating protein; ONE-GC, olfactory neuroepithelium guanylate cyclase; ROS-GC, rod outer segment guanylate cyclase; GCIP, guanylate cyclase inhibitory protein.

# Table of Contents


Gleb Barmashenko, Jens Buttgereit, Neil Herring, Michael Bader, Cemil Özcelik,

Denise Manahan-Vaughan and Karl H. Braunewell

*140 Dysfunction of Outer Segment Guanylate Cyclase Caused by Retinal Disease Related Mutations*

Patrick Zägel and Karl-Wilhelm Koch


Emmanuel S. Buys, Lincoln R. Potter, Louis R. Pasquale and Bruce R. Ksander

*179 Guanylate Cyclase-C/cGMP: An Emerging Pathway in the Regulation of Visceral Pain*

Gerhard Hannig, Boris Tchernychev, Caroline B. Kurtz, Alexander P. Bryant, Mark G. Currie and Inmaculada Silos-Santiago

# Ca2<sup>+</sup> and Ca2+-interlocked membrane guanylate cyclase signal modulation of neuronal and cardiovascular signal transduction

Rameshwar K. Sharma<sup>1</sup> \*, Wolfgang Baehr <sup>2</sup> , Clint L. Makino<sup>3</sup> and Teresa Duda<sup>1</sup>

*<sup>1</sup> Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, Elkins Park, PA, USA, <sup>2</sup> School of Medicine, Department of Ophthalmology and Visual Sciences, University of Utah, Salt Lake City, UT, USA, <sup>3</sup> Massachusetts Eye and Ear Infirmary and Harvard Medical School, Boston, MA, USA*

Keywords: cyclic GMP, Ca2+, membrane guanylate cyclase, multiple transduction modes, signal transduction

The papers in this research topic are focused on the extraordinary tale of cyclic GMP: its recognition as a hormonal second messenger and its subsequent interplay with intracellular calcium where the two together become co-messengers in cellular signal transduction processes. In this emerging theme membrane guanylate cyclases sense intracellular as well as extracellular signals. Four original research and 12 review articles sample the extraordinary progress that has been made in the field.

The topic introduces the present status of the field with a figure illustrating that cyclic GMP is generated by two structurally different guanylate cyclases, soluble and the membrane form (Figure 1: Sharma and Duda, 2014). The synthetic machinery and the modes of their operation by which they generate cyclic GMP are entirely different, and so are their physiologically regulated processes. Since many of the membrane guanylate cyclases carry multiple names and all names remain in use, the review also provides a guide to the current nomenclature of the membrane guanylate cyclases (Introductory text Table 1). Then it moves on with a lively historical perspective, replete with unexpected twists and turns. Current research suggests that new developments will continue to follow a tortuous but always exciting path! Guanylate cyclase activating proteins or GCAPs are neuronal calcium sensing proteins that serve as subunits in the ROS-GC complex. The review article by Lim et al. (2014) delves into the molecular mechanism by which GCAPs inhibit GC activity when [Ca2+]<sup>i</sup> is high and stimulate it when [Ca2+]<sup>i</sup> is low. As it turns out, GCAP1 has a more subtle use for its myristoyl tail, compared to other, related neuronal calcium sensors. Besides the additional forms of GCAPs that are being discovered (cf. Wen et al., 2014), ROS-GC1 is subject to modulation by other Ca2<sup>+</sup> binding proteins: S100B in retinal cones (reviewed by Sharma et al., 2014) and S100B and neurocalcin δ (NCδ) in spermatozoa (Jankowska et al., 2014). Unlike GCAPs, S100B and NCδ stimulate ROS-GC1 at high [Ca2+]<sup>i</sup> . Co-expression of S100B and GCAPs with ROS-GC1 in the same cells empowers ROS-GC1 with the ability to operate as a novel, bimodal Ca2<sup>+</sup> switch wherein guanylate cyclase activity is elevated at very high and at very low [Ca2+]<sup>i</sup> . Transduction of Ca2<sup>+</sup> signals is not restricted to ROS-GCs; NCδ serves as a subunit for the transduction of the ANF signal by ANF-RGC (Duda et al., 2014). Thus, NCδ, in addition to being a "neuronal" calcium sensor, takes on responsibilities outside of neurons. If other guanylate cyclases possess the proper sequences for NCδ, S100B and/or GCAPs interactions, their calcium sensing paradigms will unfold in the future.

Interestingly, GCAP2 has an alternate binding partner that is not a guanylate cyclase. At the photoreceptor synapse, it interacts with RIBEYE to control ribbon size (reviewed by Schmitz, 2014). Recent work on a new ROS-GC binding partner, RD3, (reviewed by Molday et al., 2014) reveals that it is necessary for the intracellular transport of ROS-GC. It also inhibits ROS-GC possibly as a means of suppressing unwanted activity until the cyclase arrives at the proper cellular location.

## Edited and reviewed by:

*Robert J. Harvey, UCL School of Pharmacy, UK*

> \*Correspondence: *Rameshwar K. Sharma, rsharma@salus.edu*

Received: *02 December 2014* Accepted: *16 February 2015* Published: *06 March 2015*

#### Citation:

*Sharma RK, Baehr W, Makino CL and Duda T (2015) Ca*2<sup>+</sup> *and Ca*2+*-interlocked membrane guanylate cyclase signal modulation of neuronal and cardiovascular signal transduction. Front. Mol. Neurosci. 8:7. doi: 10.3389/fnmol.2015.00007*

In retinal rods and cones, cyclic GMP produced by ROS-GCs opens cyclic nucleotide gated cation channels in readiness for visual transduction. Wen et al. (2014) describe the rationale for expressing multiple forms of guanylate cyclases and GCAPs in order to adjust photon response amplitude and quicken photoresponse kinetics according to the requirements of the individual photoreceptor. Besides photoreceptors, there are cyclic GMP pathways in place elsewhere in the retina. Dhingra et al. (2014) have begun to probe the function of these pathways in PDE9A knockout mice by ERG recording.

Downstream signaling pathways for natriuretic peptide receptor GCs are more complex and have not yet been so well characterized (reviewed by Pandey, 2014). In different systems, cyclic GMP synthesis by natriuretic peptide receptor guanylate cyclases decreases cyclic AMP, Ca2+, and inositol triphosphate, and downregulates PKC (protein kinase C) and mitogen-activated protein kinases. Barmashenko et al. (2014) present intriguing experiments in rats with reduced expression of C-type natriuretic peptide receptor guanylate cyclase, CNP-RGC. In neuronal recordings from hippocampus, long term potentiation was enhanced while long term depression was reduced. The changes in neuronal excitability were accompanied by increased exploratory behavior and improved object recognition.

Failure to synthesize cyclic GMP properly can cause severe forms of blindness in early childhood. Disruptions in RD3 cause Leber congenital amaurosis (Molday et al., 2014). In a research article, Zägel and Koch (2014) explain how three separate point mutations that cause another form of Leber congenital amaurosis, progressive cone degeneration, and juvenile retinitis pigmentosa, respectively, alter ROS-GC1 biochemistry. Boye (2014) summarizes promising results of experiments in which gene replacement therapy was used to correct faulty ROS-GC1 function in animal models with recessive retinal disease. Prospects for clinical trials are discussed. Dominant forms of retinal degeneration call for a different approach. The application of RNA interference to attenuate faulty ROS-GC activity caused by mutations in GCAP1 suggests that it may be a viable option (reviewed by Jiang et al., 2014). Glaucoma is yet another major cause of blindness, in this case, due to compromise of retinal ganglion cell function. Buys et al. (2014) examine studies on a major risk factor, elevated intraocular pressure, and the link to cyclic GMP synthesis by both soluble and membrane guanylate cyclases. We conclude the topic with a review by Hannig et al. (2014), that outlines the involvement of STa-RGC in the sensation of visceral pain and how a synthetic peptide may provide relief to patients suffering from abdominal pain.

## 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.

Copyright © 2015 Sharma, Baehr, Makino and Duda. 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.

## Membrane guanylate cyclase, a multimodal transduction machine: history, present, and future directions

## *Rameshwar K. Sharma\* and Teresa Duda\**

*Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, Elkins Park, PA, USA*

#### *Edited by:*

*Clint Lawrence Makino, Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA*

#### *Reviewed by:*

*Kailash N. Pandey, Tulane University Health Sciences Center, USA Alexander Scholten, University of Oldenburg, Germany*

#### *\*Correspondence:*

*Distinguished Professor Rameshwar K. Sharma, Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, 8360 Old York Road, Elkins Park, PA 19027, USA e-mail: rsharma@salus.edu; Teresa Duda, Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, 8360 Old York Road, Elkins Park, PA 19027, USA e-mail: tduda@salus.edu*

A sequel to these authors' earlier comprehensive reviews which covered the field of mammalian membrane guanylate cyclase (MGC) from its origin to the year 2010, this article contains 13 sections. The first is historical and covers MGC from the year 1963–1987, summarizing its colorful developmental stages from its passionate pursuit to its consolidation. The second deals with the establishment of its biochemical identity. MGC becomes the transducer of a hormonal signal and founder of the peptide hormone receptor family, and creates the notion that hormone signal transduction is its sole physiological function. The third defines its expansion. The discovery of ROS-GC subfamily is made and it links ROS-GC with the physiology of phototransduction. Sections ROS-GC, a Ca2+-Modulated Two Component Transduction System to Migration Patterns and Translations of the GCAP Signals Into Production of Cyclic GMP are Different cover its biochemistry and physiology. The noteworthy events are that augmented by GCAPs, ROS-GC proves to be a transducer of the free Ca2<sup>+</sup> signals generated within neurons; ROS-GC becomes a two-component transduction system and establishes itself as a source of cyclic GMP, the second messenger of phototransduction. Section ROS-GC1 Gene Linked Retinal Dystrophies demonstrates how this knowledge begins to be translated into the diagnosis and providing the molecular definition of retinal dystrophies. Section Controlled By Low and High Levels of [Ca2+]i, ROS-GC1 is a Bimodal Transduction Switch discusses a striking property of ROS-GC where it becomes a "[Ca2+]i bimodal switch" and transcends its signaling role in other neural processes. In this course, discovery of the first CD-GCAP (Ca2+-dependent guanylate cyclase activator), the S100B protein, is made. It extends the role of the ROS-GC transduction system beyond the phototransduction to the signaling processes in the synapse region between photoreceptor and cone ON-bipolar cells; in section Ca2+-Modulated Neurocalcin δ ROS-GC1 Transduction System Exists in the Inner Plexiform Layer (IPL) of the Retinal Neurons, discovery of another CD-GCAP, NCδ, is made and its linkage with signaling of the inner plexiform layer neurons is established. Section ROS-GC Linkage With Other Than Vision-Linked Neurons discusses linkage of the ROS-GC transduction system with other sensory transduction processes: Pineal gland, Olfaction and Gustation. In the next, section Evolution of a General Ca2+-Interlocked ROS-GC Signal Transduction Concept in Sensory and Sensory-Linked Neurons, a theoretical concept is proposed where "Ca2+-interlocked ROS-GC signal transduction" machinery becomes a common signaling component of the sensory and sensory-linked neurons. Closure to the review is brought by the conclusion and future directions.

**Keywords: membrane guanylate cyclase, cyclic GMP, multiple transduction modes**

#### **INTRODUCTION**

In addition to cyclic AMP and inositol triphosphate (IP3), cyclic GMP is an omnipresent intracellular second messenger of prokaryotes and eukaryotes. It plays a critical role in the control of physiological processes of cardiac vasculature, smooth muscle relaxation, blood pressure, blood volume, cellular growth, sensory transduction, neural plasticity, learning, and memory. Unique to itself, this second messenger system is generated by the catalysis of GTP by two structurally different guanylate cyclases, soluble and the membrane form (**Figure 1**). The synthetic machinery and the modes of their operation by which they generate cyclic GMP are entirely different, and so are their physiologically regulated processes. This review is focused on the mammalian membrane guanylate cyclase transduction field, a principal area of research by the authors for several decades. It begins with its historical development. Briefly

narrates the efforts involved building on small projects, brick by brick, and emerged from the chasm of disbelief, through steady, continuous work. It then moves on to its present status. To make the presentation simple and chronologically continuous, the subject matters of the earlier reviews (Sharma et al., 1997; Pugh et al., 1997; Sharma, 2002; Duda et al., 2005b; Sharma, 2010; Sharma and Duda, 2012) and publication of these authors have been freely borrowed with appropriate citations.

### **HISTORY: YEAR 1963–1987**

#### **BACKGROUND**

The present day usage of the term "signal transduction" is rooted in the original discovery of cyclic AMP (Sutherland and Rall, 1958; reviewed in: Sutherland and Rall, 1960) for which Earl Wilbur Sutherland, Jr. won a Nobel Prize in physiology/medicine in 1971. Based on the studies with glycogen metabolism in skeletal muscle, Sutherland's group proposed the "second messenger" role of cyclic AMP in the hormone actions of epinephrine and glucagon (reviewed in Sharma et al., 1997). This "second messenger" concept was revolutionary because the consensus at the time was that the hormones acted directly on their target tissues, and, thus the hormonal pathway could only be studied in intact cells. Sutherland's studies demonstrated that it was possible to study the cell-free systems and monitor the generation of cyclic AMP indicative of the parent hormone action. Conceptually, the hormone action occurred in the particulate fraction of the cell in two stages; first, the hormone interacted with the particulate fraction of a cell and second, generated its second messenger, cyclic AMP.

In this concept, the hormone was the first messenger whose interaction with its receptor resulted in a stimulus that caused the production of the second messenger "cyclic AMP," which then acted as the hormone messenger in eliciting the intracellular response (reviewed in Sharma et al., 1997). Thus, the term "second messenger" was coined. Reflective of the time, the biochemical nature of the receptor was not known, because none had been purified. It was not clear whether the enzyme, termed adenylate cyclase, catalyzing the formation of cyclic AMP, was a separate molecule or part, as a subunit, of the receptor. However, it was envisioned that interaction between the receptor and adenylate cyclase generates the second messenger. So enticing was the concept that it was viewed to be applicable to all hormonal systems, i.e., that cyclic AMP is the universal second messenger of all hormones (Robison et al., 1967). Hence, the general concept evolved in which hormone, the first messenger, transmits the extracellular (biological) information to the interior of the cellular machinery through the adenylate cyclase system, and the concept supposed that the receptor was a part of the adenylate cyclase system. The receptor was conceptualized, as defined by the studies in 1930s, as a part of the cell where hormone, the stimulus, interacts (Clark, 1937; Ariëns and Simonis, 1966). The term "stimulus" was equivalent to the presently used term "signal," which, then, according to the "second messenger" concept meant that adenylate cyclase was the means to convert the extracellular ligand binding signal, into the production of the intracellular messenger cyclic AMP. Implicit in this concept was the understanding that the events leading to the transformation of the hormonal binding signal into the production of the second messenger occurred in the cell's plasma membrane.

The "cyclic AMP second messenger concept" was brought to the molecular level by the discoveries of Martin Rodbell and Alfred G. Gilman for which in 1994 they shared the Nobel Prize in physiology/medicine. They demonstrated that G-protein signaling bridged the gap between the hormonal signal and activation of the adenylate cyclase. Rodbell's group argued that the cellular system that processes the hormonal signal was made up of three separate molecular entities: (1) Discriminator (RECEPTOR); (2) Transducer (G-Protein); (3) Amplifier, which as we understand now is the adenylate cyclase. The surface receptor receives information from the exterior of the cell; the G-protein (transducer) transmigrates the information across the cell membrane; and the amplifier (adenylate cyclase) strengthens it to initiate reactions within the cell or to communicate with other cells. In formulating this "G-Protein hormonal hypothesis" the group demonstrated that GTP was obligatory for the hormonal action of glucagon in liver membranes. The term TRANSDUCTION was introduced to indicate this GTP-dependent transformation step. And usage of the term SIGNAL TRANSDUCTION entered the field of cellular signaling (these early studies are reviewed in Rodbell, 1978). The Gilman's and then Birnbaumer's group defined the sequence of these signal transduction steps (Hepler and Gilman, 1992; Birnbaumer and Birnbaumer, 1995). Two types of G proteins, one stimulatory, termed Gs, and the other inhibitory, Gi, were purified and characterized; Gs mediated the stimulatory and Gi the inhibitory transduction. The nature of the receptor defined the specificity of the hormone. The cyclic AMP signaling system therefore comprised of three distinct proteins—receptor, G-protein, and adenylate cyclase (**Figure 2**).

Following the three-component signaling template of cyclic AMP, phosphatidylinositol (IP3) cellular signaling pathway evolved. Here, the surface receptor responded to an extracellular hormonal signal and converted phosphatidyl-4,5-bisphosphate (PIP2) in two separate second messengers, diacylglycerol and inositol trisphosphate (IP3). IP3 signal transduction splits into two: (1) mobilizes the intracellular calcium and (2) results in the activation of protein kinase C (Early studies respectively reviewed in Nishizuka, 1988; Berridge and Irvine, 1989).

In this manner, these, and complimentary studies from other laboratories (reviewed in Strader et al., 1994) established the presence of two major G-proteins linked with threecomponent cellular signal transduction pathways—cyclic AMP and phosphatidylinositol—providing the means by which the extracellular hormonal signal is translated into the production of the intracellular second messengers. These second messengers, in turn, initiate a cascade of chemical reactions which ultimately translate the hormonal signal into a biological response.

## **CYCLIC GMP SIGNALING PATHWAY** *Pursuit*

Besides cyclic AMP, could another cyclic nucleotide be an alternative hormonal second messenger in cellular signaling?

Cyclic GMP was first identified in rat urine (Ashman et al., 1963). Six years later, cyclic GMP and the enzyme membrane guanylate cyclase catalyzing its synthesis from GTP were detected in all tested animal tissues (Goldberg et al., 1969, 1973; Ishikawa et al., 1969). A groundbreaking "Ying Yang" hypothesis was proposed where both cyclic AMP and cyclic GMP served as hormonal second messengers. They were intertwined and exhibited opposing biological activities of the cellular regulation (Goldberg et al., 1975).

#### *Denial*

The celebration was short lived, however. The original proponents failed to consistently demonstrate a hormonally dependent guanylate cyclase in their tested systems (Goldberg and Haddox, 1977). They observed that a variety of non-hormonal ligands like polyunsaturated fatty acids, peroxides, hydroperoxides, free radicals, ascorbic acid, sodium nitroprusside, cigarette smoke also stimulated membrane guanylate cyclase activity. Thus, the membrane guanylate cyclase was a non-specific enzyme and underwent regulation *via* the oxidation-reduction potential of the biochemical reactions (Murad et al., 1979). Also, at the time, the only known cyclic GMP-dependent cellular component was cyclic GMP-dependent protein kinase but not one of its specific, distinct from that of the cyclic AMP-dependent protein kinase, substrate was known. Hence, the revised consensus was that the cyclic GMP-dependent system lacked specificity and acted as a sub-servant to the cyclic AMP system. Therefore, no distinct hormonally dependent membrane guanylate cyclase or cyclic GMP-related specific signaling system existed in any of the biological systems and cyclic GMP had no second messenger role in cellular signaling (Gill and McCune, 1979).

**signaling systems.** The cyclic AMP (and IP3) signaling system consists of three separate components: hormone receptor, transducer (G protein) and amplifier (adenylate cyclase). The cyclic GMP signaling system consists of a single protein. The hormonal signal is recognized by the

ATP-modulated ARM domain located next to the transmembrane domain in the intracellular portion of the protein (transducer) and the signal is amplified by the cyclase catalytic domain located at the C-terminus of the protein.

With a new twist, this view was strengthened as the presence of a soluble form of guanylate cyclase activity was found in the originally tested systems (Murad et al., 1979). This form was stimulated by polyunsaturated fatty acids, peroxides, hydroperoxides, free radicals and other agents. Impacted by these new findings and a passion to pursue this new field of soluble guanylate cyclase, the role of cyclic GMP in hormonal signal transduction was totally discarded (Goldberg and Haddox, 1977; Murad et al., 1979). These views are summed up in two of many high-impact reviews:

"While the metabolism of cyclic GMP has many parallelism to that of cyclic AMP, it is becoming quite clear that the two systems have distinct differences and that the functions of one system may not necessarily provide clues to the function of the other. In retrospect, the apparent parallelisms of the two systems hindered the originality of approaches to the cyclic GMP system and may, in fact, have impeded some of the developments in the area. Certainly, many early experiments with guanylate cyclase and cyclic GMP conducted in this laboratory and others were done based on previous experience and biases with the cyclic AMP system . . . , the properties of the soluble and particulate activities in tissues are quite different . . . . Minor differences have also been noted between particulate activities and in plasma and endoplasmic reticulum preparations. In spite of these significant physical and/or kinetic differences that suggest two or more forms of guanylate cyclase may exist in tissues, it is not yet known whether or not the proteins or their subunits are different. The apparent differences may be attributable to assay conditions, other constituents in crude preparations, or possibly other factors" (Murad et al., 1979).

"Considerable effort has been put forth to demonstrate an activation of the enzyme by hormonal or other physiological agents that promote the accumulation of cGMP in intact cells. To date, no specific direct effect of such an agent on guanylate cyclase activity *in vitro* has been convincingly demonstrated. This is a characteristic of the cGMP system that contrasts sharply with its cAMP counterpart; adenylate cyclase from disrupted cells as a general rule can be activated by the agents that stimulate cellular cAMP accumulation . . . .It is becoming evident that the greatest dissimilarity between guanylate cyclase and adenylate cyclase may be that the activity of the former is modulated indirectly with respect to the cellular action of hormones (and other cell stimuli) compared with the direct effect that hormones seem to have on the activity of adenylate cyclase" (Goldberg and Haddox, 1977).

#### *Consolidation, parallel competing concept*

The above intimidating concept, lasting to the mid-1980s, was challenged mainly by the competing concept of our group. It favored the hormonal second messenger role of cyclic GMP (reviewed in Sharma, 2010). Studies with the model systems of the isolated adrenal fasciculata and the adrenocortical carcinoma cells established a membrane guanylate cyclase to be the direct and specific transducer of the adrenocorticotropic hormone (ACTH). Notably, ACTH had no effect on the activity of the soluble guanylate cyclase (reviewed in Sharma, 2002). Thus, the guanylate cyclase existed in two forms, membrane and soluble; and the membrane form was hormone-dependent. These studies are briefly outlined below.

The models of the isolated adrenal fasciculata and the adrenocortical carcinoma cells were tailored to have no phosphodiesterase activity (Kitabchi and Sharma, 1971), and they were used to investigate the second messenger roles of cyclic AMP and cyclic GMP in steroidogenesis. The tailoring created a heightened membrane response to ACTH and it was direct, reflective of membrane-specific guanylate cyclase activity. The tailored PDEdeleted isolated fasciculata cells were sensitive to as little as 0.5–10 μU ACTH in stimulating corticosterone synthesis (Sharma et al., 1974, 1976; Perchellet et al., 1978). These concentrations of ACTH did not raise the level of cyclic AMP, yet they stimulated steroidogenesis (Sharma et al., 1974, 1976; Perchellet et al., 1978). The rise in cyclic AMP level was observed only at ACTH concentrations beyond their maximal steroidogenic capacity (Sharma et al., 1974). Thus, cyclic AMP was not the physiological second messenger of ACTH.

The physiological levels of ACTH in these cells also generated an excellent temporal correlation between the sequential production of cyclic GMP, phosphorylation, and the concomitant synthesis of corticosterone, indicating that the hormone caused the production of cyclic GMP, this, in turn, led to phosphorylation, and finally to corticosterone synthesis (Sharma et al., 1976, 1977). On similar lines, addition of cyclic GMP to these cells stimulated the protein kinase activity which was followed by a rise in their steroidogenic activity (Sharma et al., 1976); and, importantly, this added cyclic GMP mimicked the exogenously supplied ACTH action in stimulating the direct transformation of cholesterol to corticosterone (Sharma et al., 1972). Consistent with the earlier findings these studies also supported the obligatory role of Ca2<sup>+</sup> in the ACTH-induced steroidogenesis (Perchellet and Sharma, 1979). The conclusions were that cyclic GMP and Ca2<sup>+</sup> were the two complementary physiological second messengers of the hormone.

The stage was set to directly demonstrate the ACTH-dependent membrane guanylate cyclase in the adrenal cell membranes and also the presence of its companion cyclic GMP system, protein kinase in the bovine adrenal cortex, to strengthen the cyclic GMP second messenger concept.

The cyclic GMP-dependent protein kinase was purified from the bovine adrenal cortex. Its important distinctive characteristic that set it apart from the cyclic AMP signaling system was that its most effective effector molecule was 8-bromo cyclic GMP (Ahrens et al., 1982). Compared to its dibutyryl analog, it was 10,000-fold more active in stimulating the protein kinase activity. This feature established its unique functional identity because the dibutyryl cyclic AMP was more potent than cyclic AMP in stimulating cyclic AMP-dependent protein kinase. Finally, the molecular structure of the cyclic GMP-dependent protein kinase was entirely different from that of the cyclic AMP-dependent protein kinase. The single subunit of the former protein contained both the non-dissociable regulatory and catalytic domains, opposite was the case with the latter protein (Ahrens et al., 1982). Besides providing clarity of the molecular elements of the cyclic GMP signaling system, these unique characteristics of these proteins helped to resolve an earlier argument that cyclic GMP signaling system was the "subservant" of the cyclic AMP signaling system *(vide supra)*. Clearly, it was not the case.

Analysis of the adrenal cortex and the adrenocortical carcinoma tissues demonstrated the direct presence of the ACTH/Ca2+-dependent membrane guanylate cyclase (Nambi and Sharma, 1981a,b; Nambi et al., 1982). In addition, the distinct features of the membrane and the soluble form of guanylate cyclases were established (Table 1 of Nambi et al., 1982). Most fundamentally, the membrane guanylate cyclase activity was not dependent on the free radical and nitric-oxide-generating agents, including hemin.

These conclusions were supported by the findings of other investigators. ACTH stimulated membrane guanylate cyclase activity in the cultured neurons (Anglard et al., 1985), and ANF peptide hormone stimulated membrane guanylate cyclase activity of several rat tissues (Hamet et al., 1984; Waldman et al., 1984). Ironically, these were the same tissues, in the hands of the same investigators, in which earlier the membrane guanylate cyclase activity was not observed (Goldberg and Haddox, 1977; Murad et al., 1979).

The issue was thus resolved. A hormonally-dependent membrane guanylate cyclase existed in the mammalian cells; it was a transducer of specific hormonal signals and cyclic GMP was their second messenger. In addition, the membrane guanylate cyclase was functionally and kinetically distinct from the soluble form. At this moment, the task at hand was to define the cyclase's molecular nature and to elucidate the mechanism of the hormonal signal transduction.

## **BIOCHEMICAL IDENTITY: ATRIAL NATRIURETIC FACTOR RECEPTOR, THE PROTOTYPE MEMBER OF THE MEMBRANE GUANYLATE CYCLASE FAMILY**

A plasma membrane guanylate cyclase was purified from the rat adrenocortical carcinoma and its absolute purity was established by these criteria: (1) protein staining: SDS-PAGE showed a single Coomassie blue- and silver-stained band; (2) isoelectric focusing, native protein and its iodinated form yielded a symmetrical protein peak, which superimposed over the guanylate cyclase activity peak; (3) Western blot analysis using a monospecific polyclonal antibody raised against the native protein yielded a single immunoreactive band which co-migrated with that of the purified protein; (4) the antibody raised against the native protein blocked almost all of its membrane guanylate cyclase activity; and (5) the protein bound ANF peptide hormone in a stoichiometric fashion (Paul, 1986; Paul et al., 1987).

Thus, the protein was pure, it was a membrane guanylate cyclase, and possessed a remarkable characteristic of being also the ANF hormonal receptor. Accordingly, it was named ANF-RGC (ANF RECEPTOR GUANYLATE CYCLASE) also known as GC-A or NPR-A.

The properties of ANF-RGC also resolved the earlier erroneous conclusions on the nature of the membrane guanylate cyclase (Goldberg and Haddox, 1977; Murad et al., 1979). The membrane guanylate cyclase activity was independent of the nitric-oxidegenerating agents—catalase, hemin, dithiothreitol, arachidonic acid, and tuftsin, including cigarette smoke. These non-specific activities apparently belonged to the soluble form of the guanylate cyclase whose constituent, critical for the functional operation, was heme (**Figure 1**).

A parallel report described the purification of a protein from the rat lung and claimed that "ANF-binding and guanylate cyclase activity resided on a single protein" (Kuno et al., 1986). These authors demanded priority on their finding and challenged the priority on the original discovery of ANF-RGC. This issue was settled in a commentary (Sharma, 1988), which revealed that in contrast to the criteria of the absolute purity established for the adrenocortical carcinoma ANF-RGC guanylate cyclase (Paul et al., 1987), the lung guanylate cyclase preparation was crude (Kuno et al., 1986)—in addition to the main and major 120 kD protein it contained at least 13 other minor bands and the authors themselves stated "that it was about 95% pure" (Kuno et al., 1986). The possibility existed that the 5% contaminant contained the separate ligand-binding or the cyclase activity. There were two other serious problems with the claim that the lung guanylate cyclase was also the ANF receptor: (1) it bound only 14.5% of ANF at the noted theoretical value; (2) the lung enzyme was stimulated by hemin, a characteristic of the soluble guanylate cyclase, not possessed by the membrane guanylate cyclase (Kuno et al., 1986). To this date, these deficiencies in the lung enzyme to categorize it as a genuine ANF-RGC have not been resolved.

With two subsequent reports on ANF-RGC purification from the adrenal cortex and its characterization (Takayanagi et al., 1987; Meloche et al., 1988), the concept that this membrane guanylate cyclase is also a surface receptor of the ANF hormone was established. And a new field of the hormone receptor membrane guanylate cyclases was borne.

This happening had a powerful impact on the cellular signal transduction field, as it was a ground-breaker. It demonstrated that the hormonally modulated membrane guanylate cyclase transduction system was radically different from the two existing cyclic AMP and IP3 signaling pathways, quoting in the original publication "coexistence of the ANF receptor and guanylate cyclase activities on a single polypeptide chain indicates that the mechanism of transmembrane signal transduction involving mediation by second messenger, cyclic GMP, is different from the well-established adenylate cyclase system. In the hormone-dependent adenylate cyclase, there is an assemblage of individual components—receptor, GTP-binding protein, and catalytic moiety—for signal transduction. In contrast, the presence of dual activities—receptor-binding and enzymic—on a single polypeptide chain indicates that this transmembrane protein contains both the information for signal recognition and its translation into a second messenger" (Paul et al., 1987) (**Figure 2**).

In the proposed model, ANF-RGC is depicted as a single transmembrane-spanning protein; the ANF-receptor domain resided on the outside, the catalytic domain protruded inside the cell and a transmembrane segment separated its two portions (Sharma et al., 1988; Sharma, 2002; modified version of Figure 1 in Sharma, 2002). Mechanistically, ANF will bind to its receptor which resided in the extracellular domain of ANF-RGC. It will generate a cascade of structural changes, they will be carried over the transmembrane domain of ANF-RGC and finally stimulate its catalytic domain residing in the interior of the cell. The cyclic GMP will be produced and serve as a second messenger of the hormonal ANF signal.

## **ANF-RGC FAMILY**

Almost 3 years after its purification, ANF-RGC mRNA was cloned from the human and rat brain (Chinkers et al., 1989; Lowe et al., 1989), adrenal gland (Duda et al., 1991) and the mouse Leydig tumor cells (Pandey and Singh, 1990). In all cases the deduced protein structure validated the original theoretical prediction (Paul et al., 1987) that ANF-RGC was a singletransmembrane-spanning protein. The transmembrane region divides it in two roughly equal portions: extracellular and intracellular (please note: a more detailed description of a representative membrane guanylate cyclase modular structure is provided in **Figure 4**).

Analysis of the two additional cloned membrane guanylate cyclases demonstrated that they were the surface receptors of the respective hormones, CNP (Chang et al., 1989; Schulz et al., 1989; Duda et al., 1993) and enterotoxin (de Sauvage et al., 1991; Singh et al., 1991); and they have the identical structural topography as ANF-RGC. According to their function, they were named CNP-RGC (CNP receptor guanylate cyclase also known as GC-B or NPR-B) and STa-RGC (enterotoxin receptor guanylate cyclase also known as GC-C). These findings established a three-member surface receptor family of membrane guanylate cyclases, which remains the case to date (reviewed in Sharma, 2010).

#### **HORMONE SIGNALING SITE RESIDES IN THE EXTRACELLULAR DOMAIN OF RECEPTOR GUANYLATE CYCLASES**

With the model system of the variant form of ANF-RGC, GCα, cloned from the adrenal cortex, the original study of Duda et al. (1991) demonstrated that the ANF binding site in ANF-RGC resides in the extracellular domain. Compared to ANF-GC, GCα in its structure had only two amino acid substitutions, Gln338His and Leu364Pro, but it did not bind ANF. Importantly, it did retain basal cyclase catalytic activity. Reconstitution of the native ANF-RGC residues restored the mutant's ANF-binding and ANF-dependent catalytic activities (Duda et al., 1991). Thus, the ANF ligand-binding domain resided in the extracellular domain of ANF-RGC; and one, or both of the residues— Gln338, Leu364—controlled its ANF-binding and ANF-dependent catalytic activities. Refined point mutation analysis demonstrated that both of these controls resided in Leu<sup>364</sup> (Duda et al., 1991). Thus, the hypothesis for the ANF signal transduction mechanism evolved. It involved the ANF binding to its extracellular domain in ANF-RGC; the binding initiated a cascade of structural changes in ANF-RGC; these changes, in turn, stimulated its catalytic domain, which resided at the opposite end on the intracellular domain of the cyclase. The generated cyclic GMP served as the second messenger of ANF signal.

This signal transduction mechanistic template was applied to CNP-RGC cloned from the human retina (Duda et al., 1993). CNP-RGC is a homologous protein to ANF-RGC, with 59% structural identity. It was found that Glu<sup>332</sup> residue of CNP-RGC was critical for the CNP binding and the CNP-dependent catalytic activity of CNP-RGC (Duda et al., 1994). Like ANF-RGC, the site had no influence on the basal guanylate cyclase activity. These studies were extended to compare the identities of the folding patterns of the external domains of these membrane guanylate cyclases (Duda et al., 1994). They were found to be almost identical (Duda et al., 1994). The counterpart of CNP-RGC-Glu<sup>332</sup> residue is the Gln338 residue of ANF-RGC, and that of ANF-RGC-Leu<sup>364</sup> residue is Val358 of CNP-RGC. Conversion of the ANF-RGC residue Gln338 to Glu resulted in a change from none to significant CNP signal transduction activity. And, the conversion of Val358 to Leu generated significant ANF signal transduction activity in CNP-RGC (Duda et al., 1994). These studies validated that the ligand-binding domain resides in the extracellular domain of ANF-RGC and CNP-RGC, provided identity of the residues that are critical in defining the ligand-binding domains and showed that the folding patterns of these domains in two guanylate cyclases are very similar.

The subsequent crystallization studies on the isolated extracellular domains of ANF-RGC and CNP-RGC validated the conclusions that the hormone-binding site, indeed, resides in the external domain of the guanylate cyclase and Leu364 of ANF-RGC and Val<sup>358</sup> of CNP-RGC constitute critical hydrophobic sites essential for the ANF or CNP signaling but the sites are not the direct hormone-binding sites (Ogawa et al., 2004; He et al., 2001).

#### **ATP IS OBLIGATORY FOR ANF AND CNP HORMONAL SIGNALING**

Prior to the molecular characterization of ANF-RGC, two studies with crude membrane preparations proposed that ATP accelerates ANF-dependent ANF-RGC activity (Kurose et al., 1987; Chang et al., 1990). Two different mechanisms for the acceleration were proposed; direct (Kurose et al., 1987) and indirect *via* an accessory protein (Chang et al., 1990). Because the preparations were crude and the guanylate cyclase was uncharacterized, the issue as to which mechanism was correct could not be resolved.

Subsequently, in parallel two independent groups established that ATP is obligatory for the ANF-dependent ANF-RGC transduction activity (Marala et al., 1991; Chinkers et al., 1991). Neither ANF, nor ATP alone, is able to stimulate ANF-RGC catalytic activity. Only when these two are together they are able to stimulate the cyclase. Because the non-hydrolyzable analogs, ATPγS and AMP-PNP, mimicked the ATP effect, both groups proposed that ATP acts directly by allosteric regulation of ANF-RGC (Marala et al., 1991; Chinkers et al., 1991). Later study with CNP-RGC demonstrated that ATP is also obligatory for its CNP-dependent activity (Duda et al., 1993).

The mechanistic models proposed for the ATP effect by these two groups were different, however. Which one is valid, was debated for over two decades. MODEL 1 proposed by the Garbers'group stated ". . . binding of ANP [ANF] to the extracellular domain of its receptor initiates a conformational change in the protein kinase-like domain (KHD), resulting in derepression of guanylate cyclase activity" (Chinkers and Garbers, 1989). Its central idea was that "KHD in native ANF-RGC suppresses its catalytic module activity; ANF functions by relieving this suppression." Sharma's group challenged this model on the grounds of their findings that partial or complete deletion of the KHD did not cause any significant elevation in the basal activity of the guanylate cyclase; therefore, ANF signaling did not involve overcoming the KHD suppression of the guanylate cyclase activity, contradicting the central theme of MODEL 1. (Goraczniak et al., 1992; reviewed in Duda et al., 2005b). They proposed an alternate MODEL 2. In it the repressor of the core catalytic domain (CCD) of the guanylate cyclase was the disulfide structural motif of ANF-RGC (Duda and Sharma, 2005). This motif was juxtaposed to the N-terminus of the TM (transmembrane of the domain). The motif repressed the ARM (ATP REGULATED MODULE) domain, which, in turn, inhibited the CCD. ATP-modulated ANF signal overcame the disulfide structural motif-dependent inhibition of ARM resulting in activation of the CCD of ANF-RGC.

The MODEL 2 has been defined in molecular terms by simulating structure of the ARM domain in 3D-terms, experimentally validating the biochemistry of the structure and its physiology through the mouse molecular genetic models (Duda et al., 2001c, 2009, 2013). The definition has been arrived at through biochemical experiments involving point mutation, time-resolved tryptophan fluorescence, Forster Resonance Energy Transfer (FRET), reconstitution, recombinant constructs, molecular modeling, immunohistochemistry and recently mouse molecular genetics (these studies are comprehensively reviewed in Duda et al., 2014).

### **LINKAGE WITH SENSORY TRANSDUCTION SIGNALS EXPANDS THE FAMILY**

Because the property of being a hormonal receptor, transducer and signal amplifier was common in all known members of the membrane guanylate cyclase family (ANF-RGC, CNP-RGC and STa-RGC), the notion was that the cyclic GMP signaling pathway operated solely in the hormonal signaling pathways and was designed to transduce only the signals generated outside the cell. However, it was not the case.

The membrane guanylate cyclase signal transduction paradigm changed with the landmark discovery of the photoreceptor ROS-GC guanylate cyclase. Here the signal initiation occurred inside the cell and the signal transduction system was composed of two separate elements, Signal receiver and the Transducer (detailed reviews Sharma and Duda, 2012; Koch et al., 2010; Pugh et al., 1997).

It was known that in phototransduction, which is the transformation of the light signal into an electrical signal, both [Ca2+]i and cyclic GMP are the critical cytosolic regulatory effectors of the photon signal in the vertebrate photoreceptors. Which one of these is the second messenger of the LIGHT signal; how these signals are generated; how they interact with each other; and what was the source of the cyclic GMP was not known, however (early reviews Pugh and Cobbs, 1986; Stryer, 1986; Lamb et al., 1986). Early reports on the successful identification of photoreceptor ROS-GC contradicted each other, and created chaos. They ranged from the guanylate cyclase being a 67 kDa subunit (Horio and Murad, 1991a,b), consisting of "separate regulatory and catalytic subunits" (Stryer, 1991) and of its molecular from of retGC (Shyjan et al., 1992).

The breakthrough came with the seminal observation that a soluble bovine rod outer segment (ROS) fraction stimulated the catalytic activity of a particulate photoreceptor guanylate cyclase in the absence of Ca2<sup>+</sup> (Koch and Stryer, 1988). This hinted that the guanylate cyclase was of the membrane form; the soluble ROS fraction was Ca2<sup>+</sup> sensitive and the two components together constituted the phototransduction-linked guanylate cyclase transduction element. However, the conceived structure of this guanylate cyclase was erroneous, believed to be composed of "separate regulatory and catalytic subunits" (Stryer, 1991).

The true identity of the photoreceptor ROS-guanylate cyclase (ROS-GC also known as GC-E) was resolved by establishing its direct purification from the bovine outer segments (OS) (Margulis et al., 1993), the site of phototransduction; and its protein-sequence-based molecular cloning, structure, and function (Goraczniak et al., 1994). ROS-GC had a theoretical molecular mass of 120,360 Da; a value similar to that reported earlier for a bovine (Koch, 1991) and a toad photoreceptor guanylate cyclase (Hayashi and Yamazaki, 1991). Unlike the other known membrane guanylate cyclases, atrial natriuretic factor receptor guanylate cyclase ANF-RGC (Paul et al., 1987; Sharma, 1988; Chinkers et al., 1989; Duda et al., 1991) and type C natriuretic peptide CNP-RGC guanylate cyclase (Chang et al., 1989; Duda et al., 1993), ROS-GC was not hormonally-responsive (Goraczniak et al., 1994). And also ROS-GC was not composed of "separate regulatory and catalytic subunits"; it was not nitric oxide sensitive nor was its structure identical to retGC (Shyjan et al., 1992), which apparently was a cloning artifact whose primary structure was eventually revised to match that of bovine ROS-GC (Accession number M92432).

Hydropathy analysis of its deduced amino acid sequence revealed that topography of this protein was similar to the other three peptide hormone receptor guanylate cyclase members (Goraczniak et al., 1994), i.e., (1) it was a single transmembrane spanning protein; (2) the TM domain divided the protein into roughly two equal segments, extracellular and intracellular; (3) it showed a significant overall homology with the other peptide hormone receptor guanylate cyclases: 27% identity with STa-RGC, 30% with ANF-RGC, 31% with CNP-RGC. The level of identity rose to 37, 40, and 41% in their intracellular regions, and there were respective identities of 25, 32, and 32% between their kinase like domains. The least identity with peptide hormone receptor guanylate cyclase existed in their extracellular domains: 14% with STa-RGC, 17% with ANF-RGC and 18% with CNP-RGC. Despite these similarities, there were two significant structural variances. The signature ATP-modulated domain Gly-X-X-X-Gly present in ANF-RGC and CNP-RGC was absent in ROS-GC. And the ROS-GC protein beyond the catalytic domain contained a C-terminal extension tail of 90 amino acids, Y965-K1054, this tail was absent in ANF-RGC and CNP-RGC. On these grounds the discovery of ROS-GC represented expansion and the generation of a new subfamily of the membrane guanylate cyclases.

It is noteworthy that in the membrane guanylate cyclase family ROS-GC is the only one whose molecular identity has been established on the basis of its protein sequence. This strategy has provided an experimental proof for the position of the Nterminus amino acid of the mature protein and demonstrated that the mature protein contains a 56 amino acid N-terminus hydrophobic signal peptide. The theoretical molecular mass of the protein with its signal peptide is 120,361 Da and without it is 114,360 Da. (Goraczniak et al., 1994).

## **ROS-GC, A CA2+-MODULATED TWO COMPONENT TRANSDUCTION SYSTEM**

Following the functional description by Koch and Stryer (1988), almost contemporaneously, two independent groups discovered two structurally dissimilar guanylate cyclase activating proteins (GCAPs), GCAP1 (Palczewski et al., 1994; Subbaraya et al., 1994; Gorczyca et al., 1994; Frins et al., 1996) and GCAP2 (Dizhoor et al., 1995). In the presence of free Ca2+, did these GCAPs inhibit catalytic activity of the photoreceptor ROS-GC?

Availability of the recombinant (r)ROS-GC made it possible to answer this question. Expressed in the heterologous system of COS cells, ROS-GC was responsive at 10 nM [Ca2+]i to the GCAP1 stimulation in a dose dependent manner. The stimulation was inhibited cooperatively by free Ca2<sup>+</sup> with a K1*/*<sup>2</sup> of 100 nM. Under identical conditions, GCAP1 had no effect on the catalytic activity of the recombinant peptide hormone ANF-RGC (Duda et al., 1996b). An important characteristic of the transduction system was that GCAP1 remained bound to ROS-GC at the low and high Ca2<sup>+</sup> levels, consistent with physiological observations (Koutalos et al., 1995). Similar reconstitution studies established GCAP2 as another Ca2+-sensor of ROS-GC signaling. With the subsequent discovery of a second ROS-GC, termed ROS-GC2, or GC-F (Lowe et al., 1995) to distinguish it from the original ROS-GC which was renamed as ROS-GC1 (GC-E), it became clear that a Ca2+-modulated system composed of a pair of ROS-GCs and a pair of GCAPs sub-serve phototransduction in the outer segments of photoreceptors. Which of the two ROS-GCs is physiologically linked with phototransduction?

This problem has not yet been solved for ROS-GC2, but by all criteria ROS-GC1 is a vital component of the phototransduction machinery. (1) It is the only one purified directly from the ROS, the site of phototransduction. (2) Direct quantitative estimation of the two isozyme guanylate cyclases in the bovine ROS demonstrates that their ratios are 96% ROS-GC1 and 4% ROS-GC2, suggesting ROS-GC2 presence is minor (Helten et al., 2007). The latest estimation in the mouse shows the same pattern, 76 vs. 24% (Peshenko et al., 2011). Notably, in contrast to the bovine, the mouse study was made with the total OS instead of the isolated membrane fractions. (3) The multiple cone rod dystrophies in humans have only been linked with the ROS-GC1 mutations, none so far with ROS-GC2 (reviewed in Hunt et al., 2010). Finally, (4) it has been demonstrated that the ROS-GC1 gene deletion results sequentially in the cone and then the rod degenerations, indicating that the guanylate cyclase is directly linked with the vitality and the function of photoreceptors (Yang et al., 1999).

## **GCAPS INTERLACE CA2<sup>+</sup> AND ROS-GC**

Together with recoverin (Dizhoor et al., 1991; Hurley et al., 1993), GCAPs are the prototype members of the NCS (neuronal Ca2<sup>+</sup> sensor) proteins family (reviewed in Sharma and Duda, 2012; Koch et al., 2010). Except for S100B, which has two, a general characteristic of this family is that its members have four Ca2+-specific structural motifs, termed EF hands. They define the general Ca2<sup>+</sup> sensor property of the family. While the teleost fish expresses up to 8 forms, the mammals express 2-3 GCAPs (reviewed in: Rätscho et al., 2010).

GCAP1 and GCAP2 are expressed in the same concentration of 3μM in the bovine ROS (Hwang et al., 2003). Their quantitative, relative, and cellular concentrations in the cones are not known. Yet immunocytochemical studies show that both GCAPs are present in the outer and inner segments of the rods and cones (Gorczyca et al., 1994; Frins et al., 1996). In mouse cones, the expression of GCAP1 exceeds significantly the expression of GCAP2 (Wen et al., 2014). Notably, GCAP1 presence in the cone synaptic pedicles can be clearly defined (Venkataraman et al., 2003).

True to the NCS family trait, GCAPs harbor four EF-hand Ca2+-binding motifs, yet only three are functional (reviewed in Koch et al., 2010; Sharma and Duda, 2012). They are second, third and fourth. They sense Ca2<sup>+</sup> in the range of tens to hundreds nanomoles and are the Ca2+-sensor elements of the ROS-GC linked phototransduction machinery.

The knowledge on the [Ca2+]i sensing properties of GCAP1 is more advanced than that of GCAP2. In a recent proposal, the transition from the DARK State to the illuminated state of the photoreceptors occurs by the substitution of bound Ca2<sup>+</sup> with the bound Mg2<sup>+</sup> to ROS-GC1 (Peshenko and Dizhoor, 2006, 2007). Ca2<sup>+</sup> associates with GCAP1 with a rate of <sup>∼</sup> <sup>2</sup> <sup>×</sup> <sup>10</sup><sup>8</sup> <sup>M</sup>−<sup>1</sup> <sup>s</sup> −1 (*kon*) which is close to the diffusion limit (Sokal et al., 1999). The apparent dissociation constants of each EF-hand for Ca2<sup>+</sup> are between 0.08 and 0.9μM and between 0.1 and 1.6μM in the absence and in the presence of 2 mM Mg2+, respectively (Lim et al., 2009). Thus, these nanomolar affinities result in fast dissociation rates (for example, *koff* <sup>=</sup> *kon* <sup>×</sup>*KD* <sup>=</sup> <sup>2</sup> <sup>×</sup>10<sup>8</sup> <sup>M</sup>−<sup>1</sup> <sup>s</sup> <sup>−</sup><sup>1</sup> <sup>×</sup>0*.*<sup>2</sup> <sup>×</sup>10−<sup>6</sup> *<sup>M</sup>* <sup>=</sup> <sup>40</sup>*<sup>s</sup>* <sup>−</sup>1; 1/*koff* <sup>=</sup> 25 ms), consistent with the kinetics of the photoresponse.

GCAP2 senses Ca2<sup>+</sup> signals differently through its second, third, and fourth EF-hands. They display an apparent *KD* of 300 nM (Ames et al., 1999), but no specific assignment for the affinity of each EF-hand has been made so far. It appears, however, that EF hands two and four influence Ca2<sup>+</sup> sensitivity of GCAP2 more than EF-hand three (Dizhoor and Hurley, 1996). Cysteine accessibility study with GCAP2 mutants shows its restricted reactivity toward Cys111 at sub-micromolar Ca2+-concentrations, indicating that within the Ca2<sup>+</sup> concentration range where ROS-GCs are regulated, Ca2+-induces conformational changes in GCAP2 (Helten and Koch, 2007). Thus, these fluctuations in the free Ca2<sup>+</sup> concentrations of the rods and cones control and signal GCAPs to transmit their messages to the bound ROS-GC and be transduced in the generation of cyclic GMP, the second messenger of phototransduction.

#### **DIFFERENTIAL ROS-GC1 MODULATION BY GCAPS**

In a significant conceptual advancement, the original study of Krishnan et al. (1998) proposed and demonstrated that "the intracellular region of ROS-GC1 is composed of multiple modules, each designed to mediate a specific Ca2<sup>+</sup> signaling pathway." This meant that the sensing mechanisms of [Ca2+]i by the ROS-GC1 modules are more than one. This concept was arrived at through the use of ROS-GC1 deletion and ANF-RGC/ROS-GC1 hybrid mutants which retained only the catalytic domain of ROS-GC1. The GCAP1- and GCAP2- modulated domains in ROS-GC1 were separate and they resided on the opposing ends of the catalytic domain of ROS-GC1, GCAP2 at the C-terminal side and the GCAP1 at the N-terminal.

This concept was fine-tuned by mapping the GCAP1- and GCAP2-modulated sites of ROS-GC1 by the use of a comprehensive technology involving soluble constructs of ROS-GC1, direct binding measurements by surface plasmon resonance (SPR) spectroscopy, coimmunoprecipitation and functional reconstitution utilizing progressive deletion constructs and peptide competition studies. The GCAP1 site is composed of two subdomains: transduction, M445-L456 and binding, L503-I522 (Lange et al., 1999), and the GCAP2 of one continuous domain, Y965-N981, containing both the transduction and the binding site (Duda et al., 2005a).

These two GCAP sites have different sensitivities to capture [Ca2+]i signals and activate ROS-GC1. The [Ca2+] K1*/*<sup>2</sup> for GCAP1 is 707 nM and for GCAP2, 100 nM (Hwang et al., 2003). Thus, the latter is about one-order of magnitude more sensitive to [Ca2+]i in signaling ROS-GC1 activation.

At a secondary structural level, there also exists a striking difference between the activities of the two GCAPs. The indication is that dimerization of GCAP2 is a necessary requisite to activate the ROS-GC (Olshevskaya et al., 1999). In contrast, the dimeric form of GCAP1 is inactive and does not result in activation of the ROS-GC (Hwang et al., 2004).

## **GCAP-MODULATED ROS-GC1 PHOTOTRANSDUCTION MODEL (FIGURE 3)**

The biochemical process by which the rods and cones convert the incoming LIGHT signal into the generation of the electrical signal is termed photo-transduction. Educated by the ROS specific photoreceptor features of the interlocked GCAP/ROS-GC transduction system, a model for its role in operation of the phototransduction machinery is proposed (**Figure 3**). This model explains two principal characteristics of photo-transduction machinery, recovery and its variable luminosity dependent operation; the latter characteristic adds an advanced feature to the previous models (Pugh et al., 1999; Burns and Baylor, 2001; Koch et al., 2002, 2010; Sharma, 2002; Luo et al., 2008; Stephen et al., 2008; Wensel, 2008) and depicts the differential involvement of GCAPs in its operation. The luminosity-dependent phototransduction principles of the present model incorporate a core theme of the previously proposed "Ca2+-relay model" (Koch, 2006; reviewed in Koch et al., 2010; Sharma, 2010), which explained the results of a transgenic GCAPs null mice study where

**operation of the ROS-GC-GCAP transduction system. Left panel.** An illustration of a typical vertebrate rod. In the DARK a circulating current (arrows) is present. It is outward in the inner segment and carried primarily by K+; in the outer segment the net charge is inward, with about 90% of the inward flow carried by the Na<sup>+</sup> and 10% by Ca2<sup>+</sup> ions. Na+/K<sup>+</sup> exchange pumps in the inner segment membrane and Na+/K+-Ca2<sup>+</sup> exchangers in the

outer segment membrane (see also right panels) maintain the overall ionic gradients against the dark flows. The capture of a photon (hν) by a rhodopsin molecule in one of the disc membranes of the outer segment initiates the photo-transduction cascade. **Right upper panel. DARK.** The components of the Photo-Transduction cascade are shown in the dark/resting steady-state. Cytoplasmic cyclic GMP, generated by the basal catalytic activity of ROS-GC, *(Continued)*

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

keeps a fraction of CNG channels in the plasma membrane open. ROS-GC1 via its 415M-L456 segment is GCAP1- and via 965Y-N981 is GCAP2-bound. Ca2<sup>+</sup> ions enter the cell via the CNG-channel and are extruded via the Na+/K+, Ca2+-exchanger. Synthesis and hydrolysis of cyclic GMP by ROS-GC and PDE, respectively, occur at a low rate. The heterotrimeric G protein transducin is in its GDP-bound state and is inactive. The Ca2<sup>+</sup> binding proteins calmodulin (CaM), recoverin (Rec) are bound to their target proteins, the CNG-channel, rhodopsin kinase (Rhk), respectively. **Right middle panel.** Absorption of **BRIHT LIGHT** by the visual pigment rhodopsin leads to the activation of the transduction cascade: the GTP-bound α-subunit of transducin activates PDE that rapidly hydrolyzes cGMP. Subsequently the CNG-channels close and the

the role of GCAP2 was questioned (Howes et al., 2002; Wen et al., 2014).

## **MIGRATION PATTERNS AND TRANSLATIONS OF THE GCAP SIGNALS INTO PRODUCTION OF CYCLIC GMP ARE DIFFERENT**

After settling the issue that the biochemical modes of two GCAPs in Ca2<sup>+</sup> signaling of the photoreceptor ROS-GC1 are different and autonomous of each other, the next task was to analyze their migratory pathways at the structural levels, determine the mechanisms involved in transmission and translation of the signals at the CCD of ROS-GC1 into the production of cyclic GMP.

To comprehend these issues and their resolutions, **Figure 4** is presented and the parts of the discussion are excerpted (Duda et al., 2012) and briefly recounted. The study involved a programmed domain deletion, expression, *in vivo* fluorescence spectroscopy, and *in vitro* reconstitution experiments. The findings validated (1) in structural terms that operational modes of two GCAPs in signaling of ROS-GC1 activity are different. GCAP1 signal transduction occurs through M445-L456 and L503- I <sup>522</sup> subdomains of JMD and of GCAP2 signal transduction via the Y965-N981 subdomain of CTE. These subdomains are at the opposite ends of CCD and they are the sites where the signal of the respective GCAPs originates.

(2) In living cells, both GCAPs by themselves are soluble proteins. Only when they are present with ROS-GC do they become membrane bound through their specific target sites on the ROS-GC1. Thus, ROS-GC1 is the one that bestows on them the property of being membrane bound; the anchoring sites are M445-L456 and L503-I522 for GCAP1 and Y965-N981 for GCAP2. In accordance with these conclusions, disruption of the GCAP1 anchoring site on ROS-GC1 through mutagenesis, results in GCAP1 being a soluble protein. In contrast, GCAP2 remains bound to this ROS-GC1 as its activator. These results prove that the two GCAPs signal ROS-GC1 activation through different modes and the differences reside in the spatial characteristics of ROS-GC1; and because orientations of the two domains of their signal origins are different, their migration pathways are different: GCAP1 downstream from M445-L456 and L503-I522 to the P808-K1054. This assumption was validated through the studies with the SHD-deleted ROS-GC1 mutant. Being downstream from the JMD, the site of the GCAP1 signal, and being a component of the GCAP1 signal trajectory, deletion of the SHD disrupted the GCAP1 signaling of ROS-GC1. But the GCAP2 signaling of the ROS-GC remained intact.

Ca2+-concentration falls. The fall in cytoplasmic [Ca2+]i is sensed by Ca2+-binding proteins: CaM dissociates from the CNG-channel what leads to an increase in cyclic GMP sensitivity of the channel; recoverin stops inhibiting rhodopsin kinase; rhodopsin becomes phosphorylated. Both Ca2+-free GCAPs in their changed configurations activate ROS-GC and synthesis of cyclic GMP increases. Arrestin (Arr) binds to phosphorylated rhodopsin and interferes with the binding and further activation of transducin. Enhancement of cyclic GMP synthesis brings it to its original DARK state level and termination of the cascade, which leads to reopening of CNG channels. **Right bottom panel, DIM LIGHT**. The initial fall of [Ca2+]i is selectively detected only by GCAP1. In its Ca2+-free state GCAP1 attains the activated mode and stimulates ROS-GC activity. GCAP2 remains Ca2+-bound and in its inhibitory mode.

Strikingly, these results also revealed the existence of an in intriguing signaling pathway, never observed before for the membrane guanylate cyclase family. This pathway is unique to GCAP2 and runs opposite to that of GCAP1, upstream from the Y965- N981 site in CTE to the CCD for translation of the signal into the production of cyclic GMP (**Figure 4**).

(3) The CCD is a common and conserved translation site for the GCAP1, GCAP2 and all other signals generated by the ligands of the membrane guanylate cyclase family. The current model, based on biochemical analysis, is one in which CCD has intrinsic basal catalytic activity and in its isolated form exists as a homodimer (Venkataraman et al., 2008). Through molecular modeling, this secondary structure has been advanced to its 3D-form, which reveals that its two subunits are in antiparallel conformation (Liu et al., 1997; Venkataraman et al., 2008). Is this model applicable in depicting the 3D structure of CCD in a living cell?

It is. The studies using the bimolecular fluorescence complementation (BiFC) technique demonstrated for the first time that in living cells, the CCD is present in its homodimeric form; and the two subunits are fused in the antiparallel orientations (Duda et al., 2012).

#### **ROS-GC1 GENE LINKED RETINAL DYSTROPHIES**

The disclosure of true identity, complete structure and organization of the ROS-GC1 gene (Duda et al., 1998) and the mechanism by which Ca2<sup>+</sup> signal modulates its encoded ROS-GC1 enzyme activity made it possible to investigate its role in the human retinal diseases at the genetic levels and explain their biochemistry in molecular terms. For recent comprehensive reviews on this topic the reader is referred to (Hunt et al., 2010; Karan et al., 2010; Sharma, 2010; earlier findings to the year 2002 are covered in Duda and Koch, 2002; Newbold et al., 2002).

In their original study Perrault et al. (1996) demonstrated that there is a point mutation in the human ROS-GC1 gene, F565S. Patients inheriting this mutation are inflicted with Leber's congetinal amauresis type 1 (LCA1); they are born blind or become blind soon after birth. Identity of the GCAP1 regulatory regions in bovine ROS-GC1 provided an opportunity to define this dystrophy in molecular terms (Lange et al., 1999). The human F565S residue corresponds to the bovine F514S residue. This residue is located in the mapped L503-I522 ROS-GC1 binding region for GCAP1. Biochemical analysis on the heterologusly expressed LCA1-linked ROS-GC1 mutant in COS cells demonstrated that the point mutation causes almost complete loss (84%) of the basal

**panel:** Schematic representation of the structural topography of ROS-GC1. It is a single transmembrane homodimer protein. The dashed lines on the right show the defined boundaries of its segments: LS, leader sequence; ExtD, extracellular domain; TM, transmembrane domain; ICD, intracellular domain. All its functional domains are housed in ICD; their designated names and the amino acid residues constituting their boundaries are indicated at the left: JMD, juxtamembrane domain housing the indicated GCAP1-targeted domain; KHD, kinase homology domain; SHD-signaling helix domain; CCD-core catalytic domain; CTE-C-terminal extension housing the GCAP2-targeted

catalytic activity of ROS-GC1 and also of its Ca2<sup>+</sup> sensitivity to modulation by GCAP1 (Duda et al., 1999b). Thus, the basic structural integrity of ROS-GC1 has been deranged by the mutation and this has also incapacitated GCAP1-modulated Ca2<sup>+</sup> signaling mechanism of the guanylate cyclase.

Following this lead, similar approach has been used to investigate the second type of retinal disease that correlates with mutations of the ROS-GC1 gene. Named cone-rod dystrophy type 6 (CORD 6), patients suffering with this disease carry one or several point mutations in the dimerization domain of ROS-GC1 (Kelsell et al., 1998). One example is ROS-GC1-E786D, R787C, T788M mutation. The mutation deranges the dimer formation of the guanylate cyclase and reduces the basal catalytic activity of the ROS-GC (Duda et al., 1999a, 2000; Tucker et al., 1999). This basic structural derangement in the guanylate cyclase, intriguingly, increases its sensitivity to GCAP1 and GCAP2. Physiological consequences are that there is a change in the trigger of the Ca2+-feedback causing a shift in the response-intensity curve domain. It is noteworthy, that the sites targeted by the two GCAPs (encircled) are at the opposite ends of CCD. **Right panel:** The two GCAP pathways run in opposite directions. The trajectory of the GCAP1 pathway shown in blue dashed arrow is downstream. From its origin in the JMD, it passes through the structural domains of KHD and SHD in its course to CCD. In contrast, the trajectory of the GCAP2 pathway (shown in pink dashed arrow) is upstream. From its origin in the CTE, it directly flows to CCD. The CCD exists as an antiparallel homodimer. Both GCAP signals are translated at the CCD into the production of cyclic GMP, which serves the second messenger of the LIGHT signals.

to lower light intensities. These effects explain the photophobia often reported by CORD 6 patients and the accumulating light damage of the retina leading to loss of cone and rod vision over a time period of decades.

## **CONTROLLED BY LOW AND HIGH LEVELS OF [CA2+]I, ROS-GC1 IS A BIMODAL TRANSDUCTION SWITCH**

Discovered contemporaneously with GCAPs, CD-GCAP (Ca2+ dependent guanylate cyclase activator) is the conformational isomer of S100B protein (Pozdnyakov et al., 1995, 1997; Margulis et al., 1996; Duda et al., 1996a, 2002; Wen et al., 2012). In contrast to the commercial S100B that is apparently Zn2+-bound, CD-GCAP native to the retinal neurons is Zn2+-free (Pozdnyakov et al., 1997).

S100B possesses a striking property toward ROS-GC1 stimulation that is opposite to that of GCAPs. This finding started a new paradigm of Ca2<sup>+</sup> signaling to the cyclic GMP pathway. Recombinant ROS-GC1 expressed in a heterologous system of COS cells responds at 10 nM [Ca2+]i to GCAP1 stimulation in a dose dependent manner. Free Ca2<sup>+</sup> inhibits this stimulation with a K1*/*<sup>2</sup> of 100 nM (Duda et al., 1996a). Zn2+-free S100B, however, stimulates native and recombinant ROS-GC1 with a [Ca2+] K1*/*<sup>2</sup> of 400 nM. Presence and ability of the ROS-GC to couple to GCAPs and/or to S100B places a neuronal cell under stringent [Ca2+]i modulation. The cell can accelerate or de-accelerate the generation of cyclic GMP; and in this manner can operate as a Ca2<sup>+</sup> bimodal switch. In this hypothetical scheme, the bimodal Ca2<sup>+</sup> ROS-GC transduction switch would be an elegant general mechanism for neural transmissions.

This hypothesis was put to test (Duda et al., 2002). Direct interaction between ROS-GC1 and S100B was demonstrated by cross-linking studies utilizing bis-(sulfosuccinimidyl)suberate. A cross-linked product between ROS-GC1 dimer and S100B occurred. Studies with ROS-GC1 deletion constructs confined an interaction of S100B with the C-terminus of ROS-GC1, aa731– 1054. An EC50 of 395 nM, calculated using SPR spectroscopy, was comparable to the EC50 of 800 nM for stimulation of ROS-GC1 by S100B. Refined analysis revealed that aa952–991 segment of ROS-GC1 defines the S100B binding site with half maximal binding at 198 nM. The site contained a 966RIHVNS972 motif that was obligatory for binding and a flanking cluster of four aa residues, 1039RRQK1042 forming a transduction site. This transduction site did not contribute to S100B binding but promoted maximal activation of ROS-GC1.

These interactive physical parameters between S100B and ROS-GC1 were brought to the functional level in the synaptic layers of the retina where the co-presence of ROS-GC1, GCAP1 and S100B existed (Liu et al., 1994; Cooper et al., 1995; Duda et al., 2002). Below 200 nM, Ca2<sup>+</sup> elicited a dose-dependent decrease in guanylate cyclase activity with an IC50 for Ca2<sup>+</sup> of 100 nM. In contrast, at higher concentrations, ROS-GC activity began to climb with an EC50 of 0.8μM. That GCAP1 is involved in the inhibitory phase of ROS-GC1 activity was confirmed by peptide competition experiments. A sequence motif L503-I522 in ROS-GC1 is critical and specific for the cyclase activation by GCAP1 (Lange et al., 1999). The core motif within this domain consists of D507-R518.

Thus, in the synapse ROS-GC1 acts as a bimodal Ca2<sup>+</sup> transduction switch. GCAP1 has the higher affinity for Ca2<sup>+</sup> and upon binding, lowers ROS-GC1 activity. S100B captures the high Ca2<sup>+</sup> signal and upon binding, stimulates ROS-GC1. This ROS-GC1 signal transduction model is schematically diagrammed in (Wen et al., 2012) (**Figure 4**: Sharma et al., 2014).

#### **RECENT FINDINGS**

ERG recordings on the retinas of the mouse S100B−*/*<sup>−</sup> model suggest that S100B up-regulates ROS-GC1 Ca2+-dependent catalytic activity and modulates the transmission of neural signals to cone ON-bipolar cells (Wen et al., 2012).

## **CA2+-MODULATED NEUROCALCIN** δ **ROS-GC1 TRANSDUCTION SYSTEM EXISTS IN THE INNER PLEXIFORM LAYER (IPL) OF THE RETINA**

Besides S100B, a NCS family member, neurocalcin δ (NCδ), is present in bovine IPL (Krishnan et al., 2004; details reviewed in Sharma and Duda, 2012). Co-expressed with ROS-GC1 in the heterologous system of COS cells, NCδ stimulates ROS-GC1 in a Ca2+-dependent manner with a K1*/*<sup>2</sup> of 0.8μM. Its target site in the ROS-GC is V837-L858 (Venkataraman et al., 2008). In accordance with their mutual physiological interaction, NCδ and ROS-GC1 are present together in IPL; added presence of [Ca2+]i stimulates native IPL ROS-GC1 activity and nM range of Ca2<sup>+</sup> keeps NCδ ROS-GC1 bound. The Ca2+-dependent NCδ interactive ROS-GC1 kinetic parameters are: without Ca2+, NCδ has no affinity for ROS-GC1; in the presence of Ca2+, it binds ROS-GC1 with a K*<sup>A</sup>* of 2*.*<sup>3</sup> <sup>×</sup> 106 <sup>M</sup>−<sup>1</sup> and a *KD* of 4.6 <sup>×</sup>10−<sup>7</sup> M.

These analyses reveal that the steps of NCδ binding to and dissociation from ROS-GC1 are Ca2+-dependent, they occur within the physiological levels of Ca2+, they are direct and of moderate affinity; and they occur within the time span of the visual transduction steps. And significantly, the target site of NCδ for ROS-GC1 is unique to itself; it does not overlap with that of S100B or the GCAPs.

An extraordinary characteristic about the NCδ binding domain in ROS-GC is that it resides directly within the CCD (Venkataraman et al., 2008). This finding demonstrated that NCδ-modulated Ca2<sup>+</sup> signaling of ROS-GC1 occurs through a new model, its principles are (1) NCδ directly interacts with CCD; (2) it does not require the adjacent N-terminally located α-helical dimerization domain structural element for its interaction; (3) the CCD module, housing the site, is intrinsically active, i.e., it has basal guanylate cyclase activity. (4) The CCD module by itself is dimeric; it does not require the dimerization domain structural element for being so; and (5) the core dimeric form of the catalytic module is directly regulated by the Ca2+-bound NCδ; Ca2+-unbound NCδ is ineffective (Venkataraman et al., 2008).

#### **ROS-GC LINKAGE WITH OTHER THAN VISION-LINKED NEURONS**

The very brief following sections are meant to demonstrate that the interlocked elements of the Ca2<sup>+</sup> sensors and ROS-GC family are not unique to just the visual transduction processes they expand their signaling roles to many other sensory and sensorylinked transduction processes.

#### **PINEAL GLAND**

The bovine pineal gland embodies the Ca2+-modulated GCAP1 and S100B-modulated ROS-GC1 transduction systems. They exist in two separate sets of pinealocytes. In one set, in the GCAP1 mode, the ROS-GC catalytic activity is modulated by free [Ca2+]i K1*/*<sup>2</sup> of 100 nM. The second set is in the S100B mode, the ROS-GC activity is modulated by [Ca2+]i with K1*/*<sup>2</sup> of about 0.5μM. It has been proposed that these two processes may be involved in the dark and light states of the two types of pinealocytes (Venkataraman et al., 1998, 2000).

#### **OLFACTION**

Following the lead observation of Duda et al. (2001b) that GCAP1-modulated Ca2+- signaling of ROS-GC1 transduction system is expressed in the rat olfactory bulb, its presence and a role in odorant transduction was investigated (Duda et al., 2001a). Prior to these studies role of the membrane guanylate cyclase transduction system was denied in the odorant transduction and the system was exclusively believed to be "involved in behavioral responses induced by hormones or pheromones, possibly related to reproduction, rather than a response to specific odorants" (Juilfs et al., 1997; Meyer et al., 2000). Yet, these studies revealed the presence of selected components of the membrane guanylate cyclase in the ONE-GC (GC-D) neurons. These neurons were segregated from the dominant odorant linked cyclic AMP signaling neurons.

#### *Odorant Transduction Model (Duda and Sharma, 2009, Figure 4)*

A small population of the olfactory receptor neurons (ORN) contains a cyclic GMP signal transduction pathway (Juilfs et al., 1997; Leinders-Zufall et al., 2007; Duda and Sharma, 2008; Pertzev et al., 2010). This pathway resides at the apical region of the cilia. Present in this region is ONE-GC. Its outer domain is a receptor for uroguanylin (Duda and Sharma, 2008). In its inner domain, at the C-terminus, resides the catalytic domain (Duda and Sharma, 2008). The M880-L921 segment of this domain is bound to Ca2<sup>+</sup> sensor component NCδ. In the resting state, the ORN is in a 60–100 nM range of [Ca2+]i and ONE-GC is in its basal state. The uroguanylin odorant signal starts by its interaction with the receptor domain of ONE-GC (Duda and Sharma, 2008). It is processed through two sequential steps. In step one, ONE-GC is primed and activated minimally. In step 2, [Ca2+]i rises. With a K1*/*<sup>2</sup> of 0.3–0.8μM, Ca2<sup>+</sup> binds NCδ facilitating its interaction with the ONE-GC's segment M880-L921 (Duda and Sharma, 2008; Duda et al., 2004); Ca2+-bound NCδ than signals full activation of ONE-GC and maximal synthesis of cyclic GMP (reviewed in Sharma and Duda, 2010a,b; Zufall and Munger, 2010).

It is envisioned that the operation of step 2 starts with the generation of a small amount of cyclic GMP in step 1. This pool of cyclic GMP opens a limited number of the cyclic GMP-gated channels causing influx of Ca2<sup>+</sup> in the ORN. Ca2<sup>+</sup> binds NCδ, which then fully activates ONE-GC.

#### **SPECIAL NOTE**

There are clues that, besides NCδ, two additional Ca2<sup>+</sup> sensors of ONE-GC are involved in the odorant transduction. One is hippocalcin (Hpca) (Krishnan et al., 2009) and the other is GCAP1 (Pertzev et al., 2010). The studies with the Hpca−*/*<sup>−</sup> mouse model and with the GCAP1-specific antibodies demonstrate that both Ca2<sup>+</sup> sensors are physically linked with ONE-GC and are present in the rat and the mouse olfactory neuroepithelium region. In addition, 35% of the total ONE-GC transduction activity is controlled by GCAP1, 27% by NCδ, and 38% by Hpca (reviewed in: Sharma and Duda, 2010a).

#### **GUSTATION**

Similar to olfaction, gustation (taste) belongs to chemical senses. In response to the sweet, sour, salty, bitter, and umami, taste receptor cells generate electric signal in a process called gustatory transduction (Avenet and Lindemann, 1989; Gilbertson et al., 2000, 2001; Herness, 2000; Lindemann, 1996). The molecular steps in this transduction process are ill defined. Yet in a general scheme Ca2<sup>+</sup> is pivotal for the process; it depolarizes plasma membrane of the given taste cell, releases the transmitter release and functions through the CNG-gated channels, cyclic AMP and cyclic GMP (Kolesnikov and Margolskee, 1995).

The expression of a Ca2<sup>+</sup> signaling ROS-GC1 transduction system in the anterior portion of the bovine gustatory epithelium has been demonstrated at the biochemical, molecular and functional levels (Duda and Sharma, 2004). The system is composed of two components: the Ca2+-sensor protein, S100B and the transducer, ROS-GC1. Co-immunoprecipitation experiments reveal that ROS-GC1 and S100B physically interact with each other. The precise operational mechanism of this signal transduction mechanism and its physiology has yet not been decoded (reviewed in Sharma and Duda, 2010b).

## **EVOLUTION OF A GENERAL CA2+-INTERLOCKED ROS-GC SIGNAL TRANSDUCTION CONCEPT IN SENSORY AND SENSORY-LINKED NEURONS**

Extraordinary feature of the ROS-GC signal transduction system is that it is an extremely delicate Ca2+-modulated machinery. It is crafted impressively to do so. In a range of nM, [Ca2+]i signals produce the ROS-GC activity to its peak level and then they decline it with a [Ca2+]i K1*/*<sup>2</sup> of 100 nM. In a bimodal fashion, the signals then accelerate it with a Ca2<sup>+</sup> K1*/*<sup>2</sup> of 500 nM to 800 nM. The ROS-GC accomplishes the former mode through its Ca2<sup>+</sup> sensor partners, GCAPs and the latter function through its CD-GCAP partners. This mechanism of bimodal regulation enables ROS-GC to generate the Ca2+-dependent pulsated levels of cyclic GMP, which functions as a second messenger of the sensory and sensory-linked Ca2<sup>+</sup> transduction mechanisms. Presented illustrations are: (1) LIGHT signal induced GCAP-modulated ROS-GC transduction mode in the photoreceptor cells; (2) bimodal GCAP1 and S100B modes in the photoreceptor-cone synapse; (3) the odorant signal–GCAP1, S100B and Hpca–regulated modes in ONE-GC neurons and (4) not yet well-understood processes of the pineal gland and gustation. Bestowed with this design and clues of its linkage with physiology of the various sensory neurons, these investigators have proposed that "Ca2+-modulated ROS-GC" machinery is a central signal transduction component of all the sensory and/or sensory-linked secondary neurons. Accompanied by a nearby CNG channel, the machinery is empowered to de-or hyper-polarize neurons, thereby affecting the generation of action potentials (model in Figure 4 of Sharma, 2010). It becomes an elegant theoretical general mechanism for neural transmissions.

## **CONCLUSION AND FUTURE DIRECTIONS**

This review has briefly chronicled the events that have resulted in the step-by-step development of the field of membrane guanylate cyclase. Its foundation rests on the initial seminal finding made five decades ago (1963) that cyclic GMP exists in rat urine. Based on the conceptual framework of the cyclic AMP second messenger, it implied the existence of a hormonally-dependent membrane guanylate cyclase transduction system that generates cyclic GMP in the mammalian cells. For about a decade, this concept underwent a tortuous evolution; first, the euphoria and then a chaos. Barring ours, and few others, most of the fashionable groups denied existence of this transduction system. The field rekindled with the first purification and characterization of a membrane guanylate cyclase, ANF-RGC. Surprisingly, besides being a guanylate cyclase, it was also a hormone receptor. And, the field of membrane guanylate cyclase was born. Yet, the notion evolved that the membrane guanylate cyclase family is solely represented by its hormone-receptors. The field expanded and changed with the discovery of a Ca2+-modulated ROS-GC transduction system. Yet it remained restrictive, believed to be linked with the single physiological function of phototransduction in the photoreceptor cells. This belief also changed with the discovery that besides GCAPs, which are selectively linked with phototransduction, other Ca2<sup>+</sup> sensors coupled with ROS-GC exist beyond the photoreceptor cells. Consistent with its broader role, Ca2+ sensor S100B modulated ROS-GC transduction system is linked with the photoreceptor-bipolar neural transmissions. In addition, the system is also present in the pinealocytes and a ROS-GCvariant (ONE-GC) transduction system is expressed in the rodent olfactory systems.

Exponential expansion of the field has generated new problems which have to be solved. Some of them are briefly outlined below.

The first at hand deals with the modes of guanylate cyclases modulation. Have all the modulators of each cyclase been discovered? Recent studies suggest that it may not be the case as exemplified by ONE-GC. For over a decade this cyclase was considered by some groups an orphan receptor (Fulle et al., 1995; Juilfs et al., 1997) but it is now known that it is the receptor for odorant uroguanylin/guanylin (Leinders-Zufall et al., 2007; Duda and Sharma, 2008; Zufall and Munger, 2010). In this new paradigm, the extracellularly-generated odorant signal needs to be further processed in the intracellular domain of the cyclase in a Ca2+-dependent manner (Duda and Sharma, 2009). Thus, ONE-GC becomes the first cross-over functional guanylate cyclase, being modulated by both odorant peptide receptor domain and the Ca2+-modulated domain. How many other guanylate cyclases possess this feature and what is the molecular mechanism underlying the process of dual regulation?

Second, a new model of ANF-RGC signal transduction has evolved (Duda et al., 2012). ANF-RGC is the bimodal switch; one switch is the traditional ANF hormone and the other, [Ca2+]i. [Ca2+]i signal targets directly the CCD of ANF-RGC and increases the production of cyclic GMP. What is its mechanism of operation? Are other hormone receptor guanylate cyclases also bimodal switches?

Third, the ANF-RGC receptor subfamily is multifunctional. Besides controlling the well-known processes linked with cardiac vasculature—diuresis, fluid secretion, smooth muscle relaxation, it also influences the processes involved in metabolic syndrome and obesity control (Martel et al., 2010). Similarly, CNP-RGC triggered by its hormonal CNP signal affects such diverse processes as bone growth (Bartels et al., 2004) and axonal sprouting in neurons (Schmidt et al., 2007, 2009). How does it achieve this characteristic?

Transcending to the ROS-GC subfamily, it is also multifunctional; linked with the physiology of visual transduction in the retina, photoentrainment in the pineal gland, as well as odorantand gustatory transduction. Except for phototransduction, the underlying molecular principles of its linkage with the other sensory processes are lacking.

Fourth, a new model of ROS-GC1 signaling specific to cone photoreceptors has emerged. Its key molecule is the S100B protein. Molecular principles of this model need to be deciphered.

Fifth, ill-defined information suggests that GC-G is linked with the processes of capacitation (Kuhn et al., 2004), atmospheric CO2 detection (Chao et al., 2010) and is also modulated by cool ambient temperature (Mamasuew et al., 2008). The molecular explanations for these processes are not available.

Sixth, a common trait of the membrane guanylate cyclase family is that it is a multi-modular protein. Experimentally validated evidence indicates that each of its modular domains is precisely crafted to control the cyclase's functional specificity (Sharma and Duda, 1997). Complete function of each domain is lacking, however. Consequently, none of the intact membrane guanylate cyclase structures have been solved by X-ray crystallography.

Seventh, and final, the basic knowledge gained by the preceding studies will be invaluable in defining the physiology of each of the guanylate cyclases, linking them with respective pathological states, and in the long run, in designing targeted therapies against these pathologies.

## **ACKNOWLEDGMENTS**

RKS gratefully acknowledges the continuous support for the past 35 years by the numerous USPHS awards from the National Institutes of Health, the beginning awards from the National Science Foundation and the Damon Runyon Walter Winchell Cancer Fund. TD acknowledges National Heart, Blood and Lung Institute grant HL084584. The authors are thankful to Ms Sanya Sharma for editorial assistance.

## **REFERENCES**


of bovine photoreceptor guanylate cyclase. *Biochemistry* 35, 6263–6266. doi: 10.1021/bi960007m


the atrial natriuretic factor. *FEBS Lett*. 219, 375–379. doi: 10.1016/0014- 5793(87)80256-9


*Biophys. Res. Commun*. 100, 508–514. doi: 10.1016/S0006-291X(81) 80206-9


**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: 14 March 2014; accepted: 30 May 2014; published online: 02 July 2014. Citation: Sharma RK and Duda T (2014) Membrane guanylate cyclase, a multimodal transduction machine: history, present, and future directions. Front. Mol. Neurosci. 7:56. doi: 10.3389/fnmol.2014.00056*

*This article was submitted to the journal Frontiers in Molecular Neuroscience. Copyright © 2014 Sharma and Duda. 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.*

## Structural diversity of neuronal calcium sensor proteins and insights for activation of retinal guanylyl cyclase by GCAP1

## *Sunghyuk Lim1, Alexander M. Dizhoor <sup>2</sup> and James B. Ames 1\**

<sup>1</sup> Department of Chemistry, University of California at Davis, Davis, CA, USA

<sup>2</sup> Basic Sciences, Pennsylvania College of Optometry, Salus University, Elkins Park, PA, USA

#### *Edited by:*

Rameshwar K. Sharma, Salus University, USA

#### *Reviewed by:*

Baojin Ding, University of Massachusetts Medical School, USA Frank Schmitz, Saarland University, Germany

#### *\*Correspondence:*

James B. Ames, Department of Chemistry, University of California at Davis, One Shields Avenue, Davis, CA 95616, USA e-mail: jbames@ucdavis.edu

Neuronal calcium sensor (NCS) proteins, a sub-branch of the calmodulin superfamily, are expressed in the brain and retina where they transduce calcium signals and are genetically linked to degenerative diseases. The amino acid sequences of NCS proteins are highly conserved but their physiological functions are quite different. Retinal recoverin controls Ca2+-dependent inactivation of light-excited rhodopsin during phototransduction, guanylyl cyclase activating proteins 1 and 2 (GCAP1 and GCAP2) promote Ca2+-dependent activation of retinal guanylyl cyclases, and neuronal frequenin (NCS-1) modulates synaptic activity and neuronal secretion. Here we review the molecular structures of myristoylated forms of NCS-1, recoverin, and GCAP1 that all look very different, suggesting that the attached myristoyl group helps to refold these highly homologous proteins into different three-dimensional folds. Ca2+-binding to both recoverin and NCS-1 cause large protein conformational changes that ejects the covalently attached myristoyl group into the solvent exterior and promotes membrane targeting (Ca2+-myristoyl switch). The GCAP proteins undergo much smaller Ca2+-induced conformational changes and do not possess a Ca2+ myristoyl switch. Recent structures of GCAP1 in both its activator and Ca2+-bound inhibitory states will be discussed to understand structural determinants that control their Ca2+-dependent activation of retinal guanylyl cyclases.

**Keywords: calcium, EF-hand, Ca2+-myristoyl switch, NCS-1, recoverin, GCAP1, NCS protein, NMR**

#### **INTRODUCTION**

Intracellular calcium ions (Ca2+) regulate neuronal signaling in the central nervous system (Berridge et al., 2000; Augustine et al., 2003). Neuronal Ca2<sup>+</sup> signals are detected by a family of neuronal calcium sensor (NCS) proteins (Ames et al., 1996, 2012; Braunewell and Gundelfinger, 1999; Burgoyne and Weiss, 2001; Burgoyne et al., 2004; Weiss et al., 2010) that contain EF-hand motifs (Moncrief et al., 1990; Ikura, 1996; Ikura and Ames, 2006) as well as by a family of C2-domain containing proteins (synaptotagmin and protein kinase C isoforms; Nalefski and Falke, 1996; Corbalan and Gomez, 2014). At least sixteen different NCS proteins are known (Weiss and Burgoyne,2002; Burgoyne and Haynes, 2012) and are conserved from yeast to humans (**Figure 1**). Recoverin (Dizhoor et al., 1991) and guanylyl cyclase activating proteins 1 and 2 (GCAP1 and GCAP2; Dizhoor et al., 1994; Palczewski et al., 1994) are expressed in the retina, where they regulate phototransduction in photoreceptor cells (Palczewski et al., 2000; Ames and Ikura, 2002; Stephen et al., 2008; Ames et al., 2012). NCS proteins are also expressed in the brain such as neurocalcin (Hidaka and Okazaki, 1993), frequenin (NCS-1; Pongs et al., 1993; McFerran et al., 1998), visinin-like proteins (VILIPs; Bernstein et al., 1999; Braunewell and Klein-Szanto, 2009), K+ channel interacting proteins (KChIPs;An et al.,2000), calsenilin/DREAM (Buxbaum et al., 1998; Carrion et al., 1999), and hippocalcin (Kobayashi et al., 1992, 1993; Tzingounis et al., 2007).

Recoverin also called S-modulin (Dizhoor et al., 1991; Kawamura and Murakami, 1991), the first NCS protein to be discovered, controls the lifetime of photo-excited rhodopsin (Kawamura, 1993; Erickson et al., 1998; Makino et al., 2004) by regulating rhodopsin kinase (Calvert et al., 1995; Chen et al., 1995; Klenchin et al., 1995; Komolov et al., 2009). Recoverin decreases the lifetime of rhodopsin at low Ca2<sup>+</sup> levels to control visual recovery and promote photoreceptor adaptation to background light. More recent evidence indicates that recoverin can also modulate the decay of the light-activated phsophodiesterase activity. Such modulation may help accelerate visual recovery in the presence of background light (Chen et al., 2012). Recoverin is also located in the rod inner segment (Strissel et al., 2005) and is associated with cancer-associated retinopathy (Polans et al., 1991; Subramanian and Polans, 2004).

Guanylyl cyclase activating proteins 1 and 2 are also expressed in photoreceptor cells where they activate retinal guanylyl cyclase at low cytosolic Ca2<sup>+</sup> levels upon light activation (Dizhoor et al., 1994; Palczewski et al., 1994, 2004). The EF-hand motifs in GCAPs can bind both Mg2<sup>+</sup> and Ca2<sup>+</sup> (Peshenko and Dizhoor, 2004, 2006). Mg2<sup>+</sup> binding stabilizes a structural form of GCAPs that activates cyclase activity (Peshenko and Dizhoor, 2006; Lim et al., 2009), whereas Ca2+-bound GCAPs inhibit the cyclase (Dizhoor and Hurley, 1996; Dizhoor et al., 1998). GCAPs are important for regulating the recovery phase of visual excitation and particular mutants are linked to various forms of retinal degeneration (Semple-Rowland et al., 1996; Sokal et al., 1998; Baehr and Palczewski, 2007; Bondarenko et al., 2010; Jiang and Baehr, 2010).

Neuronal calcium sensor proteins (frequenin or NCS-1) are expressed in other tissues beside the brain (Kapp et al., 2003) and in lower organisms including flies (Pongs et al., 1993), worms (Gomez et al., 2001), and yeast (Frq1; Hendricks et al., 1999; Huttner et al., 2003; Hamasaki et al., 2004). Yeast NCS homologs (called Frq1) activate a phosphatidyl inositol 4-OH kinase isoform (Pik1; Hendricks et al., 1999; Kapp et al., 2003; Strahl et al., 2003, 2007) required for vesicle trafficking and secretion (Hama et al., 1999; Walch-Solimena and Novick, 1999). Mammalian NCS-1 interacts with voltagegated Ca2<sup>+</sup> and K<sup>+</sup> channels (Weiss et al., 2000; Nakamura et al., 2001) and activates inositol trisphosphate receptors (Boehmerle et al., 2006).

NCS proteins typically contain about 200 amino acid residues in chain length with four EF-hand motifs, a first EF-hand that does not bind Ca2+, and a myristoylation consensus sequence at the N-terminus. NCS proteins have similar sequences, ranging from 35 to 60% identity (**Figure 1**). EF-hand residues are the most highly conserved, particularly in the Ca2<sup>+</sup> binding loops. The fourth EF-hand sequence is variable, and Ca2<sup>+</sup> is able to bind to EF4 in frequenin (Cox et al., 1994; Ames et al., 2000) and GCAPs (Peshenko and Dizhoor, 2007; Stephen et al., 2007) but Ca2<sup>+</sup> does not bind to EF4 in recoverin (Ames et al., 1995a) and VILIPs (Cox et al., 1994; Li et al., 2011). Ca2+-binding to EF4 in GCAP1 controls whether GCAP1 can activate or inhibit guanylyl cyclase (Peshenko and Dizhoor, 2007). The residues near the C-terminus and linker between EF3 and EF4 are non-conserved, suggesting that these regions may play a role in target specificity for recoverin but not for GCAPs.

Retinal recoverin and most other NCS proteins are myristoylated at the amino terminus (Dizhoor et al., 1992; Kobayashi et al., 1993; Ladant, 1995). Recoverin and GCAPs contain a saturated myristoyl (14:0) or related fatty acyl group (12:0, 14:1, 14:2), because N-myristoyl transferase (Gordon et al., 1991) can efficiently utilize C12 and/or C14 acetyl-CoA as fatty acyl donors in the retina. In tissues other than the retina, myristoylation is the predominant modification. Myristoylated recoverin binds to cell membranes only at high Ca2<sup>+</sup> levels (Zozulya and Stryer, 1992; Dizhoor et al., 1993), whereas unmyristoylated recoverin does not bind to membranes. Likewise, bovine neurocalcin (Ladant, 1995) and hippocalcin (Kobayashi et al., 1993) both are myristoylated and exhibit Ca2+-induced localization at the plasma membrane in response to neuronal stimulation. Ca2+-induced membrane targeting by NCS proteins has been termed, Ca2+-myristoyl switch. The attached fatty acyl group is buried inside the protein structure of Ca2+-free recoverin (Tanaka et al., 1995). Ca2<sup>+</sup> binding to recoverin causes extrusion of the fatty acid, enabling it to interact with lipid bilayer membranes. Recoverin's Ca2+-myristoyl switch may control its light-induced movement into the rod inner segment (Strissel et al., 2005). GCAP proteins are also myristoylated (Palczewski et al., 1994; Frins et al., 1996; Olshevskaya et al., 1997). However, unlike recoverin, GCAPs do not possess a functional Ca2+-myristoyl switch (Olshevskaya et al., 1997; Hwang and Koch, 2002). Instead the N-terminal myristoyl group remains

sequestered inside GCAP1 in both Ca2+-free and Ca2+-bound states (Hughes et al., 1998; Lim et al., 2009). Indeed, the crystal structure of Ca2+-bound GCAP1 shows the myristoyl group surrounded by the protein (Stephen et al., 2007), and a recent nuclear magnetic resonance (NMR) structural analysis reveals that the activator state of GCAP1 has an overall structure similar to that of Ca2+-bound inhibitory state in which the N-terminal myristoyl group is buried in both the Ca2+-free and Ca2+-bound states (Lim et al., 2013).

Atomic-resolution structures are known for myristoylated forms of recoverin (Ames et al., 1997), GCAP1 (Stephen et al., 2007), and NCS-1 (Lim et al., 2011) that each fold differently around the attached myristoyl group (**Figure 2**). For NCS-1, the attached myristoyl group is located in a protein crevice formed by helices from EF3 and EF4 near the C-terminus (**Figure 2A**). The covalently attached fatty acyl group in NCS-1 protrudes in a parallel fashion between four helices from EF3 and EF4 (**Figure 2B**). The C-terminal location of the myristoyl binding site in NCS-1 is quite different from that of recoverin in which the myristate is positioned inside a cavity near the N-terminus (**Figure 2C**). The attached fatty acyl chain in recoverin is wedged between the helices of EF1 and EF2 in a perpendicular fashion (**Figure 2D**), which contrasts with the parallel arrangement of thefatty acyl chain in NCS-1 (**Figure 2B**). For GCAP1, the myristoyl group is sequestered in a cavity formed by the N-terminal domain with participation of a C-terminal helix (**Figure 2E**). The myristoyl group in GCAP1 bridges the N-terminal and C-terminal ends of the protein by contacting helices at each end (**Figure 2F**). In short, the protein structural environment around the myristoyl group is quite different in recoverin, GCAP1 and NCS-1 (**Figure 2**). This suggests that each NCS protein folds differently around the N-terminal myristoyl group by contacting non-conserved patches of hydrophobic residues that are unique to each NCS protein. However, myristoylation of GCAP2 is not essential for its ability to activate its target (retinal guanylyl cyclase), because unmyristoylated GCAP2 can activate cyclase activity nearly as well as myristoylated GCAP2 (Olshevskaya et al., 1997).

In this review, we discuss recent structures of GCAP1 in both Ca2+-free activator and Ca2+-bound inhibitor states to understand structural determinants that control Ca2+-dependent activation of retinal guanylyl cyclases.

## **Ca2<sup>+</sup>-DEPENDENT ACTIVATION OF RETINAL GUANYLYL CYCLASE**

### **GUANYLYL CYCLASE ACTIVATION AND PHOTORECEPTOR RECOVERY**

In vertebrate rods and cones, photon absorption by rhodopsin or cone visual pigments triggers a phototransduction cascade that hydrolyzes cGMP, resulting in the closure of cGMP-gated cation channels on the plasma membrane that causes membrane hyperpolarization [see reviews (Pugh et al., 1997, 1999)]. To reset the resting dark-state of retinal rods (known as visual recovery), cGMP levels are replenished very quickly (Burns and Baylor, 2002) by retina-specific guanylyl cyclases (RetGCs), a membrane enzyme present in rod and cone outer segments (Dizhoor et al., 1994; Lowe et al., 1995). RetGC is a Ca2+-regulated enzyme (Koch and Stryer, 1988; Koutalos and Yau, 1996) whose activity is controlled by intracellular domains (Laura et al., 1996; Duda et al., 2005) that interact with soluble EF-hand Ca2<sup>+</sup> sensor proteins, called guanylyl cyclase activating proteins (GCAPs: GCAP1 and GCAP2; Dizhoor et al., 1994, 1995; Gorczyca et al., 1994, 1995; Koutalos et al., 1995).

Light-induced channel closure in photoreceptor cells causes a decrease in the cytosolic free Ca2<sup>+</sup> concentration (Gray-Keller and Detwiler, 1994), in mammals from 250 nM in the dark to 25 nM in the light (Woodruff et al., 2002). The catalytic activity of RetGC in the dark is negatively controlled by Ca2+-bound GCAPs (Dizhoor and Hurley, 1996; Dizhoor et al., 1998; Burns and Baylor, 2002), whereas the release of Ca2<sup>+</sup> from GCAPs at low Ca2<sup>+</sup> levels in light-activated photoreceptors causes activation of RetGC (Dizhoor et al., 1994, 1995; Gorczyca et al., 1994, 1995; Dizhoor and Hurley, 1996; Mendez et al., 2001). Light stimulation of the rod cell causes a more than 10-fold increase in cGMP production due to the activation of RetGC by GCAPs (Hodgkin and Nunn, 1988; Burns and Baylor, 2002) and is a critical step for controlling the recovery rate of a single-photon response (Pugh et al., 1999; Burns and Baylor, 2002) as well as the cone response to stronger light stimuli (Sakurai et al., 2011). In mouse rods, GCAPs have been demonstrated to have different Ca2<sup>+</sup> sensitivities (Dizhoor et al., 1998; Hwang et al., 2003) and therefore contribute to the recovery by activating guanylyl cyclase at different steps of excitation and recovery, thus imparting proper recovery kinetics to the rod response (Mendez et al., 2001; Makino et al., 2008, 2012).

#### **Mg2+ AND Ca2+ BINDING TO GCAPs RECIPROCALLY CONTROL CYCLASE ACTIVATION**

Guanylyl cyclase activating proteins activate RetGC at low Ca2<sup>+</sup> levels (less than 50 nM) and only in the presence of physiological Mg2<sup>+</sup> levels (Peshenko and Dizhoor, 2004, 2006, 2007; Dizhoor et al., 2010). This Mg2<sup>+</sup> requirement for RetGC activation by GCAPs initially suggested that Mg2<sup>+</sup> binding to GCAPs might be important for their activation of RetGC. Indeed, Mg2<sup>+</sup> was shown to bind directly to at least two of the EF-hands in GCAP1 (Lim et al., 2009), and NMR studies showed that Mg2<sup>+</sup> binding to GCAP1 at EF2 and EF3 was needed to stabilize the overall tertiary fold of the protein (Lim et al., 2009). By contrast, the Ca2+-free/Mg2+-free GCAP1 (apo-state) forms a molten globulelike structure, that contains regular secondary structure (Dell'Orco et al., 2010) but does not form a stable tertiary fold (Peshenko and Dizhoor, 2004; Lim et al., 2009). The flexible and unstructured molten-globule apo-protein could explain in part why GCAPs do not activate RetGC in the absence of Mg2<sup>+</sup> (Dizhoor et al., 2010). Thus, Mg2<sup>+</sup> binding to GCAP1 stablizes its protein structure in order to bind and activate RetGC (Dizhoor et al., 1994; Peshenko and Dizhoor, 2004), whereas Ca2<sup>+</sup> binding to GCAP1 stabilizes a distinct structure important for the inhibition of RetGC (Dizhoor et al., 1998).

The four EF-hands in the GCAPs have quite distinct divalent metal binding properties that control whether GCAPs can activate or inhibit RetGC. The first EF-hand (EF1) does not bind to either Ca2<sup>+</sup> or Mg2<sup>+</sup> because the residue at the 3-position in the EF-hand binding loop (Cys29 in GCAP1, see **Figure 1**) is not suitable for ligating either Ca2<sup>+</sup> or Mg2+. Ca2<sup>+</sup> binds to GCAP1 at the other three EF-hands (EF2, EF3, and EF4) in an independent fashion

(Lim et al., 2009) in contrast to the cooperative binding of two Ca2<sup>+</sup> to recoverin (Ames et al., 1995a). The apparent dissociation constant for Ca2<sup>+</sup> binding to GCAPs is in the submicromolar range (Lim et al., 2009; Dizhoor et al., 2010), whereas Mg2<sup>+</sup> binds with ∼1000-fold lower affinity (Gifford et al., 2007) in the submillimolar range (Peshenko and Dizhoor, 2004, 2007; Lim et al., 2009). These binding affinities imply that three Ca2<sup>+</sup> bind per

mole of GCAP1 in dark-adapted rod cells, which have relatively high cytosolic Ca2<sup>+</sup> levels [Ca2+]free <sup>=</sup> 250–500 nM [(Woodruff et al., 2002; Matthews and Fain, 2003) and [Mg2+] <sup>∼</sup>1 mM (Chen et al., 2003)]. Light-activation of the rod cell causes a dramatic lowering of the cytosolic Ca2<sup>+</sup> level [Ca2+]free <sup>=</sup> 5–50 nM (Gray-Keller and Detwiler, 1994; Sampath et al., 1998; Woodruff et al., 2002) while the Mg2<sup>+</sup> level remains fixed at [Mg2+]free <sup>∼</sup>1 mM

(Chen et al., 2003). Therefore in light-adapted rods, GCAPs do not bind Ca2<sup>+</sup> but instead bind to at least two Mg2+. Thus, Ca2+ free/Mg2+-bound GCAPs activate RetGC in light exposed rods (Dizhoor et al., 1994, 1995; Gorczyca et al., 1995), whereas Ca2+ bound GCAPs (with Ca2<sup>+</sup> bound at EF2, EF3, and EF4) inhibit RetGC in dark-adapted rods (Dizhoor and Hurley, 1996; Dizhoor et al., 1998).

#### **CONSTITUTIVELY ACTIVE MUTANTS OF GCAP1 CAUSE RETINAL DISEASE**

Various point mutations in the EF-hand motifs of GCAP1 that weaken Ca2<sup>+</sup> binding (but do not affect Mg2<sup>+</sup> binding) cause GCAP1 to constitutively activate RetGC in rods and cones, which is genetically linked to various retinal diseases (Jiang and Baehr, 2010). These mutations in the EF-hand motifs [Y99C (Dizhoor et al., 1998; Payne et al., 1998) and E155G (Wilkie et al., 2000, 2001)] weaken the Ca2<sup>+</sup> binding affinity beyond the photoreceptor Ca2<sup>+</sup> concentration and cause the Ca2+-free/Mg2+-bound GCAP1 activator state to persist even at high Ca2<sup>+</sup> levels in dark-adapted rods, which causes persistent activation of RetGC (Sokal et al., 1998; Olshevskaya et al., 2004). The GCAP mutants that constitutively activate RetGC then cause elevated cGMP levels in photoreceptor cells that promotes apoptosis and disease (Olshevskaya et al., 2004, 2012; Woodruff et al., 2007).

Mutagenesis studies of the individual EF-hands in GCAP1 have revealed that Ca2+-binding to EF4 is critical for controlling Ca2+-dependent activation of RetGC (Peshenko and Dizhoor, 2007). Mutants of GCAP1 that weaken or abolish Ca2<sup>+</sup> binding to EF4 but retain Ca2<sup>+</sup> binding at EF2 and EF3 [D144N/D148G (Peshenko and Dizhoor, 2007 and E155G Wilkie et al., 2000, 2001)] are constitutively active even in the presence of high Ca2<sup>+</sup> levels in dark-adapted photoreceptors. Furthermore, these mutants that disable Ca2<sup>+</sup> binding to EF4 (but not EF2 and EF3) are unable to inhibit RetGC at high Ca2<sup>+</sup> levels in the dark (Peshenko and Dizhoor, 2007). In summary, GCAP1 can activate RetGC even if Ca2<sup>+</sup> is bound to EF2 and EF3 (but not bound to EF4). Also, Ca2<sup>+</sup> binding to EF4 is essential for having Ca2+ induced inhibition of RetGC. Therefore, Ca2<sup>+</sup> binding to EF4 is critical for controlling whether GCAP1 can activate or inhibit RetGC.

#### **GCAPs DO NOT POSSESS A Ca2+-MYRISTOYL SWITCH**

Ca2<sup>+</sup> binding to GCAP1 and GCAP2 does not cause ejection of the covalently attached myristoyl group (Hughes et al., 1998; Lim et al., 2009, 2013) and Ca2<sup>+</sup> binding to GCAPs do not promote their membrane targeting (Olshevskaya et al., 1997; Hwang and Koch, 2002). This is in stark contrast to the Ca2+-induced exposure of the N-terminal myristoyl group [termed Ca2+ myristoyl switch (Zozulya and Stryer, 1992; Dizhoor et al., 1993)] that promotes membrane targeting of recoverin (Zozulya and Stryer, 1992; Dizhoor et al., 1993; Valentine et al., 2003), neurocalcin (Ladant, 1995), hippocalcin (O'Callaghan et al., 2003), VILIPs (Spilker et al., 1997, 2002), and NCS-1 (Hamasaki et al., 2004). Instead, the covalently attached myristoyl group of GCAP1 remains sequestered inside the protein hydrophobic core in both Ca2+-free and Ca2+-bound forms of GCAP1 (Lim et al., 2009, 2013). NMR studies on the myristate attached to recoverin (Ames et al., 1995b, 1997; Hughes et al., 1995), VILIP1 (Li et al., 2011), and NCS-1 (Lim et al., 2011) reveal that the covalently attached myristoyl group is buried inside these proteins only in the Ca2+ free state. Ca2+-binding causes protein conformational changes that lead to exposure of the fatty acyl chain in recoverin (Ames et al., 1997), VILIP1 (Li et al., 2011), and NCS-1 (Lim et al., 2011). By stark contrast, NMR studies on GCAP1 indicate that the covalently attached myristoyl group is buried inside both Ca2+-free and Ca2+-bound GCAP1 (Lim et al., 2009, 2013). Further evidence that GCAP1 lacks a Ca2+-myristoyl switch comes from SPR studies that show myristoylation of GCAP1 has little effect on membrane binding (Hwang and Koch, 2002). Finally, the recent atomic level structures of GCAP1 directly demonstrated that the myristoyl group is buried inside the protein in a similar environment in both the Ca2+-free activator (Lim et al., 2013) and Ca2+-bound inhibitor states (Stephen et al., 2007).

Solid-state NMR and other spectroscopic studies on GCAP2 have suggested that Ca2+-free GCAP2 might be targeted to cell membranes by a reversed Ca2+-myristoyl switch (Theisgen et al., 2010, 2011). The covalently attached fatty acyl group has been suggested to become exposed in Ca2+-free GCAP2 in the presence of lipid bilayer membranes (Theisgen et al., 2011), in contrast to having the myristoyl group sequestered inside Ca2+-bound GCAP2 (Schroder et al., 2011). However, other studies on GCAP2 indicate that the myristoyl group remains sequestered inside the protein environment in both Ca2+-free and Ca2+-bound GCAP2 (Hughes et al., 1998). Also, unmyristoylated GCAP2 activates RetGC with nearly the same potency as that of myristoylated GCAP2, and myristoylation of GCAP2 is not essential for its targeting to the membrane-bound cyclase (Olshevskaya et al., 1997; Hwang and Koch, 2002).

#### **Ca2+-INDUCED PROTEIN CONFORMATIONAL CHANGES IN GCAPs**

Atomic level structures are known for Ca2+-bound forms of GCAP1 (Stephen et al., 2007) and GCAP2 (Ames et al., 1999). The four EF-hands in GCAP1 (**Figures 1** and **3**) are grouped into two globular domains: N-domain includes EF1 and EF2 (residues 18–83) and C-domain contains EF3 and EF4 (residues 88–161). Ca2<sup>+</sup> is bound to GCAP1 at EF2, EF3 and EF4, and the structure of each Ca2+-bound EF-hand in GCAP1 (**Figure 3B**) adopts the familiar open conformation of EF-hands as seen in calmodulin and other Ca2+-bound EF-hand proteins (Ikura, 1996). Indeed, the interhelical angles for each Ca2+-bound EF-hand in GCAP1 are similar to those of recoverin (Ames et al., 1997) and NCS-1 (Bourne et al., 2001). Although the internal structure of each EF-hand in GCAP1 is similar to that of recoverin, the overall three-dimensional packing arrangement and spatial organization of the four EF-hands is very different for GCAP1 vs. recoverin. Indeed, the overall root-mean-squared deviation of main chain atoms is 3.4 Å when comparing the structures of GCAP1 and recoverin (**Figure 2**). A unique structural feature of GCAP1 is that the N-terminal α-helix (residues 5–15) upstream of EF1 and C-terminal helix (residues 175–183) downstream of EF4 are held closely together by their mutual interaction with the N-terminal myristoyl group (**Figure 2F**). Thus, the covalently

attached myristoyl group in GCAP1 is sequestered within a unique environment inside the Ca2+-bound protein, which is quite different from that of recoverin as described above in **Figure 2**. The myristoyl group attached to GCAP1 makes contacts with N-terminal residues (V9, L12, and F42) and the C-terminal helix (L174, V178 and I181; **Figure 2F**). In essence, the myristoyl group serves to bridge both the N-terminal and C-terminal ends of the protein, which explains how Ca2+-induced conformational changes in the C-terminal domain (particularly in EF4) might be transmitted to a possible target binding site in EF1. A Ca2+-myristoyl tug mechanism (Peshenko et al., 2012) has been proposed to explain how Ca2+-induced conformational changes in EF4 serve to "tug" on the adjacent C-terminal helix that connects structurally to the myristoyl group and EF1. This tug mechanism helps explain how Ca2+-induced structural changes in EF4 might be relayed to the cyclase binding region in EF1 (Lim et al., 2013). The Ca2+-induced structural changes involving the C-terminal helix might also be related to Ca2+ dependent phosphorylation of S201 in GCAP2 (Peshenko et al., 2004).

The atomic level structure of Ca2+-free/Mg2+-bound activator form of GCAP1 or GCAP2 is currently not known. The main difficulty is that Ca2+-free/Mg2+-bound GCAP proteins form dimers and higher order protein oligomers that causes considerable sample heterogeneity at high protein concentrations needed for NMR or to make crystals for X-ray crystallography (Ames et al., 1999; Lim et al., 2009, 2013). Ca2+-dependent dimerization of GCAP2 has been suggested to be important for activating the cyclase (Olshevskaya et al., 1999b; Ermilov et al., 2001). Protein dimerization was also reported for GCAP1, and a GCAP1 mutant (V77E) that prevents protein dimerization also abolishes its ability to activate RetGC, suggesting that dimerization of Ca2+-free/Mg2+ bound activator state might be important for activating RetGC (Lim et al., 2013). However, protein dimerization of GCAP1 is not Ca2<sup>+</sup> dependent and only occurs at relatively high protein concentrations above 100 μM (Lim et al., 2013). Future studies are needed to investigate whether GCAPs might form a functional dimer upon binding to the dimeric RetGC to form a 2:2 complex (Ramamurthy et al., 2001; Peshenko et al., 2010), and

whether the dimeric quaternary structure may play a regulatory role.

A GCAP1 mutant, D144N/D148G called EF4mut, (Lim et al., 2013) that binds Ca2<sup>+</sup> at EF2 and EF3 (and does not bind Ca2<sup>+</sup> at EF4) can activate RetGC at high Ca2<sup>+</sup> concentrations (Peshenko and Dizhoor, 2007). Therefore, EF4mut (with Ca2<sup>+</sup> bound at EF2 and EF3) serves as a model of the Ca2+-free/Mg2+-bound activator state. The EF4mut protein is more soluble and stable than Ca2+-free/Mg2+-bound wild type, and EF4mut exhibits NMR spectra with much higher resolution and sensitivity compared to that of Ca2+-free/Mg2+-bound wild type (Lim et al., 2009, 2013). NMR structural studies on EF4mut provide some insights for a structural model of the GCAP1 activator state (**Figure 3A**). The overall structure of EF4mut is similar to the crystal structure of Ca2+-bound GCAP1 (root mean squared deviation of main chain atoms is 1.3 Å when comparing the two structures). However, residues at the domain interface (between EF2 and EF3) are structured somewhat differently in EF4mut compared to the crystal structure of Ca2+-bound GCAP1. Many of the GCAP1 residues at the domain interface have quite broad NMR resonances, suggesting that these residues are conformationally dynamic (Lim et al., 2013). The corresponding residues in recoverin (**Figure 1**) also exhibited broad NMR resonances and 15N NMR relaxation dispersion studies reveal that these domain interface residues exhibit millisecond exchange kinetics (Xu et al., 2011). Ca2+-induced rearrangement of residues at the domain interface in recoverin gives rise to a 45◦ swiveling of the two domains (Ames et al., 1997). A structural comparison between EF4mut and Ca2+-bound GCAP1 suggests a related but much smaller Ca2+-induced structural change at the domain interface in GCAP1 (**Figure 3**). The most noteworthy Ca2+-induced structural difference in GCAP1 can be seen in the entering helix of EF3 that unravels a half turn in EF4mut, which causes a repositioning of the W94 side-chain at the domain interface. A Ca2+-induced change in the structural environment around W94 is consistent with previous tryptophan fluorescence and electron paramagnetic resonance studies of GCAP1 (Sokal et al., 2001; Peshenko and Dizhoor, 2006, 2007). We suggest that Ca2+-induced rearrangement of residues at the domain interface (V77, A78, L82, K85, and

W94) plays a role in modulating Ca2+-dependent contacts with RetGC1.

#### **ACTIVATION MECHANISM FOR RetGC BY GCAPs**

The structural information for the GCAPs above provides insights into the activation mechanism of RetGC (**Figure 4**). GCAP1 residues in EF1 (Krylov et al., 1999; Olshevskaya et al., 1999a,b; Ermilov et al., 2001) and at the domain interface (Krylov et al., 1999; Lim et al., 2013; Peshenko et al., 2014) are suggested to make direct contact with RetGC (see labeled residues in **Figure 4**). In our model (**Figure 4**), Ca2+-induced conformational changes in EF4 (Lim et al., 2013) are transmitted to the cyclase binding site (see labeled residues in EF1, EF2, and EF3, **Figure 4**) by a Ca2+-myristoyl tug mechanism as described by (Peshenko et al., 2012). In the GCAP1 activator state under physiological conditions (**Figure 4**, left panel), EF2 and EF3 are bound to Mg2<sup>+</sup> (blue circles in **Figure 4**) with EF1 and EF4 unoccupied. The Ca2+-free state of EF4 forms a loose and dynamic structure (Lim et al., 2013), which allows the adjacent C-terminal helix to reach all the way to the N-terminal myristoyl group (magenta in **Figure 4**) and thus indirectly form hydrophobic contacts with residues in EF1 and EF2. In essence, the myristoyl group forms a bridge between Ca2+ induced conformational changes in the C-terminal domain and the cyclase binding site in the N-terminal domain. In the GCAP1 activator state, residues in the cyclase binding site [see **Figure 4**, labeled residues in EF1, EF2 and EF3 (Peshenko et al., 2014)] are spatially close together and form particular contacts with RetGC that require close proximity between W94 and V77 (see arrow in **Figure 4**). In the Ca2+-bound GCAP1 inhibitor state (**Figure 4**, right panel), Ca2<sup>+</sup> is bound to EF2, EF3, and EF4 (orange circles in **Figure 4**). Ca2+-binding to EF4 causes local conformational changes that in turn "tug" on the C-terminal helix which causes a slight reorientation of the N-terminal domain (EF1 and EF2) with respect to the C-terminal domain (EF3 and EF4). This domain

**FIGURE 4 | Schematic diagram of retinal guanylyl cyclase activation by GCAP1.** RetGC is shown in gray. The four EF-hands in GCAP1 are colored coded as defined in **Figure 1**, bound Mg2<sup>+</sup> are depicted by blue circles, bound Ca2<sup>+</sup> are orange circles, the C-terminal helix is white, and N-terminal myristoyl group is magenta. The labeled GCAP1 residues (K23, M26, G32, F73, V77, A78, and W94) have been implicated in the RetGC binding site (Lim et al., 2013; Peshenko et al., 2014). The binding of Ca2<sup>+</sup> at EF4 causes conformational changes that are transmitted to the RetGC binding site by a Ca2+-myristoyl tug mechanism (see arrows) modified from Peshenko et al. (2012).

swiveling causes key residues in EF3 at the domain interface (K85, W94, and K97) to move farther away from cyclase binding site residues in EF2 [F73, V77, and A78 (Peshenko et al., 2014)], which disrupts key contacts to RetGC that we suggest may be important for regulating cyclase activation.

## **CONCLUSION**

N-terminal myristoylation serves to remodel the structure of NCS proteins as seen for recoverin, GCAP1 and NCS-1. Each NCS protein folds differently around the attached myristoyl group, which causes each NCS protein to adopt a unique structure. Most NCS proteins possess a functional Ca2+-myristoyl switch that promotes their Ca2+-induced membrane targeting. By contrast, GCAP1 contains a sequestered myristoyl group in both its Ca2+-free and Ca2+-bound states and undergoes rather small Ca2+-induced protein conformational changes. Ca2<sup>+</sup> binding to the fourth EF-hand in GCAP1 triggers conformational changes in the N-terminal domain via a Ca2+-myristoyl tug mechanism that controls the exposure of residues in EF1 and EF2 that we propose may serve as a target binding site for guanylyl cyclase.

## **ACKNOWLEDGMENTS**

This work was supported by grants to James B. Ames (EY012347) and Alexander M. Dizhoor (EY11522) from the National Institutes of Health.

## **REFERENCES**


immunoreactivities (VILIP-1 and VILIP-3) in human brain. *J. Neurocytol.* 28, 655–662. doi: 10.1023/A:1007056731551


in the yeast *Saccharomyces cerevisiae*. *J. Biol. Chem.* 274, 34294–34300. doi: 10.1074/jbc.274.48.34294


myristoylated guanylate-cyclase activating protein-2. *Eur. Biophys. J.* 40, 565–576. doi: 10.1007/s00249-011-0680-9


**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 January 2014; paper pending published: 17 February 2014; accepted: 27 February 2014; published online: 17 March 2014.*

*Citation: Lim S, Dizhoor AM and Ames JB (2014) Structural diversity of neuronal calcium sensor proteins and insights for activation of retinal guanylyl cyclase by GCAP1. Front. Mol. Neurosci. 7:19. doi: 10.3389/fnmol.2014.00019*

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

*Copyright © 2014 Lim, Dizhoor and Ames. 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.*

## ROS-GC interlocked Ca2+-sensor S100B protein signaling in cone photoreceptors: review

## *Rameshwar K. Sharma1\*, Clint L. Makino2 \*, David Hicks <sup>3</sup> and Teresa Duda1\**

<sup>1</sup> Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, Elkins Park, PA, USA

<sup>2</sup> Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, Harvard Medical School, Boston, MA, USA

<sup>3</sup> Department of Neurobiology of Rhythms, Institute for Cellular and Integrative Neuroscience, CNRS UPR 3212, Strasbourg, France

#### *Edited by:*

Wolfgang Baehr, University of Utah, USA

#### *Reviewed by:*

Karl-Wilhelm Koch, Carl von Ossietzky University Oldenburg, Germany Wolfgang Baehr, University of Utah, USA

#### *\*Correspondence:*

Rameshwar K. Sharma and Teresa Duda, Research Divisions of Biochemistry and Molecular Biology, The Unit of Regulatory and Molecular Biology, Salus University, 8360 Old York Road, Elkins Park, PA 19027, USA e-mail: rsharma@salus.edu; tduda@salus.edu; Clint L. Makino, Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, Harvard Medical School, 234 Charles Street, Boston, MA 02114, USA e-mail: Clint\_Makino@meei. harvard.edu

Photoreceptor rod outer segment membrane guanylate cyclase (ROS-GC) is central to visual transduction; it generates cyclic GMP, the second messenger of the photon signal. Photoexcited rhodopsin initiates a biochemical cascade that leads to a drop in the intracellular level of cyclic GMP and closure of cyclic nucleotide gated ion channels. Recovery of the photoresponse requires resynthesis of cyclic GMP, typically by a pair of ROS-GCs, 1 and 2. In rods, ROS-GCs exist as complexes with guanylate cyclase activating proteins (GCAPs), which are Ca2+-sensing elements. There is a light-induced fall in intracellular Ca2+. As Ca2<sup>+</sup> dissociates from GCAPs in the 20–200 nM range, ROS-GC activity rises to quicken the photoresponse recovery. GCAPs then progressively turn down ROS-GC activity as Ca2<sup>+</sup> and cyclic GMP levels return to baseline. To date, GCAPs mediate the only known mechanism of ROS-GC regulation in the photoreceptors. However, in mammalian cone outer segments, cone synapses and ON bipolar cells, another Ca2<sup>+</sup> sensor protein, S100B, complexes with ROS-GC1 and senses the Ca2<sup>+</sup> signal with a K1/<sup>2</sup> of 400 nM. Unlike GCAPs, S100B stimulates ROS-GC activity when Ca2<sup>+</sup> is bound. Thus, the ROS-GC system in cones functions as a Ca2<sup>+</sup> bimodal switch; with rising intracellular Ca2+, its activity is first turned down by GCAPs and then turned up by S100B. This presentation provides a historical perspective on the role of S100B in the photoreceptors, offers a pictorial model for the "bimodal" operation of the ROS-GC switch and projects future tasks that are needed to understand its operation. Some accounts of this review have been adopted from the original publications of these authors.

**Keywords: ROS-GC guanylate cyclase, cyclic GMP, phototransduction, cones, S100B**

#### **INTRODUCTION**

A seminal observation in the field of visual transduction was that a soluble bovine rod outer segment (ROS) fraction kicked the catalytic activity of a photoreceptor guanylate cyclase into high gear in the absence of Ca2<sup>+</sup> (Koch and Stryer, 1988). The structure of this guanylate cyclase was believed to be composed of "separate regulatory and catalytic subunits" (Stryer, 1991); however, elucidating the molecular identity of the guanylate cyclase and its regulator turned out to be quite complicated. There were reports that guanylate cyclase possessed a molecular mass of 67 kDa and was nitric oxide sensitive (Horio and Murad, 1991a,b). Yet another report claimed to have solved the molecular structure of the photoreceptor retGC, the human retinal guanylate cyclase (Shyjan et al., 1992).

Later, the true photoreceptor ROS-guanylate cyclase (ROS-GC) was purified directly from bovine outer segments (Margulis et al., 1993) and its protein-sequence-based molecular cloning, structure, and function were established (Goraczniak et al., 1994). ROS-GC had a calculated molecular mass of 120,360 Da, a value similar to 112,000 Da reported earlier for a bovine (Koch, 1991) and 110,000–115,000 Da for toad photoreceptor guanylate cyclase (Hayashi and Yamazaki, 1991). Unlike the other known membrane guanylate cyclases, atrial natriuretic factor receptor guanylate

cyclase (ANF-RGC; Paul et al., 1987; Sharma, 1988; Chinkers et al., 1989; Duda et al., 1991) and type C natriuretic peptide CNP-RGC guanylate cyclase (Koller et al., 1991; Duda et al., 1993), ROS-GC was not hormonally responsive (Goraczniak et al., 1994). It thus represented a new subfamily of the membrane guanylate cyclases. Further study revealed that ROS-GC was not composed of "separate regulatory and catalytic subunits", it was not nitric oxide sensitive nor was its structure identical to retGC. The RetGC structure was eventually revised to match that of bovine ROS-GC (Accession number M92432).

Meanwhile, groups from the U.S. and the former Soviet Union joined forces to describe a newly purified Ca2<sup>+</sup> binding protein, recoverin, that appeared to regulate guanylate cyclase activity (Dizhoor et al., 1991) and adhere to the functional description of Koch and Stryer (Koch and Stryer, 1988). It was therefore considered a Ca2<sup>+</sup> modulator of ROS-GC; hence, named recoverin "because it promotes recovery of the dark state" (Dizhoor et al., 1991). However, ever purer preparations of recoverin showed a disturbing decline in guanylate cyclase stimulation. It became evident that recoverin was not the sought after regulator of guanylate cyclase activity and the initial conclusion was withdrawn (Hurley et al., 1993). Instead, recoverin exerted Ca2+-dependent control over the phosphorylation of photoexcited rhodopsin by rhodopsin

kinase (Chen et al., 1995; Klenchin et al., 1995; Kawamura, 1999). At almost the same time, two separate groups discovered similar but distinct guanylate cyclase activating proteins (GCAPs), GCAP1 (Palczewski et al., 1994; Subbaraya et al., 1994; Gorczyca et al., 1995; Frins et al., 1996) and GCAP2 (Dizhoor et al., 1995). These GCAPs did indeed stimulate ROS-GC activity as described by Koch and Stryer (1988).

Recombinant ROS-GC expressed in a heterologous system of COS cells responded at 10 nM [Ca2+]i to GCAP1 stimulation in a dose-dependent manner (Duda et al., 1996b). The stimulation was inhibited cooperatively by free Ca2<sup>+</sup> with a K1/<sup>2</sup> of 100 nM. Under identical conditions, GCAP1 had no effect on recombinant ANF-RGC, the peptide hormone receptor guanylate cyclase. An important characteristic of the transduction system was that GCAP remained bound to ROS-GC at low and high Ca2<sup>+</sup> levels, consistent with physiological observations (Koutalos et al., 1995). Similar reconstitution studies established GCAP2 as another Ca2+-sensing subunit of ROS-GC (Dizhoor et al., 1995; Gorczyca et al., 1995). When a second ROS-GC was subsequently discovered in the bovine retina, termed Ret-GC2 in human retina (Lowe et al., 1995), it was named ROS-GC2 to distinguish it from the original ROS-GC which was renamed as ROS-GC1 (Goraczniak et al., 1997). It became clear that a Ca2+-modulated system, composed of a pair of ROS-GCs and a pair of GCAPs, subserves phototransduction in the outer segments of photoreceptors. This conclusion was supported at the physiological level by the studies with the double knockout ROS-GC1/ROS-GC2 mouse model (Baehr et al., 2007). Rods and cones were non-functional, excluding the presence of a third guanylate cyclase linked with phototransduction.

Thus ROS-GCs are related to peptide hormone receptor guanylate cyclases but they do not respond to extracellular ligands. Instead, they exist in stable complexes with GCAPs that sense changes in internal [Ca2+]. The feature that incrementing [Ca2+]i progressively reduces ROS-GC1 catalytic activity was academically challenging because such a mechanism had never been observed before for any of the members of the guanylate cyclase family. All other members were solely stimulated by their effector ligands, natriuretic peptide hormones, at their extracellular domains. The molecular properties of this remarkable system and the role that the system plays in phototransduction are described in greater depth in Pugh et al. (1997), Sharma and Duda (2012).

#### **DISCOVERY OF CD-GCAP MEANT THAT ROS-GC1 COULD OPERATE AS A Ca2+ BIMODAL TRANSDUCTION SWITCH**

Contemporaneous with the discovery of GCAPs that stimulated ROS-GC1 at low Ca2+, a post-mitochondrial 100,000 *g* supernatant fraction from retina was found to stimulate recombinant ROS-GC at high Ca2<sup>+</sup> by as much as 25-fold (Pozdnyakov et al., 1995). The new factor was a peptide with an apparent molecular weight of 6–7 kDa protein that oligomerized to a functional size of 40 kDa. It stimulated native and recombinant ROS-GC with a K1/<sup>2</sup> for Ca2<sup>+</sup> of 2 μM with biochemical characteristics typical of a Ca2+-binding protein. To distinguish it from the just-discovered GCAPs, the new factor was named CD-GCAP (Ca2+-dependent guanylate cyclase activator protein; Pozdnyakov et al., 1995).

The capacity for ROS-GC to couple to GCAPs or to CD-GCAP allows for a cell to sculpt its response to Ca2+. Depending upon the cell's specific needs, it could express either GCAP or CD-GCAP to stimulate cyclic GMP synthesis at low (tens of nanomolar) or at high (sub-micromolar) ranges of [Ca2+]. An even more exciting possibility exists for the simultaneous expression of both Ca2+ regulatory subunits in the same cell. ROS-GC could then operate as a Ca2+-bimodal transduction switch, stimulated at low and at high free [Ca2+] by Ca2<sup>+</sup> -free GCAPs and by Ca2+-bound S100B, respectively. A theoretical Ca2+-BIMODAL ROS-GC transduction model embodying these features was proposed (**Figure 1**). Bimodal Ca2<sup>+</sup> sensing could be an elegant general mechanism for neural transmission.

## **CD-GCAP IS AN S100 PROTEIN**

The first indication that CD-GCAP might be related to S100 proteins was that it appeared as a 6–7 kDA Ca2+-binding protein on SDS-PAGE (Pozdnyakov et al., 1995). S100 proteins are known to run anomalously as 6–7 kDa proteins in SDS-PAGE even though their subunit molecular weight, calculated from amino acid (aa) sequence, is about 10.5 kDa (Kligman and Marshak, 1985). Additional evidence for the structural similarity of CD-GCAP to S100 emerged from mass spectrometry analysis (Pozdnyakov et al., 1997). The fragmented molecular masses of 10,580 Da for CD-GCAP and 10,582 Da for S100B were essentially the same. Tryptic digests also yielded indistinguishable fragmentation patterns. Finally, bovine CD-GCAP and bovine S100B

**FIGURE 1 | Original theoretical Ca2+ bimodal ROS-GC transduction model.** In the absence of bound Ca2+, GCAPs 1 and 2 boost ROS-GC1 catalytic activity. In contrast, S100B boosts ROS-GC1 activity in the Ca2+-bound state. GCAPs sense [Ca2+]i <sup>&</sup>lt; 200 nM while S100B senses [Ca2+]i <sup>&</sup>gt; 200 nM. Thus in the GCAPs operational mode the S100B switch is "OFF" and in the S100B operational mode the GCAP switch is "OFF" (Upgraded with permission from Sharma, 2010; Sharma and Duda, 2012).

were cloned and their predicted primary protein structures were identical (Pozdnyakov et al., 1997).

Despite being structurally identical proteins, biochemical analyses revealed two surprising functional discrepancies. As expected, S100A1–S100B and S100ββ dimer (S100B) from commercial sources stimulated cloned recombinant and wt-photoreceptor ROS-GC (Duda et al., 1996a; Margulis et al., 1996). But the Vmax of ROS-GC was about 50% higher when activated by CD-GCAP than when activated by commercially obtained S100B. In addition, the commercial S100B required about 20-times more calcium (3.2 x 10−<sup>5</sup> M vs. 1.5 x 10−<sup>6</sup> M for CD-GCAP) for half-maximal stimulation of guanylate cyclase. Clearly, CD-GCAP was not quite the same as the commercially obtained S100B.

In the search for the explanation for these functional differences, the proteins were analyzed for post-translational modifications and higher order structures (Pozdnyakov et al., 1997). After treating both proteins with hydroxylamine, a deacylating agent, CD-GCAP no longer activated guanylate cyclase while S100B activation retained its ability to do so. Furthermore, hydroxylamine fragmented CD-GCAP while S100B was unaffected. These results indicated that the conformations of the two proteins were different. Such a difference could have arisen from variations in the procedures for purifying the two proteins: CD-GCAP purification entailed heating at 75◦C in 5 mM Ca2+, while S100B purification included zinc affinity chromatography (commercial sources refused to provide these details but their references indicated so). To investigate, CD-GCAP was subjected to zinc affinity chromatography (CD-GCAP-to-S100B) and S100B was heated to 75◦C in Ca2<sup>+</sup> (S100B-to-CD-GCAP). Guanylate cyclase activation, calcium–sensitivity, and hydroxylamine lability measurements demonstrated that CD-GCAP-to-S100B was identical to S100B and that S100B-to-CD-GCAP was identical to CD-GCAP. Apparently zinc inhibits the native S100B potency toward stimulation of ROS-GC. An additional finding was that S100B monomer is about twice as effective as the dimer in activating the ROS-GC catalytic activity. Hence, CD-GCAP and commercially obtained S100B are conformational isomers. The native form of S100B present in the retinal neurons is CD-GCAP.

### **CHARACTERIZATION OF THE S100B BINDING SITE OF ROS-GC1**

Direct interaction between ROS-GC1 and S100B was verified by cross-linking experiments with bis-(sulfosuccinimidyl)suberate. A cross-linked complex, consisting of ROS-GC1 dimer and S100B, was detected by both ROS-GC1 and S100B antibodies on Western blots (Duda et al., 2002). Biochemical experiments with ROS-GC1 deletion constructs localized an interaction of S100B with the C-terminus of ROS-GC1, aa 731–1054 (Duda et al., 1996a). An EC50 of 395 nM, measured using surface plasmon spectroscopy, was comparable to the EC50 of 800 nM for stimulation of ROS-GC1 by S100B in a biochemical assay (Duda et al., 2002).

Since the aa 733–964 segment containing the putative dimerization and catalytic domains is highly conserved, the less conserved aa 965–1054 segment was subjected to further analyses. First, three ROS-GC1 deletion mutants: -965–1054, -972–1054, and


These studies validated the unique ability of ROS-GC1 to respond not only to low Ca2<sup>+</sup> signals mediated by GCAPs but also to high Ca2<sup>+</sup> signals mediated by S100B. To analyze the possible interplay of these signals, the aforementioned ROS-GC1 deletion mutants were analyzed for their responses to GCAP1 (Duda et al., 2002) and GCAP2 (Duda et al., 2005). At 10 nM Ca2+, activation of all mutants was identical to that of wild-type ROS-GC1. Thus, the S100B and GCAP1 regulatory sites of ROS-GC1 are distinct and do not interfere with one other. This result was not surprising because the binding site for GCAP1 was previously mapped to the N-terminal part of the ROS-GC1 cytoplasmic domain (Duda et al., 1999; Lange et al., 1999).

The case for GCAP2 was not so clear. Initial mapping had localized the GCAP2 binding site closer to the C-terminus of ROS-GC1 (Krishnan et al., 1998). In reconstitution experiments with the ROS-GC1 deletion mutants (the same as used for mapping the S100B site: *vide supra*), the -965–1054 mutant failed to respond to GCAP2, the -972–1045 responded with Vmax of ∼50% of that of the wild-type ROS-GC1, and the response of the -1016–1054 mutant matched that of the wildtype cyclase, indicating that the functional GCAP2-modulated domain resided within aa 965–1016. Detailed peptide competition assays narrowed down the GCAP2-modulated site to the sequence motif Y965-N981 (Duda et al., 2005). As expected, this site is distinct from the GCAP1 site. It, however, partially overlaps with the S100B-regulatory site. These results provide a hint that, when GCAP1 and S100B are co-expressed, a single ROS-GC1 complex could operate as a bimodal Ca2<sup>+</sup> switch. On the other hand, when GCAP2 and S100B are co-expressed, the switch might involve separate populations of ROS-GC1 complexes. Furthermore, identification of the binding sites for the Ca2<sup>+</sup> sensors disclosed an intriguing topographical feature of ROS-GC1; GCAP2 and S100B transmit the Ca2<sup>+</sup> signals to the catalytic domain of ROS-GC1 from the C-terminal side while GCAP1 does so from a more distant location on the N-terminal side. These intramolecular mechanics are specific to ROS-GC1 and are different from the peptide hormone receptor subfamily, where the signal to the catalytic site migrates solely downstream from a site on the N-terminus located on the opposite side of

the membrane. A schematic model of the system is depicted in **Figure 2**.

ROS-GC2 has an ∼8x lower affinity for S100B and with S100B bound at high Ca2+, its maximal activation is twofold lower than that of ROS-GC1. Substitution of aa 731–1053 of ROS-GC1 into the corresponding region of ROS-GC2 and vice versa reversed the selectivity (Duda et al., 1998). These results suggest that ROS-GC2 may be an S100B partner of lesser importance than ROS-GC1.

#### **S100B IS EXPRESSED IN CONES BUT NOT IN RODS**

Guided by the knowledge that ROS-GC and GCAPs are present in photoreceptors, including their outer segments where they serve as critical components of phototransduction (Cuenca et al., 1998; Kachi et al., 1999; reviewed in Pugh et al., 1997; Koch et al., 2010; Sharma and Duda, 2012), and that S100B is also present in photoreceptors (Rambotti et al., 1999; reviewed in: Rambotti et al., 2002), questions were raised as to whether a ROS-GC1 bimodal Ca2<sup>+</sup> -regulated system operates in visual transduction and whether it is present in both rods and cones. These issues are being addressed in mouse (e.g., Wen et al., 2012), for which

 $\text{There exist useful knockouts including S100B}^{-}\text{ ${}^{\text{-}}$ , GOAP1}^{-}\text{ ${}^{\text{-}}$ , GOAP2}^{-}\text{ ${}^{\text{-}}$ , GOAP2}^{-}\text{ ${}^{\text{-}}$ , ROAP2}^{-}\text{ ${}^{\text{-}}$ , ROS-GC1}^{-}\text{ ${}^{\text{-}}$ }$ 

Affinity purified S100B antibody recognized on Western blots of mouse photoreceptors, a protein of molecular weight corresponding to S100B. In biochemical assays, guanylate cyclase activity of mouse ROS membranes exhibited bimodal Ca2<sup>+</sup> switch behavior; activity decreased as [Ca2+] rose from 10 to 300 nM and then increased as [Ca2+] began to exceed 300 nM. The"inhibitory" phase of the guanylate cyclase activity was consistent with the loss of ROS-GC1 activation by GCAPs upon Ca2<sup>+</sup> binding to the latter and the stimulatory phase with Ca2+-induced activation of ROS-GC1 activity by S100B. The increase in ROS-GC activity at low [Ca2+] disappeared in ROS membranes from GCAP1/GCAP2−/<sup>−</sup> mice, whereas the increase in ROS-GC activity at high [Ca2+] was missing in ROS membranes from S100B−/<sup>−</sup> mice and from ROS-GC1−/<sup>−</sup> mice. The significance of the result from ROS-GC1−/<sup>−</sup> mice will be revealed below. Yet, regulation of ROS-GC1 by S100B was not functionally linked to phototransduction in rods because flash responses from wild-type and S100B-KO rods were indistinguishable (Wen et al., 2012). The possibility that S100B might

**FIGURE 2 |The modular nature of ROS-GC1.** Left Panel. The dashed lines on the right show the domain boundaries: LS, leader sequence (note: the leader sequence is absent from the mature protein shown in the right panel); ExtD, extracellular domain; TM, transmembrane domain; ICD, intracellular domain. The designated names and the amino acid residues constituting their boundaries are indicated on the left: JMD, juxtamembrane domain housing the indicated GCAP1-binding site; KHD, kinase homology domain; SHD, signaling helix domain; CCD, core catalytic domain; CTE, C-terminal extension housing the S100B-binding site. It is noteworthy that the ICD contains all of the functional domains and that

GCAP1 and S100B target sites on opposite sides of the CCD. Right panel. The GCAP1 and S100B signaling pathways within ROS-GC propagate in opposite directions. The trajectory of the GCAP1 pathway is shown with the red dashed arrow. From its origin in the JMD, it passes "downstream" through the structural KHD and SHD in its course to the CCD. The trajectory of the S100B pathway shown with the blue dashed arrow. Originating in the CTE, it directly flows "upstream" to the CCD. The CCD exists as an antiparallel homodimer. Both GCAP1 and S100B signals are translated at the CCD into faster production of cyclic GMP (Modified from Duda et al., 2012).

contribute to the rod photoresponse at high free [Ca2+] outside the normal physiological range was tested and not found.

Detailed histochemical analysis of S100B expression in the mouse retina brought all these observations into accord (**Figure 3**). As expected, the S100B-KO retina showed no S100B immunoreactivity (Wen et al., 2012), affirming the specificity of the S100B antibody, because photoreceptor outer segments also express S100A1, another member of the S100 family (Rambotti et al., 1999). S100B immunoreactivity was present in wild-type photoreceptors, however, it was not present in all of them. The scattered S100B staining corresponded to that for cone-specific arrestin. On the other hand, in retinas of Nrl−/<sup>−</sup> mice, all outer segments labeled for S100B as well as for cone arrestin. Because Nrl−/<sup>−</sup> retina is rod-less and populated exclusively with cone photoreceptors (Mears et al., 2001) these results strongly indicate that among photoreceptors, S100B expression is restricted to cones. Unimodal switch behavior in ROS-GC1−/<sup>−</sup> outer segments (above) is now easy to understand. In ROS-GC1−/<sup>−</sup> retinas, cones degenerate but the rods remain intact (Yang et al., 1999). Rods however, do not express S100B, so the only Ca2+-regulated guanylate cyclase activity in ROS-GC1−/<sup>−</sup> retinas is the GCAPs modulated ROS-GC2 activity.

Our ongoing immunohistochemical studies with the Nile Rat (*Arvicanthis ansorgei*) retina in which 33% of the photoreceptors are cones, in contrast to the 3% cone population in the mouse (Bobu et al., 2006) appear to corroborate the conclusion derived from analyses of murine retinas that S100B co-exists with ROS-GC1 exclusively in the cone outer segments.

At present, the question about the role of the ROS-GC1 bimodal switch in cone phototransduction is still unanswered because cones from S100B−/<sup>−</sup> mice have yet to be recorded. A model illustrating some of the difficulties in understanding the operation

showing that cone arrestin and S100B labeling are restricted to cones. **(D)** Bright field. OS, outer segments; IS, inner segments; ONL, outer nuclear layer containing nuclei of rods and cones; OPL, outer plexiform of the bimodal switch is shown in **Figure 4**. One issue is that in darkness (DARK), Ca2<sup>+</sup> will not bind to S100B and activate the switch if [Ca2+]i never exceeds 250 nM. However, the value of 250 nM for [Ca2+]i in darkness was taken from mouse rods (Woodruff et al., 2002) because measurements from mouse cones have not yet been made. Thus the true [Ca2+]i for cones in darkness may be higher. A second issue is that without proper restraint, operation of the bimodal switch at high [Ca2+]i would ignite an explosive positive feedback: increased cyclic GMP production, followed by opening of more cyclic GMP-gated channels, which then allow a greater influx of Ca2+, etc. Cones could accumulate lethal levels of Ca2<sup>+</sup> or experience a collapse in ion gradients (DEPOLARIZATION-S100B Mode). Indeed, genetic mutations that result in a hyperactive ROS-GC complex cause the photoreceptors to degenerate (Olshevskaya et al., 2002; Zägel and Koch, 2014).

#### **THE BIMODAL SWITCH EXISTS AT THE PHOTORECEPTOR-BIPOLAR CELL SYNAPSE**

Immunohistochemical studies also place ROS-GC1, GCAP1 and S100B in the synaptic layers of the retina (Liu et al., 1994; Cooper et al., 1995; Duda et al., 2002; Venkataraman et al., 2003). To test whether the ROS-GC1 transduction system utilizes a Ca2<sup>+</sup> bimodal switch at the photoreceptor-bipolar cell synapse (presumably the cone to bipolar synapse, see **Figure 3**), bovine retinal synaptosomes (P1) were prepared and analyzed. Purity of the preparation was ascertained by Western blotting with rhodopsin kinase antibody. No immunoreactivity for rhodopsin kinase was detected, assuring that the preparation was not contaminated with ROSs, the original residential site of the ROS-GC1. Western blots were positivefor ROS-GC1, GCAP1 and S100B demonstrating that indeed the synaptosomal fraction contained all the components

red at high magnification. Conditions were adjusted to optimize the detection of outer segment labeling Courtesy of A. Pertzev, Salus

**Frontiers in Molecular Neuroscience www.frontiersin.org** March 2014 | Volume 7 | Article 21 |

University.

of the ROS-GC1 bimodal system and was suitable for functional analyses.

To be physiologically relevant, the components of the ROS-GC1 bimodal system, ROS-GC1, GCAP1 and S100B must reside in proximity and not, e.g., in different populations of synaptosomes. Because the Western analyses described above did not provide information about the proximity of ROS-GC1, GCAP1 and S100B molecules, immunohistochemical analyses (Duda et al., 2002; Venkataraman et al., 2003) and cross-linking experiments were performed.

Cryosections of bovine retina were doubly labeled with antibodies against GCAP1 or ROS-GC1 and SV2, a synaptic vesicle protein that strongly stains photoreceptor termini (Buckley and Kelly, 1985). The results revealed co-expression of ROS-GC1 and GCAP1 in cone synaptic termini (Venkataraman et al., 2003). Similarly, co-localization of ROS-GC1 and S100B was shown (Duda et al., 2002).

Cross-linking experiments with a non-cleavable imidoester confirmed the proximity of ROS-GC1 and GCAP1 in the synaptosomal fraction (Venkataraman et al., 2003). Without the cross-linker, the ROS-GC1 and GCAP1 antibodies identified separate bands at ∼120 kDa and ∼21 kDa, respectively, corresponding to the appropriate monomers. But in presence of the cross-linker, both GCAP1 and ROS-GC1 antibodies identified a common immunoreactive band at ∼250 kDa, the approximate size for a ROS-GC1 dimer-GCAP1 complex. The results demonstrated that ROS-GC1 and GCAP1, indeed, resided in proximity. Because ROS-GC1 resides with GCAP1 in cone synaptic terminals and ROS-GC1 also resides with S100B, it was inferred that all three components of the bimodal system co-reside in the photoreceptor synaptic terminals, specifically cone synaptic terminals.

Functionality of the ROS-GC1 bimodal switch in the P1 fraction was determined by measuring guanylate cyclase activity at Ca2<sup>+</sup> concentrations ranging from 10 nM to 10 μM. Below 200 nM, Ca2<sup>+</sup> elicited a dose-dependent decrease in guanylate cyclase activity with an "inhibitory" value (IC50) for Ca2<sup>+</sup> of 100 nM. But at higher concentrations, ROS-GC activity began to climb with an EC50 of 0.8 μM. Side by side analysis of photoreceptor outer segment and P1 membranes demonstrated that the cyclic GMP synthetic activities were inhibited identically by free Ca2<sup>+</sup> with an IC50 of 100 nM.

That GCAP1 was involved in the "inhibitory" phase of ROS-GC1 activity was confirmed by peptide competition experiments. A sequence motif L503–I522 in ROS-GC1 is critical and specific for the cyclase activation by GCAP1 (Lange et al., 1999). The core motif within this domain consists of D507–R518. Analysis of the P1 and photoreceptor outer segments membranes in parallel demonstrated that the incremental concentrations of the L503–I522 and D507–R518 peptides at 10 nM Ca2<sup>+</sup> inhibited their guanylate cyclase activities with the same dose-dependence. A control peptide of L503–I522 with a scrambled sequence had no effect on GCAP-dependent ROS–GC1 activity. Both fractions showed a reversal of the peptide (D507–R518)-dependent inhibition upon the addition of an excess of GCAP1 (Venkataraman et al., 2003). Thus Ca2<sup>+</sup> signaling in the synapse region is modulated by GCAP1 and S100B. GCAP1 has the higher affinity for Ca2<sup>+</sup> and upon binding, lowers ROS-GC1 activity. S100B captures the high Ca2<sup>+</sup> signal and upon binding, stimulates ROS-GC1.

#### **A ROLE IN SIGNAL PROCESSING IN THE RETINA?**

ERG recordings from wild-type and S100B−/<sup>−</sup> mice provided the first clues as to the physiological significance of S100B in retinal function (Wen et al., 2012). The ERG is a field potential across the retina generated by the electrical activity of all cells, including that of rods and cones (reviewed in: Peachey and Ball, 2003; Weymouth and Vingrys, 2008). The corneal negative a-wave, which sums the photocurrent responses of rods, is normal in the S100B-KO mice because S100B is not present in rods. The corneal positive b-wave, generated mainly by ON-bipolar cells, displays slower kinetics in S100B−/<sup>−</sup> mice; it peaks later and takes longer to recover. Oscillatory potentials that arise from the activity of inner retinal neurons are smaller and there are disturbances in their timing. Gaps in our knowledge about what S100B is doing in the cone outer segment, the cone synapses and for that matter, in other retinal neurons preclude any attempt to interpret the changes in the ERG at this time.

#### **FUTURE DIRECTIONS**

The composition and the operational principles of cones and rods differ in ways that support their specialized visual functions: extremely high sensitivity in rods, rapid responsiveness over a wide dynamic range in cones. In addition to the GCAPs, S100B serves as an additional Ca2<sup>+</sup> sensor component of ROS-GC1 in cones. Future tasks will be to decode this signal transduction system at the molecular and physiological levels, to link it with cone-related dystrophies and to design therapeutic options. Progress would be facilitated if several important objectives were to be met. (1) A suitable mammalian model with abundant cones needs to be selected. No such mouse model is available at present. Through genetic manipulation, development of the rods has been suppressed in Nrl−/<sup>−</sup> mice and their retinas are cone-exclusive. But these cones may not be exact copies of the natural, although biochemically and physiologically, they appear to be very close (Nikonov et al., 2005, 2006). One alternative is the Nile rat (*A. ansorgei*), whose retina contains about 30% cones in contrast to the 3% population in the mouse (Bobu et al., 2006). (2) The molecular architecture of the ROS-GC complex needs to be understood. Which

combinations of Ca2<sup>+</sup> sensor proteins are possible in the complex? Which actually occur? How are their relative levels within different cell types regulated? (3) Ca2+-modulated signal transduction by the ROS-GC complex needs to be defined at a molecular level. For the S100B native to cones, what are the kinetic parameters of the S100B transduction switch? By what mechanism does S100B sense Ca2<sup>+</sup> and how does it effect a change in the catalytic domain of ROS-GC? (4) The physiological role of S100B in the cone photoresponse needs to be assessed. Does it accelerate the recovery of cone response and increase the dark current? What limits S100B stimulated ROS-GC activity at high Ca2+? Does S100B increase the rate of vesicular release of neurotransmitter at the synapse? S100A1 is also expressed in outer segments, what is its function? (5) S100B is expressed in inner retinal neurons but nothing is known about what it does there. (6) S100B protein does not fit in the presently classified NCS protein family, sharing only 15% sequence identity and containing only two EF-hands. Therefore, the NCS family needs to be expanded to include S100B as a branch.

#### **ACKNOWLEDGMENTS**

The authors gratefully acknowledge the continuous past support by the numerous USPHS award from the National Institutes of Health, the beginning award from the National Science Foundation and the Damon Runyon Walter Winchell Cancer Fund and the Howe Laboratory Endowment of the Massachusetts Eye and Ear Infirmary.

#### **REFERENCES**


Stryer, L. (1991). Visual excitation and recovery. *J. Biol. Chem.* 266, 10711–10714.


cones but not in rods of the murine retina. *Cell. Physiol. Biochem.* 29, 417–430. doi: 10.1159/000338496


**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: 20 December 2013; accepted: 05 March 2014; published online: 26 March 2014.*

*Citation: Sharma RK, Makino CL, Hicks D and Duda T (2014) ROS-GC interlocked Ca2+-sensor S100B protein signaling in cone photoreceptors: review. Front. Mol. Neurosci. 7:21. doi: 10.3389/fnmol.2014.00021*

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

*Copyright © 2014 Sharma, Makino, Hicks and Duda. 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 andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Ca**2+**-modulated ROS-GC1 transduction system in testes and its presence in the spermatogenic cells

## *Anna Jankowska1\*, Rameshwar K. Sharma2 and Teresa Duda2 \**

<sup>1</sup> The Unit of Molecular Biology, Department of Cell Biology, Poznan University of Medical Sciences, Poznan, Poland

<sup>2</sup> The Unit of Regulatory and Molecular Biology, Research Divisions of Biochemistry and Molecular Biology, Salus University, PA, USA

#### *Edited by:*

Clint Lawrence Makino, Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA

#### *Reviewed by:*

Rosario Donato, University of Perugia, Italy Andrzej Lukaszyk, Poznan University

of Medical Sciences, Poland

#### *\*Correspondence:*

Anna Jankowska, The Unit of Molecular Biology, Department of Cell Biology, Poznan University of Medical Sciences, 5D Rokietnicka Street, 60-781 Poznan, Poland e-mail: ajanko@ump.edu.pl; Teresa Duda, The Unit of Regulatory and Molecular Biology, Research Divisions of Biochemistry and Molecular Biology, Salus University, 8360 Old York Road, Elkins Park, PA 19027, USA e-mail: tduda@salus.edu

ROS-GC1 belongs to the Ca2+-modulated sub-family of membrane guanylate cyclases. It primarily exists and is linked with signaling of the sensory neurons – sight, smell, taste, and pinealocytes. Exceptionally, it is also present and is Ca2+-modulated in t he non-neuronal cells, the sperm cells in the testes, where S100B protein serves as its Ca2<sup>+</sup> sensor. The present report demonstrates the identification of an additional Ca2<sup>+</sup> sensor of ROS-GC1 in the testes, neurocalcin δ. Through mouse molecular genetic models, it compares and quantifies the relative input of the S100B and neurocalcin δ in regulating the Ca2<sup>+</sup> signaling of ROS-GC1 transduction machinery, and via immunochemistry it demonstrates the copresence of neurocalcin δ and ROS-GC1 in the spermatogenic cells of the testes. The suggestion is that in more ways than one the Ca2+-modulated ROS-GC1 transduction system is linked with the testicular function. This non-neuronal transduction system may represent an illustration of the ROS-GC1 expanding role in the trans-signaling of the neural and non-neural systems.

**Keywords: testes, neurocalcin δ, membrane guanylate cyclase, ROS-GC1, calcium ions**

#### **INTRODUCTION**

Rod outer segment membrane guanylate cyclase, ROS-GC1 (known also as Ret-GC1 or GC-E), belongs to the family of membrane guanylate cyclases. Its discovery and molecular characterization was a land mark event in the field of phototransduction (reviewed in Pugh et al., 1997; Koch et al., 2010) as it identified the source of cyclic GMP that serves as a second messenger of the LIGHT signal. It also impacted the entire membrane guanylate cyclase field by dividing the guanylate cyclase family into two subfamilies, one comprising the hormone receptor cyclases, and the other, cyclases modulated by intracellular [Ca2+]*<sup>i</sup>* signals (reviewed in: Sharma, 2010). ROS-GC1 belongs to the second subfamily together with another rod outer segment guanylate cyclase ROS-GC2 (Ret-GC2 or GC-F), and the olfactory neuroepithelial guanylate cyclase, ONE-GC (GC-D; reviewed in: Sharma, 2010).

The best-documented physiological function of ROS-GC1 and ROS-GC2 is in the recovery phase of phototransduction, to return the illuminated photoreceptors to the dark, resting state. Illumination of photoreceptors leads to activation of cyclic GMP phosphodiesterase, depletion of cyclic GMP, closure of the cyclic GMP gated (CNG) channels, lowering the free Ca2<sup>+</sup> concentration, and hyperpolarization of the plasma membrane (reviewed in: Pugh et al., 1997; Koch et al., 2010). The ROS-GCs task is to restore the dark-level of cyclic GMP allowing opening of the CNG channels, increase of Ca2<sup>+</sup> influx, and depolarization of plasma membrane. Ca2<sup>+</sup> concentration, thus, determines

the activities of ROS-GCs but in an indirect way. Guanylate cyclase activating proteins (GCAPs; GCAP1 and GCAP2) sense the post-illumination fall in Ca2<sup>+</sup> and stimulate ROS-GCs to synthesize cyclic GMP at a faster rate and restore its dark level (reviewed in: Detwiler, 2000; Koch et al., 2010).

GCAPs, however, are not the only Ca2<sup>+</sup> sensing modulators of ROS-GC activity. While increasing Ca2<sup>+</sup> concentrations inhibit ROS-GCs activity through GCAPs, two other Ca2<sup>+</sup> sensors, S100B and neurocalcin δ stimulate ROS-GC1 in a Ca2+-dependent fashion (Pozdnyakov et al., 1995; Margulis et al., 1996; Duda et al., 1996a; Kumar et al., 1999). The Ca2+-dependent S100B-mediated activation of ROS-GC1 operates in cones including their outer segments and pedicles (Duda et al., 2002; Wen et al., 2012). Its role in photo- and visual transductions remains to be established, but existing data indicate its involvement in transmission of the visual signal from cone ON-bipolar cells (Wen et al., 2012). Ca2<sup>+</sup> signaling of ROS-GC1 activity mediated by neurocalcin δ is operative in retinal ganglion cells (Krishnan et al., 2004); and of ONE-GC in the olfactory neuroepithelium (Duda et al., 2001a, 2004).

Beyond the retina, ROS-GC1 is expressed in the pineal gland where in one subset of pinealocytes it co-localizes with GCAP1 and in another, with S100B (Venkataraman et al., 2000); in the mitral cells of the olfactory bulb where it co-localizes with GCAP1 (Duda et al., 2001b); and it co-immunoprecipitates with S100B in the gustatory epithelium (Duda and Sharma, 2004).

Neurocalcin δ (NCδ) belongs to a subfamily of neuronal calcium sensor (NCS) proteins called visinin like proteins (VSNLs). Similar to other, but not all NCS proteins, it is acylated at the N-terminus by myristic acid and undergoes a classical calcium-myristoyl switch (Ladant, 1995) e.g., it buries the myristoyl group in a hydrophobic pocket in a Ca2<sup>+</sup> free form and exposes it in Ca2+-bound form, a phenomenon first observed for recoverin (Zozulya and Stryer, 1992). Myristoylation of neurocalcin δ allows its membrane association in response to changes in the intracellular Ca2<sup>+</sup> concentration. However, once NCδ binds to the cellular membranes in a Ca2+-dependent fashion, part of it remains membrane associated even after removing Ca2<sup>+</sup> by the addition of ethylene glycol tetraacetic acid (EGTA; Krishnan et al., 2004). Although the highest level of NCδ has been detected in neuronal tissues, its expression in the periphery is also observed. Functionally, NCδ has been linked to a wide variety of processes such as receptor endocytosis through interaction with α- and β-clathrin and β-adaptin (Ivings et al., 2002), trafficking and membrane delivery of glutamate receptors of the kainate type (Coussen and Mulle, 2006), and with microtubule assembly (Iino et al., 1995). The presence of NCδ has been found in the inner plexiform layer of the retina, e.g., in the amacrine and ganglion cells (Krishnan et al., 2004), olfactory sensory neurons (Duda et al., 2001b, 2004) and in type II cells of mouse circumvallate taste papillae (Rebello et al., 2011).

In the inner retinal layer NCδ acts as Ca2+-dependent modulator of membrane guanylate cyclase ROS-GC1 (Krishnan et al., 2004) and in the olfactory neuroepithelium, of ONE-GC (Duda et al., 2001a, 2004). The exact physiological significance of the ROS-GC1-NCδ signaling system in the retina is not known yet, it has, however, been proposed that the system may be involved in synaptic processes (Krishnan et al., 2004). In the olfactory neuroepithelium neurocalcin δ has been proposed as a Ca2<sup>+</sup> sensor component of the two-step odorant uroguanylin signaling machinery (Duda and Sharma, 2009).

Outside the neuronal system, NCδ is expressed in the adrenal glomerulosa cells (Duda et al., 2012a,b). There it co-localizes with the member of receptor guanylate cyclase sub-family, atrial natriuretic factor receptor guanylate cyclase (ANF-RGC), and has been proposed to be involved in the inhibition of aldosterone synthesis (Duda et al., 2012a,b). Accordingly, a mouse model in which one copy of NCδ gene is deleted is inflicted with hyperaldosteronism and hypertension (Duda et al., 2012b).

Preliminary studies localize also ROS-GC1 outside the neuronal system, in male gonads, where it co-localizes with GCAP1 and S100B (Jankowska et al., 2007). It remains to be determined, whether, and how ROS-GC1 and its companion calcium binding proteins are involved in the physiology of the testis. Considering, however, that cyclic GMP and Ca2<sup>+</sup> are genuine mediators of signal transduction in the testes in the physiology of fertilization (reviewed in: Garbers, 1989; Jankowska and Warchol, 2010) it is highly possible that ROS-GC1, stimulated or inhibited as needed, in a Ca2+-dependent fashion by Ca2+-binding proteins, is the provider of the physiologically necessary quantities of cyclic GMP. We now present new observations on co-presence of NCδ and ROS-GC1 in the spermatogenic cells and Ca2+-dependent modes of NCδ and S100B in modulating ROS-GC1 signaling the mammalian testes.

## **MATERIALS AND METHODS**

#### **TISSUES**

The study group consists of testes from human (obtained from consenting organ donors, *n* = 5), bovine (purchased from local slaughter house, *n* = 3), and Wistar rat (obtained form the Vivarium of Poznan University of Medical Sciences, Poznan, Poland, *n* = 12). The study was approved by the IACUC and Ethics Review Boards of all the involved Institutions.

#### **GENETICALLY MODIFIED MICE**

Care of the experimental animals conformed to the protocols approved by the IACUC at Salus University and was in strict compliance with the NIH guidelines. Construction of heterozygous NCδ-KO (NCδ+/−) mice is described in (Duda et al., 2012b). The S100B-KO mice are described in (Wen et al., 2012).

#### **ANTIBODIES**

Rabbit ROS-GC1 and NCδ antibodies were produced, characterized and affinity purified as in (Venkataraman et al., 2003; Krishnan et al., 2004). Secondary antibodies, AP-conjugated goat anti-rabbit IgG, preimmune rabbit serum, NBT/BCIP, and Cy3-conjugated sheep anti-rabbit IgG were purchased from Sigma-Aldrich.

#### **REVERSE TRANSCRIPTION POLYMERASE CHAIN REACTION (RT-PCR)**

Total RNA was isolated from seminiferous tubules and interstitial tissue using TriPure isolation reagent (Roche Diagnostics), according to the manufacturer's protocols. The cDNA library was constructed using Advantage RT for PCR kit (BD-Bioscience) and used for the amplification of ROS-GC1 and neurocalcin δ fragments. Sequences of primers used for the amplification are listed in **Table 1**. The amplified fragments were purified on agarose gel and sequenced to confirm their identities.

### **WESTERN BLOTTING**

The procedure was carried out according to the previously published protocols (Venkataraman et al., 2000; Duda and Sharma, 2004; Jankowska et al., 2007). 150 mg of membrane fraction proteins were denatured in gel-loading buffer [62.5 mM Tris-HCl (pH 7.5), 2% SDS, 5% glycerol, 1 mM β-mercaptoethanol, and 0.005% bromophenol blue] in 95◦C for 10 min. Samples were then subjected to SDS-PAGE in a buffer containing 0.025 mM Tris-HCl (pH 8.3), 0.192 M glycine, and 0.1% SDS. Afterwards the resolved proteins were transferred to nitrocellulose membranes. To avoid non-specific interactions membranes were incubated in Tris buffered saline containing 0.05% Tween 20 (TBS-T), 5% non-fat milk (blocking buffer) overnight at 4◦C. ROS-GC1 and NCδ proteins were detected with specific rabbit polyclonal antibodies diluted 1:1500 and 1:1000, respectively. After 1 h incubation the blot was rinsed three times with TBS-T and incubated with anti-rabbit secondary antibodies conjugated with horseradish peroxidase (1:10 000).

To visualize the immunoreactive bands SuperSignal blaze chemiluminescent substrate (Pierce) was used according to the manufacturer's protocol. Signal was detected by exposing the blot



to Kodak X-ray film for 15 s. Then the X-ray film was scanned and processed using Photoshop 6.0 software.

#### **IMMUNOHISTOCHEMISTRY**

Paraffin sections of the testes fixed in 4% paraformaldehyde were used for immunohistochemical detecting of ROS-GC1 and NCδ. To block the non-specific binding the sections were first washed with TBS (100 mM Tris-HCl, 0.9% NaCl) and incubated in blocking solution consisting of TBS buffer containing 0.05% Tween 20 (TBS-T) and 1% BSA for 1 h at room temperature. After washing with TBS-T the sections were incubated with the respective primary antibodies for 60 min at 37◦C and washed four times (15 min each) in TBS-T. Anti ROS-GC1 antibodies were diluted 1:200 and anti neurocalcin δ, 1:100. AP-conjugated anti-rabbit IgG, diluted 1:200 and NBT/BCIP as the substrate were used for detection. Controls included detection reactions carried out under identical conditions except that the primary antibodies were replaced by preimmune serum.

#### **ISOLATION OF THE PARTICULATE FRACTION OF THE TESTES**

Membrane fraction of the testes was isolated according to the protocol described previously (Marala and Sharma, 1988). The tissue was homogenized in a buffer containing 250 mM sucrose, 10 mM Tris-HCl (pH 7.4) and protease inhibitors. The homogenate was centrifuged at 400 *g* and then the supernatant at 10,000 *g* and finally, at 40,000 *g.* The resulting pellet represented the membrane fraction.

#### **GUANYLATE CYCLASE ACTIVITY ASSAY**

The particulate fraction was assayed for guanylate cyclase activity as described previously (Paul et al., 1987; Jankowska et al., 2007, 2008). Briefly, membranes were preincubated on an ice-bath with or without NCδ in an assay system containing 10 mM theophylline, 15 mM phosphocreatine, 20 μg creatine kinase, and 50 mM Tris-HCl (pH 7.5), adjusted to the appropriate free Ca2<sup>+</sup> concentration with precalibrated Ca2/EGTA solutions (Molecular Probes). The reaction was initiated by the addition of a substrate solution (4 mM MgCl2 and 1 mM GTP, final concentration) and continued by

incubation at 37◦C for 10 min. The reaction was terminated by the addition of 50 mM sodium acetate (pH 6.2) followed by heating in a boiling water-bath for 3 min. The amount of cyclic GMP formed was determined by radioimmunoassay (Nambi et al., 1982; Jankowska et al., 2007, 2008).

#### **STATISTICAL ANALYSES**

The activity of guanylate cyclase was calculated as mean of at least six separate values ± SD.

#### **RESULTS**

General consensus has been that ROS-GC1 is expressed exclusively in the sensory as well as in the second order neurons of the retina and in the neurons of the pineal gland and the olfactory bulb (Hayashi and Yamazaki, 1991; Goraczniak et al., 1994; Liu et al., 1994; Yang et al., 1995; Venkataraman et al., 2000; Duda et al., 2001a). However, earlier results of these investigators provided the first indication that ROS-GC1 is also expressed outside the neuronal system, in bovine testes and sperm (Jankowska et al., 2007, 2008). In the sperm, it is co-expressed with calcium sensor S100B and neuronal calcium sensor proteins GCAP1 and NCδ (Jankowska et al., 2008). These results were striking enough to warrant more detailed studies. Present investigation represents a step in that direction.

#### **MAMMALIAN TESTES EXPRESS ROS-GC1 AND NEUROCALCIN δ: ANALYSES AT THE mRNA LEVELS**

The presence of ROS-GC1 and NCδ transcripts was analyzed in human, bovine, and rat testes. Using total RNA isolated from these tissues individual cDNA libraries were constructed and used for amplification of specific fragments of ROS-GC1 and NCδ cDNAs by polymerase chain reaction. For the amplification, distinct, species-specific primers for ROS-GC1 or NCδ were designed. These primers corresponded to the sequences located within so called "kinase-like domain" of ROS-GC1 cDNA and for NCδ they overlapped with the 5 - and 3 - ends of its coding sequence. The amplification yielded fragments of 553, 548, and 566 bp of the human, bovine, and rat ROS-GC1, respectively, and 526 bp fragment of neurocalcin δ (**Table 1**). Sequencing of all amplified products confirmed their identities with ROS-GC1 or NCδ cDNAs. The ROS-GC1 fragments represented indeed sequence coding for the kinase-like domain, and the NCδ fragment had sequence identical with the predicted part of NCδ coding region. Amplification of uninterrupted coding sequences for ROS-GC1 and NCδ documented that the RNA used was not contaminated with genomic DNA. Thus both, ROS-GC1 and NCδ transcripts are present in human, bovine, and rat testes.

#### **MAMMALIAN TESTES EXPRESS ROS-GC1 AND NEUROCALCIN δ: ANALYSES AT THE PROTEIN LEVELS**

At the protein level, the expression of ROS-GC1 and NCδ in mammalian testes was tested in two ways. First, a rudimentary assessment of the expression of both types of proteins was obtained by Western blotting; this was followed by immunocytochemical localization of ROS-GC1 and NCδ in human and rat testes. In both types of analyses appropriate affinity purified antibodies, anti-ROS-GC1, or anti-NCδ were used. Immunostaining of bovine testis was performed in parallel as positive control.

After establishing that ROS-GC1 and NCδ transcripts are present in mammalian testes, their protein identity was scrutinized by Western blotting. The mobility of the ROS-GC1 or NCδ immunoreactive band, ∼116 kDa and ∼20 kDa, respectively, (**Figures 1A,B**) was identical to that observed previously for ROS-GC1 expressed in photoreceptor outer segments (Duda et al., 1996b) and NCδ expressed in the inner retina (Krishnan et al., 2004). These results clearly confirm that both proteins, ROS-GC1 and NCδ are expressed in human, rat, and bovine testes.

Detailed analyses of the localization of ROS-GC1 and NCδ in the mammalian testes were performed by immunocytochemistry. In human and rat testes ROS-GC1 immunostaining was randomly distributed in the seminiferous tubules (**Figures 2A,B**). The pattern of this staining was identical to that observed for bovine testes (**Figure 2C**). Visual analysis allowed localizing the staining to spermatogenic cells, especially to primary spermatocytes and spermatids. The majority of primary spermatocytes (**Figure 2**; indicated as Sc), spermatids (**Figure 2**; indicated as Sd) and single spermatogonias (**Figure 2**; indicated as Sg) were immunostained in all species tested, human, rat, and bovine. To verify specificity of the ROS-GC1 immunostaining a control reaction was performed in which ROS-GC1 antibody was omitted and substituted with the preimmune serum. Under these conditions no labeling was observed (**Figures 2A'–C'**).

In a similar manner the expression of NCδ in the human and rat testes was analyzed (**Figures 3A,B**) and compared with that in bovine testis (**Figure 3C**). NCδ immunoreactivity was observed in all types of germinal cells: spermatogonias (**Figure 3**: Sg), spermatocytes (**Figure 3**: Sc) and spermatids (**Figure 3**: Sd). Extremely strong staining was observed in spermatids localized close to seminiferous tubule lumen. No labeling was observed when the sections were incubated with preimmune rabbit serum instead of the NCδ antibody, attesting to the specificity of the staining (**Figures 3A'–C'**).

The results presented clearly demonstrate that both ROS-GC1 and NCδ are expressed in the mammalian testes and in the germinal cells. Importantly, the patterns of ROS-GC1 and NCδ immunoreactivities are identical in the species analyzed (compare **Figures 2** and **3**) strongly indicating that they co-localize with each other. Notably, the intensities of the immunostaining for ROS-GC1 or NCδ increase with the maturation of the spermatogenic cells.

Since the increase in the staining is observed across the cells in different stages of spermatogenesis process – the highest accumulation of analyzed proteins was observed in spermatocytes and spermatids – it indicates that ROS-GC1 and NCδ expression correlates with the stage of the seminiferous cycle.

#### **NEUROCALCIN δ STIMULATES ROS-GC1 CATALYTIC ACTIVITY EXPRESSED IN THE TESTES: EVIDENCE FROM RECONSTITUTION EXPERIMENTS**

To demonstrate that neurocalcin δ stimulates ROS-GC1 expressed in the testes, their membranes were analyzed for guanylate cyclase activity in the presence of NCδ. Membranes of human, bovine, and rat testes were isolated and incubated with increasing concentrations of NCδ and constant 100 μM Ca2+. The amount of cyclic GMP formed was determined as a measure of guanylate cyclase activity. NCδ stimulated the cyclase activity in a dosedependent fashion (**Figure 4**). Half-maximal stimulation (EC50) was observed at ∼0.8 μM NCδ in all three species tested. The maximal stimulation of ∼2.2-fold above the basal value was at 2 μM NCδ. Importantly, the stimulatory profiles of guanylate cyclase in the analyzed membranes were very similar to those

observed for the recombinant ROS-GC1 (Kumar et al., 1999) and ROS-GC1 native to the retinal photoreceptor outer segments and inner plexiform layer (Krishnan et al., 2004). These results show that NCδ indeed stimulates ROS-GC1 activity in the testes.

#### **ROS-GC1 AND NCδ ARE FUNCTIONALLY LINKED: EVIDENCE FROM GENETICALLY MODIFIED MICE**

Because the experiments described above were conducted on the native membranes of the testes but with exogenously supplied NCδ, they prove the ability of ROS-GC1 to respond to NCδ but they do not prove that these two together form a functional unit. To determine this, the mouse model with deletion of one copy of the NCδ gene, NCδ+/−, was used [mice with deletion of both copies of NCδ are not born (Duda et al., 2012b)]. It was reasoned that if NCδ is indeed the Ca2+-sensor modulator of ROS-GC1 in the testes of these mice the NCδ-modulated Ca2<sup>+</sup> signaling pathway should be half as active as in those of the wild type mice [wt (NCδ+/+)]. To test this prediction, the particulate fractions of the testes from wild type and NCδ+/<sup>−</sup> mice were isolated in the presence of 100 μM Ca2<sup>+</sup> and tested for guanylate cyclase activity in the presence and absence of Ca2+. The reason for isolating the membranes in the presence of Ca2<sup>+</sup> was that NCδ

the testes were incubated with primary antibodies against NCδ. NCδ immunoreactivity was observed in all spermatogenic cells of human **(A)**, rat **(B)**, and bovine **(C)** testes. Particularly strong signal was detected in spermatocytes and spermatids localized close to lumen of the seminiferous tubule (indicated as Sc and Sd, respectively). Control staining was performed according to the procedure for NCδ detection, except that preimmune serum was used instead of the primary antibody **(A'–C')**. Original magnification 400X.

exhibits the property of "calcium-myristoyl switch" which means that in the presence of Ca2<sup>+</sup> it is membrane bound and therefore would be present in the membrane fraction (Ladant, 1995; Krishnan et al., 2004). The guanylate cyclase activity assayed in the absence of Ca2<sup>+</sup> (1 mM EGTA was present in the assay mixture) was 58 <sup>±</sup> 7 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the wt and 63 <sup>±</sup> 8 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the NCδ+/<sup>−</sup> mice (**Figure 5**: panel "Ca2+"). However, the cyclase activity determined in the presence of Ca2<sup>+</sup> was strongly dependent on the mice genotype. With 1 μM Ca2<sup>+</sup> present in the assay mixture the activity was 248 <sup>±</sup> 17 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the wild type mice and 148 <sup>±</sup> 16 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the NCδ+/<sup>−</sup> mice (**Figure 5**: panel "+Ca2<sup>+</sup> – basal"). These results, as predicted, demonstrate that the Ca2+-dependent NCδ-modulated ROS-GC1 signaling pathway in the mice with one copy of NCδ gene deleted (NCδ+/−) is about half as active as in the wild type mice. To further validate that the lowering of the Ca2+-dependent cyclase activity in the membranes of NCδ+/<sup>−</sup> mice is the exclusive consequence of lower NCδ expression, 4 μM exogenous NCδ was added to the wild type and NCδ+/<sup>−</sup> membranes and the cyclase activity was assessed in the presence of 1 μM Ca2+. The cyclase activity in the wild type membranes increased only minimally, from 248 to 273 ± 25 pmol cyclic GMP

Original magnification 400X.

min−<sup>1</sup> (mg prot)−1but in the NCδ+/<sup>−</sup> membranes, from 148 to 258 <sup>±</sup> 28 (**Figure 5**: panel "+Ca2<sup>+</sup> <sup>+</sup> <sup>4</sup> <sup>μ</sup>M NCδ"). Thus, the activity achieved was practically the same for both types of membranes. These results demonstrate that addition of exogenous NCδ to the NCδ+/<sup>−</sup> membranes restores the guanylate cyclase catalytic activity and brings it to the level of activity in the wild type membranes. The slight activity increase in the wild type membranes can be explained by a partial loss of the native NCδ during the membrane preparation procedure. Together these results demonstrate that functional ROS-GC1-NCδ-Ca2<sup>+</sup> transduction system is operative and is the functional transduction element in the testes.

#### **WHICH CA2+ SENSOR, NCδ OR S100B, IS THE MORE POTENT REGULATOR OF ROS-GC1 ACTIVITY?**

Our previous investigation has shown that S100B is expressed in the testes (Jankowska et al., 2007) and the present results demonstrate that NCδ is also expressed there. Thus, two Ca2+-dependent modulators of ROS-GC1 activity are present in the testes. Taking advantage of the availability of two types of genetically modified mice, NCδ+/<sup>−</sup> and S100B−/<sup>−</sup> we attempted comparing the relative inputs of NCδ and S100B into regulation of ROS-GC1 transduction machinery.

Testes were removed from the wild type, NCδ+/<sup>−</sup> and S100B−/<sup>−</sup> mice, their membrane fractions were isolated in the presence of 100 μM Ca2<sup>+</sup> and assessed for guanylate cyclase activity. The results of this experiment are shown in **Figure 6**. In the absence of Ca2<sup>+</sup> all three types of membranes exhibited comparable activity of <sup>∼</sup>60 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup>

(**Figure 6A**: panel "basal") and, as expected, the activity was not affected by the addition of NCδ or S100B (**Figure 6A**: panels "+ncalc δ" and "+S100B"). However, when the cyclase activity was assayed in the presence of Ca2<sup>+</sup> the values differed between the genotypes: 250 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the wild type, 150 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the NCδ+/<sup>−</sup> and 200 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the S100B−/<sup>−</sup> (**Figure 6B**: panel "basal"). These values describe the ROS-GC1 activity stimulated together by NCδ and S100B in the case of the wild type membranes; stimulated by half of wild type amount of NCδ and S100B in the case of NCδ+/<sup>−</sup> membranes; and stimulated by NCδ only in the case of S100B−/<sup>−</sup> membranes. Thus, the difference between ROS-GC1 activity in testes of the wild type mouse and the activity in the genetically modified mice (100 and 50 pmol pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> for the NCδ+/<sup>−</sup> and S100B−/−, respectively) describes the Ca2+-dependent stimulatory effects lost due to the absence of half of the NCδ present in the wild type testes or to the total absence of S100B. Higher loss of activity resulting from deletion of one copy of NCδ gene than resulting from deletion of two copies of S100B gene indicates that NCδ is more potent than S100B in modulating the Ca2+-dependent ROS-GC1 activity. One could argue, however, that this comparison might not be totally reflective of the *in vivo* contributions of NCδ and/or S100B, because there is a possibility that some of these proteins were lost during the membrane preparation process. To settle this possibility the ROS-GC1 activity was measured in the presence of individually added NCδ or S100B.

Addition of 4 μM NCδ (**Figure 6B**: panel "+ncalc δ") resulted in a small increase of the cyclase activity exhibited by the membranes isolated from the wild type testes (from ∼250 to ∼270 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> and isolated from the S100B−/<sup>−</sup>

**FIGURE 5 | NCδ modulates ROS-GC1 activity in mouse testes.** Testes were removed from the wild type (WT) and NCδ+/<sup>−</sup> mice and their particulate fractions were isolated in the presence of 100 μM Ca2+. These were assayed for guanylate cyclase activity in the presence of 1 mM EGTA or 10 <sup>μ</sup>M Ca2<sup>+</sup> (panel "−Ca2+"- and "+Ca2+"-basal). 4 <sup>μ</sup>M myristoylated NCδ was added to the membranes and the guanylate cyclase activity was assessed in the presence of 10 μM Ca2+. The experiment was done in triplicate and repeated with separate membrane preparations. The results are mean ± SD of these experiments.

testes [from <sup>∼</sup>200 to <sup>∼</sup>220 pmol cyclic GMP min−<sup>1</sup> (mg prot)−1]. This non-significant to minimal increase of ROS-GC1 catalytic activity in these membranes indicates that if any NCδ was lost during the membrane preparation, it was only the insignificant amount. For the NCδ+/<sup>−</sup> membranes the increase was significant, from <sup>∼</sup>150 to <sup>∼</sup>260 pmol cyclic GMP min−<sup>1</sup> (mg prot)−1. Importantly, addition of 4μM NCδ brought the ROS-GC1 activity in these membranes very close to the ROS-GC1 activity in the wild type membranes. Because only residual amount of NCδ was lost during membrane preparation, as concluded from the wild type and S100B−/<sup>−</sup> membranes, practically all the increase of activity in the NCδ+/<sup>−</sup> membranes accounts for the activity lost due to the expression of only half of the NCδ present in the membranes from wild type testes.

When the ROS-GC1 activity was assayed with 4 μM S100B added to the membranes (**Figure 6B**: panel"+S100B") it increased by <sup>∼</sup>40 pmol cyclic GMP min−<sup>1</sup> (mg prot)−<sup>1</sup> in the wild type and by 25 pmol cyclic GMP min−<sup>1</sup> (mg prot)−1in the NCδ+/<sup>−</sup> membranes indicating that some S100B was lost in membranes preparation. The increase was <sup>∼</sup>80 pmol cyclic GMP min−<sup>1</sup> (mg prot)−1for the S100B−/<sup>−</sup> membranes [from 205 to 282 pmol cyclic GMP min−<sup>1</sup> (mg prot)−1] and this increase of activity in the S100B−/<sup>−</sup> membranes accounts for the activity lost due to the loss of S100B expression in the testes. Comparing the activity loss because of the absence of half of NCδ with the activity loss due to the total absence of S100B expression shows that neurocalcin δ is more significant contributor to ROS-GC1 Ca2+-dependent activity than S100B. It could be estimated that NCδ contributes about 75% to ROS-GC1 Ca2+-dependent activity and S100B, about 25%.

#### **DISCUSSION**

The core finding of this study is that it establishes the presence and molecular nature of a Ca2+- modulated ROS-GC transduction system in all types of the germinal cells of the testes: spermatogonias, spermatocytes, and spermatids. Because abundance of this transduction system is present in the spermatids residing close to the seminiferous tubule lumen, it hints that the system is linked with the processes of spermatogenesis.

This Ca2+-dependent cyclic GMP generating transduction system is composed of a transducer component ROS-GC1 and two Ca2<sup>+</sup> sensor components, NCδ and S100B. NCδ is the major signal contributor responsible for about 75% of the guanylate cyclase stimulated catalytic activity, and S100B the relatively minor contributor, about 25% of signaling activity.

Cyclic GMP appears to be a critical second messenger in the physiology of the testes. In particular, it has been shown to influence motility in spermatozoa, development of testicular germ cells, relaxation of peritubular lamina propia cells, testosterone synthesis in Leydig cells, and dilatation of testicular blood vessels (Singh et al., 1988; Armstrong et al., 1994; Middendorff et al., 1997, 2000). Our immunostaining results demonstrate that both ROS-GC1 and NCδ immunoreactivity is present in all germ cells but its intensity is lowest in spermatogonias and highest in spermatids, hence, it increases with maturation, indicating that development of these cells may be directed by the activity of ROS-GC1.

The present study, of basic biochemical nature, brings to the fore several issues for future research. (1) What is the exact composition of the transduction system: does it represent a single pathway, composed of one ROS-GC1 and two Ca2<sup>+</sup> sensors, NCδ, and S100B operating simultaneously? Or, two independent, but complementary pathways, each composed of one ROS-GC1 and one Ca2<sup>+</sup> sensor, NCδ, or S100B? (2) What is the precise morphological residence of any of these transduction systems in the testicular cells? (3) What is the exact stoichiometry of these components in the given cell types? (4) What are the mechanistic details of the operational modes of this/these ROS-GC transduction pathways? (5) What is the physiological significance of these pathways to the testicular function, endocrine, fertility, including spermatogenesis; (6) what is the interplay between the Ca2+-modulated ROS-GC1 signaling pathway and ANF-dependent ANF-RGC pathway, another cyclic GMP generating machinery functioning in the mammalian testes (Marala and Sharma, 1988; Pandey and Singh, 1990; Pandey et al., 1999; Mourdjeva et al., 2001) and finally, (7) how this information can be translated to the clinical level to search for the pathway/s-linked diseases and finding their cures?

Handicapped with these informational gaps, it is difficult to construct a proper Ca2+-modulated signal transduction model for any of the specific physiological function/s of the testes. However, a preliminary signal transduction scheme is conceptualized where [Ca2+]*<sup>i</sup>* modulation of ROS-GC1 activity can occur through three modes. MODE 1 involves interaction of ROS-GC1 with its Ca2<sup>+</sup> sensors, NCδ and S100B. The ROS-GC is stimulated in [Ca2+]*<sup>i</sup>* – dependent manner and generates cyclic GMP. MODE 2 involves interaction of the ROS-GC1 with GCAP1 which is also present in the testes (Jankowska et al., 2007). It is recalled that GCAP1 signaling is opposite to that of NCδ or S100B; it stimulates ROS-GC1 catalytic activity in the absence of Ca2<sup>+</sup> and progressively inhibits it with increasing Ca2<sup>+</sup> concentration (Duda et al., 1996b). Thus, through GCAP1 [Ca2+]*<sup>i</sup>* keeps the ROS-GC activity in the suppressed state. MODE 3, it is the most interesting scenario. ROS-GC1 is present together with the Ca2+-dependent guanylate cyclase activators (CD-GCAPs)– NCδ, S100B – and with GCAP1. The ROS-GC activity oscillates between the GCAP1-dependent state and CD-GCAPs-dependent state. In these states the cyclase is a bimodal Ca2<sup>+</sup> transduction switch. And it functions according to the principles established for the synapse region between the photoreceptor and ON-bipolar cells (Duda et al., 2002; Venkataraman et al., 2003). This mode of ROS-GC1 activity regulation is possible because each of the three modulators targets its individual domain in ROS-GC1. These domains have been mapped and are schematically shown in **Figure 7**. Neurocalcin δ targets the ROS-GC1 region V836–D856 (Venkataraman et al., 2008); S100B, G962–N981and I1030–Q1041 (Duda et al., 2002); GCAP1, M445–L456 and L503–1522(Lange et al., 1999).

#### **MODE 3 – MODEL**

*Basal state – low Ca*2+: Ca2+-free GCAP1 stimulates ROS-GC1 catalytic activity. Cyclic GMP formed in response to this stimulation opens some of the CNG channels present in the sperm membranes (Wiesner et al., 1998; Revelli et al., 2002). At least some of the Ca2<sup>+</sup> ions entering the sperm through the open CNG channels are bound by neurocalcin δ and/or S100B. It is important to note that the half-maximal activation of ROS-GC1 by GCAP1 occurs at 707 nM Ca2<sup>+</sup> (Hwang et al., 2003), the Ca2<sup>+</sup> concentration at which neurocalcin δ and S100B are able to activate ROS-GC1 with approximately half-maximal effectiveness. With increasing Ca2<sup>+</sup> concentrations GCAP1 inhibits ROS-GC1 activity while neurocalcin δ and S100B stimulate it. *Active state* – Ca2+-bound NCδ and/or S100B further stimulate ROS-GC1 and cyclic GMP is formed in higher quantities leading to massive opening of CNG channels and influx of Ca2<sup>+</sup> (Shapiro et al., 1990; Revelli et al., 2002). In this phase, the ROS-GC1 activity is resultant of two opposing effects, inhibitory through GCAP1 and stimulatory, through NCδ and S100B. *Signal termination*: when cyclic GMP and [Ca2+]*<sup>i</sup>* reach the optimal level necessary for sperm membrane hyperpolarization and physiological function (e.g., capacitation and/or acrosome reaction; Galinado et al., 2000) combined

effects of protein phosphorylation and phosphodiesterase activity (Su and Vacquier, 2006) lead to lowering of cyclic GMP levels, closure of the CNG channels and cessation of the Ca2<sup>+</sup> influx.

## **CONCLUSION**

Co-expression of NCδ, S100B, and GCAP1 with ROS-GC1 in mammalian spermatogenic cells confers bimodal Ca2<sup>+</sup> sensitivity to cyclic GMP synthesis. The stimulatory arm of the switch operated by NCδ and S100B is essential for timely opening of the CNG channels by cyclic GMP produced and vast influx of Ca2<sup>+</sup> necessary for the acrosome reaction. In this stimulatory arm NCδ appears to be a significantly more important player than S100B. The inhibitory arm operated by GCAP1 stops the cyclic GMP synthesis and therefore Ca2<sup>+</sup> influx. Our findings demonstrate that ROS-GC1 transduction system and its bimodal Ca2<sup>+</sup> modulation is not unique to neurosensory and neurosensory-linked systems and has a more widespread role in general Ca2+-dependent signaling and may be a part of the fertilization machinery.

#### **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: 27 February 2014; accepted: 08 April 2014; published online: 29 April 2014. Citation: Jankowska A, Sharma RK and Duda T (2014) Ca2+-modulated ROS-GC1 transduction system in testes and its presence in the spermatogenic cells. Front. Mol. Neurosci. 7:34. doi: 10.3389/fnmol.2014.00034*

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

*Copyright © 2014 Jankowska, Sharma and Duda. 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.*

## Atrial natriuretic factor receptor guanylate cyclase, ANF-RGC, transduces two independent signals, ANF and Ca2**<sup>+</sup>**

## *Teresa Duda\*, Alexandre Pertzev and Rameshwar K. Sharma*

The Unit of Regulatory and Molecular Biology, Research Divisions of Biochemistry and Molecular Biology, Salus University, Elkins Park, PA, USA

#### *Edited by:*

Clint L. Makino, Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA

#### *Reviewed by:*

Peter Jedlicka, Goethe University, Germany Baojin Ding, University of Massachusetts Medical School, USA

#### *\*Correspondence:*

Teresa Duda, The Unit of Regulatory and Molecular Biology, Research Divisions of Biochemistry and Molecular Biology, Salus University, 8360 Old York Road, Elkins Park, PA 19027, USA e-mail: tduda@salus.edu

Atrial natriuretic factor receptor guanylate cyclase (ANF-RGC), was the first discovered member of the mammalian membrane guanylate cyclase family.The hallmark feature of the family is that a single protein contains both the site for recognition of the regulatory signal and the ability to transduce it into the production of the second messenger, cyclic GMP. For over two decades, the family has been classified into two subfamilies, the hormone receptor subfamily with ANF-RGC being its paramount member, and the Ca2<sup>+</sup> modulated subfamily, which includes the rod outer segment guanylate cyclases, ROS-GC1 and 2, and the olfactory neuroepithelial guanylate cyclase. ANF-RGC is the receptor and the signal transducer of the most hypotensive hormones, ANF– and B-type natriuretic peptide (BNP). After binding these hormones at the extracellular domain it, at its intracellular domain, signals activation of the C-terminal catalytic module and accelerates the production of cyclic GMP. Cyclic GMP then serves the second messenger role in biological responses of ANF and BNP such as natriuresis, diuresis, vasorelaxation, and anti-proliferation. Very recently another modus operandus for ANF-RGC was revealed. Its crux is that ANF-RGC activity is also regulated by Ca2+. The Ca2<sup>+</sup> sensor neurocalcin d mediates this signaling mechanism. Strikingly, the Ca2<sup>+</sup> and ANF signaling mechanisms employ separate structural motifs of ANF-RGC in modulating its core catalytic domain in accelerating the production of cyclic GMP. In this review the biochemistry and physiology of these mechanisms with emphasis on cardiovascular regulation will be discussed.

**Keywords: atrial natriuretic factor, atrial natriuretic factor receptor guanylate cyclase, calcium, cyclic GMP, neurocalcin δ, signal transduction**

#### **INTRODUCTION**

In the 1980s, the field of cellular signal transduction underwent total metamorphosis. Until then only two major paradigms of signal transduction, that of cyclic AMP and that of IP3 (inositol triphosphate) were known and were used to explain cellular signaling mechanisms and the second messenger concepts. Each of these two paradigms constituted an assemblage of three components – receptor, GTP binding protein, and the transducer catalyst – necessary for signal transduction. The astonishing discovery of a novel membrane protein that was simultaneously a receptor of a hormone and the transducer of its signal added a new dimension to our understanding of cellular signal transduction. The hormone was the newly described atrial natriuretic factor (ANF; de Bold, 1985) and the membrane protein was both its receptor and guanylate cyclase enzyme, termed therefore atrial natriuretic factor receptor guanylate cyclase (ANF-RGC; Kuno et al., 1986; Paul, 1986; Paul et al., 1987; Takayanagi et al., 1987; Meloche et al., 1988). Other terms to describe the protein are GC-A and NPR-A. ANF-RGC responded to ANF binding with accelerated synthesis of its second messenger cyclic GMP. The novelty of the system was 2-fold; first, a single protein, ANF-RGC, which contained both the ability to recognize the ANF hormonal signal and the activity to translate the hormonal information into the production of its second messenger, cyclic GMP was identified; and second, cyclic GMP was recognized as *bona fide* hormonal second messenger. Until then, the majority of laboratories forcefully denied the hormonal second messenger role of cyclic GMP (reviewed in Sharma, 2002, 2010).

The concept that ANF-RGC is indeed both hormone receptor and a catalyst was further supported by cloning studies (Chinkers et al., 1989; Lowe et al., 1989; Pandey and Singh, 1990; Duda et al., 1991; Marala et al., 1992). Homology cloning made possible finding other receptor membrane guanylate cyclases: C-type

**Abbreviations:** ANF, atrial natriuretic factor; ANF-RGC, atrial natriuretic factor receptor guanylate cyclase; ARM, ATP-regulatory module; ATP, adenosine triphosphate; BNP, B-type natriuretic peptide; CNP, C-type natriuretic peptide; CNP-RGC, C-type natriuretic peptide receptor guanylate cyclase; cyclic AMP, 3- ,5- -cyclic adenosine monophosphate; cyclic GMP, 3- ,5- -cyclic guanosine monophosphate; CNG channel, cyclic nucleotide gated channel; GCAP, guanylate cyclase activating protein; GC-A, guanylate cyclase-A; GC-B, guanylate cyclase-B; GC-C, guanylate cyclase-C; GC-D, guanylate cyclase-D; GC-E, guanylate cyclase-E; GC-F, guanylate cyclase-F; GC-G, guanylate cyclase-G; KHD, kinase homology domain; K-LD, kinase-like domain; NCδ, neurocalcin δ; NPR-A, natriuretic peptide receptor-A; ONE-GC, olfactory neuroepithelial guanylate cyclase; STa-RGC, heat-stable enterotoxin (also guanylin and uroguanylin) receptor guanylate cyclase; ROS-GC, rod outer segment guanylate cyclase.

natriuretic peptide receptor guanylate cyclase, CNP-RGC (also known as GC-B; Chang et al., 1989; Schulz et al., 1989; Duda et al., 1993b) and heat-stable enterotoxin (also guanylin and uroguanylin) receptor guanylate cyclase, STa-RGC (GC-C; Schulz et al., 1990; de Sauvage et al., 1991; Singh et al., 1991; Hamra et al., 1993; Khare et al., 1994). Identification of these three structurally and functionally related membrane guanylate cyclases that were receptors for hormonally active peptides resulted in a notion that all membrane guanylate cyclases, even those still to be discovered, were receptors for specific extracellular ligands.

The notion was short-lived, however. The next four, in chronological order, identified membrane guanylate cyclases: the rod outer segment guanylate cyclases, ROS-GC1 (also known as RetGC1 or GC-E; Koch, 1991; Shyjan et al., 1992; Goraczniak et al., 1994) and ROS-GC2 (RetGC2 or GC-F; Lowe et al., 1995; Yang et al., 1995; Goraczniak et al., 1998), the olfactory neuroepithelium guanylate cyclase, ONE-GC (also termed as GC-D) (Fulle et al., 1995; Duda et al., 2001a) and GC-G (Schulz et al., 1998), did not respond with increased activity to any extracellular ligand. They were therefore branded as "orphan receptors" (Fulle et al., 1995; Yang et al., 1995; Wedel and Garbers, 1997; Schulz et al., 1998).

ROS-GC1 and ROS-GC2 were not orphan RGC, however. Biochemical and physiological findings began to reveal that regulation of their catalytic activities is specific to their physiological function which is to return the illuminated photoreceptors to the dark, resting state. The illumination leads to activation of cyclic GMP phosphodiesterase, depletion of cyclic GMP, closure of the cyclic GMP gated (CNG) channels, and lowering the free Ca2<sup>+</sup> concentration (reviewed in Pugh et al., 1997; Koch et al., 2010). The ROS-GCs' task is to restore the dark-level of cyclic GMP allowing opening of the CNG channels and increase of Ca2<sup>+</sup> influx. Ca2<sup>+</sup> concentration, thus, determines the activity of ROS-GCs but it occurs in an indirect way. Guanylate cyclase activating proteins (GCAPs) sense the post-illumination fall in Ca2<sup>+</sup> and stimulate ROS-GCs to resynthesize cyclic GMP at a faster rate (reviewed in Detwiler, 2000; Koch et al., 2010). In this cyclic GMP-Ca2<sup>+</sup> feedback mechanism, ROS-GCs do not respond to an extracellular ligand but to their intracellular Ca2<sup>+</sup> sensing ligands, the GCAPs (Palczewski et al., 1994; Dizhoor et al., 1995; Duda et al., 1996b).

Soon thereafter the evidence began to emerge, primarily from our laboratory, that GCAPs are not the only Ca2<sup>+</sup> sensing modulators of ROS-GC activity. While increasing Ca2<sup>+</sup> concentrations inhibit ROS-GCs activity through GCAPs, two other Ca2<sup>+</sup> sensors, S100B, and neurocalcin δ (NCδ) stimulate ROS-GC1 in a Ca2+ dependent fashion (Pozdnyakov et al., 1995; Duda et al., 1996a; Margulis et al., 1996; Kumar et al., 1999). The Ca2+-dependent S100B-mediated activation of ROS-GC operates in cones including their outer segments and pedicles (Wen et al., 2012). Its role in photo- and visual transductions remains to be established, but existing data indicate its involvement in transmission of the visual signal from cone ON-bipolar cells (Wen et al., 2012). Ca2<sup>+</sup> signaling of ROS-GC activity mediated by NCδ is operative in retinal ganglion cells (Krishnan et al., 2004) but its linkage with the visual transduction is not clear yet.

Similar, but with additional twists, regulation of ONE-GC occurs in a subset of olfactory sensory neurons (Fulle et al., 1995; Duda et al., 2001a; Pertzev et al., 2010). The cyclase was initially classified as a member of the Ca2<sup>+</sup> modulated subfamily (Duda et al., 2001a). Its activity is modulated in a Ca2+-dependent fashion by NCδ (Duda et al., 2001a, 2004) and by GCAP1 (Duda et al., 2006, 2012a; Pertzev et al., 2010). Remarkably, the Ca2+-GCAP1 pattern of ONE-GC modulation is opposite to that of ROS-GC modulation. Sensing increasing concentrations of Ca2<sup>+</sup> GCAP1 stimulates ONE-GC while it inhibits ROS-GC (Duda et al., 2006, 2012a).

After the Ca2+-dependent modulation of ONE-GC activity was demonstrated, an extracellular ligand of the cyclase was found. The ligand was the odorant uroguanylin (Leinders-Zufall et al., 2007; Duda and Sharma, 2008; reviewed in Sharma and Duda, 2010; Zufall and Munger, 2010). Hence, ONE-GC responds to both extracellular (uroguanylin) and intracellular Ca2<sup>+</sup> signals. At this point a new, two-step "cross-talk" odorant transduction model was proposed (Duda and Sharma, 2009). In "step 1," the odorant uroguanylin interacts with the receptor domain of ONE-GC, causing its minimal activation. The cyclic GMP generated in response to uroguanylin signal opens a few of the CNG3 channels leading to some increase in [Ca2+]i in the odorant receptor cell. In "step 2," the membrane bound NCδ and GCAP1 sense the increase in [Ca2+]i and in their Ca2+-bound states fully activate ONE-GC triggering maximal influx of Ca2+and depolarization of the olfactory receptor cell membrane.

The odorant receptor ONE-GC linkage with the intracellular free Ca2<sup>+</sup> signals brought forth a question whether this cyclase is a unique case of dually regulated membrane guanylate cyclase. Latest studies from our laboratory demonstrate that it is not. The nascent member of the hormone receptor subfamily, the ANF-RGC, is also responsive to Ca2<sup>+</sup> (Duda et al., 2012a,b) and NCδ is the sensor protein which communicates the Ca2<sup>+</sup> signal to ANF-RGC and in the Ca2+-bound state, activates ANF-RGC activity. The two modes of ANF-RGC activation, hormonal and Ca2+ dependent, engage their specific and independent pathways of transmitting the stimulatory signals to the catalytic domain. The end-product, however, is common, the second messenger cyclic GMP. Because of the individual signaling mechanisms involved in Ca2<sup>+</sup> and hormonal signaling their net effects are multiplicative. Hence, in terms of the original sub-classification, the present day knowledge is that at least two cyclases, ANF-RGC and ONE-GC, are hybrids sensing both hormonal and Ca2<sup>+</sup> signals.

The last cloned membrane guanylate cyclase was GC-G. Until today, the information of its function is very scarce. It was suggested that the mouse form of GC-G is selectively expressed in the sperm and may be involved in the process of capacitation (Kuhn et al., 2004). Other reports suggest that the cyclase is expressed in Grueneberg ganglion olfactory subsystem where it is responding to CO2 (Chao et al., 2010) or to cool ambient temperature indicating its role in thermo-sensation (Mamasuew et al., 2008).

The following discussion is exclusively devoted to the ANF-RGC. It was the first template model, which established that cyclic GMP is a *bona fide* hormonal second messenger. The ongoing studies define the second ANF-RGC transduction model, Ca2+-modulated signaling. It makes ANF-RGC a bimodal switch, hormonal and Ca2+. An additional significant part of the review is coverage of the manner in which these two models have now begun to explain the biochemical principles of cardiovascular, renal and endocrine homeostasis with special emphasis on blood pressure regulation.

#### **HORMONAL SIGNALING OF ANF-RGC ACTIVATION BINDING OF LIGAND-HORMONE ANF**

Although intuitively expected, the first experimental evidence that the ANF binding site in ANF-RGC is located within the extracellular domain came first from the site-directed and deletion mutagenesis studies. A mutant ANF-RGC was cloned from rat adrenal cDNA library (Duda et al., 1991) and termed GCα. Its sequence differed from ANF-RGC by two amino acid substitutions Q<sup>338</sup> <sup>→</sup> H and L364 <sup>→</sup> P which resulted from single nucleotide mutations, CAG → CAC and CTG → CCG, respectively. Expressed in heterologous system of COS cells, GCα was properly located in the plasma membrane and exhibited basal guanylate cyclase activity; it however, neither bound nor responded to ANF or other natriuretic peptides (Duda et al.,1991). These results demonstrated that the two amino acid substitutions exclusively abolished binding of ANF and therefore ANF signaling of the cyclase activity. Restoration by site-directed mutagenesis, glutamine at position 338 and leucine at position 364, reinstated ANF binding and ANF-dependent stimulation of the cyclase.

The site of ANF binding was further systematically analyzed (McNicoll et al., 1992, 1996; He et al., 1995; Misono, 2002). By cross-linking and proteinase digestion it was determined that the amino terminus of ANF is in contact with the region M173-F188 of ANF-RGC, and the C-terminus, with D191-R198 region. The fact that the identified contact sites were very close to each other was interpreted that the N- and C- termini of ANF interface distinct subunits of ANF-RGC homodimer. Importantly, it also implied that one ANF-RGC dimer binds one molecule of ANF.

Details of ANF binding to the ANF-RGC extracellular domain were unraveled by analyses of the extracellular domain of ANF-RGC co-crystallized with ANF (Ogawa et al., 2003, 2004). These analyses revealed that (1) the extracellular domain of ANF-RGC exists as a dimer in the head-to-head configuration; (2) one ANF-RGC dimer binds one molecule of ANF (2:1 complex); (3) the binding site in one monomer differs from that in the other - one monomer binds the N-terminal part of ANF and the other binds C-terminal part; (4) there is no preference in ANF binding to a specific monomer of the extracellular domain; bound ANF occurs in two alternate orientations of equal occurrence; (5) the membranedistal portion of the extracellular domain contains chloride ion necessary for ANF binding.

#### **TRANSMEMBRANE MIGRATION OF THE ANF BINDING SIGNAL**

With the exception of enterotoxin RGC, all membrane guanylate cyclases contain a hinge region juxtaposed to the N-terminal side of the transmembrane domain. This region contains two conserved cysteine residues separated from each other by 6–8 residues. In ANF-RGC these residues are Cys<sup>423</sup> and Cys432 and were indicated as a critical structural motif in ANF signaling (Huo et al., 1999; Labrecque et al., 1999; Miyagi and Misono, 2000). There was however no consensus on the mechanism of its operation. Based on results with intact cells transfected with ANF-RGC cDNA it was proposed that the cysteines are involved in dimerization through the formation of an inter-chain S-S bridge (Huo et al., 1999; Labrecque et al., 2001) or, that they form an intra-chain disulfide bridge (Huo et al., 1999; Miyagi and Misono, 2000). Analyses of guanylate cyclase activity in isolated membranes of COS cell expressing C423S, C432S, and C423S, C432S mutants (Duda and Sharma, 2005) allowed to conclude that both C<sup>423</sup> and C<sup>432</sup> residues independently and equally control the catalytic activity of ANF-RGC and that activation of ANF-RGC does not involve interchain-disulfide bond formation. In the basal state, the disulfide motif keeps the ANF-RGC in its repressed catalytic state and ANF/ATP signaling brings it to the fully active state (Duda and Sharma, 2005).

A model of hormone-induced rotation has been proposed to explain how the extracellular signal is transmitted through the hinge region to the intracellular domain of ANF-RGC (Ogawa et al., 2004; Misono et al., 2005). Binding of ANF causes a twist motion of both ANF-RGC monomers centered on a support point close to the bound ANF. This twist motion translocates the juxtamembrane domains of both monomers with only minimal change in the distance between them. This movement constitutes a signal, which is transmitted through the transmembrane domain. Rotation of the transmembrane domain by 40 degrees occurs in response to ANF binding (Parat et al., 2010).

#### **PASSAGE OF THE ANF SIGNAL THROUGH THE INTRACELLULAR DOMAIN**

#### *ANF signaling requires ATP*

Early studies, before the biochemical characterization of ANF-RGC, demonstrated that activation by ANF is significantly amplified by ATP in the guanylate cyclase activation (Kurose et al., 1987; Cole et al., 1989; Chang et al., 1990; Larose et al., 1991; reviewed in Sharma, 2002; Duda et al., 2005). Subsequently, it was demonstrated that ATP is obligatory for ANF-dependent activation of ANF-RGC (Chinkers et al., 1991; Marala et al., 1991; Goraczniak et al., 1992; Wong et al., 1995; reviewed in Duda et al., 2005). Because the ATP effect was mimicked by its nonhydrolyzable analogs, ATPγS and AMP-PNP, it was suggested that ATP acts as an allosteric regulator (Chinkers et al., 1991; Marala et al., 1991; Goraczniak et al., 1992; Wong et al., 1995; reviewed in Duda et al., 2005). Shortly thereafter it was shown that ATP is also obligatory for the CNP-dependent stimulation of CNP-RGC (Duda et al., 1993a,b). Thus, it appears that the requirement of ATP is a common attribute of natriuretic peptide RGC signaling.

Studies with ANF-RGC deletion mutants revealed that the ATP regulated region resides in the intracellular portion of ANF-RGC between the transmembrane and the C-terminal catalytic domain (Chinkers and Garbers, 1989; Goraczniak et al., 1992). This region was named the kinase homology (or kinase-like domain; KHD or K-LD) because of its sequence similarity with the corresponding domains of tyrosine kinases (Chinkers et al., 1989; Chinkers and Garbers, 1989; Duda et al., 1991). A model for the ATP function was proposed that "binding of ANP to the extracellular domain of its receptor initiates a conformational change in the protein K-LD, resulting in de-repression of guanylate cyclase activity" (Chinkers and Garbers, 1989). The central idea behind the model was that KHD in native ANF-RGC suppresses the catalytic module activity; ANF functions by relieving this suppression. Study from our group using two KHD deletion mutants, -506–677 and -555– 762, negated this model and proposed an alternative one where ATP via ATP regulated module (ARM) potentiates the hormonal signal"(1) the signal is caused by the binding of the hormone to the receptor site; (2) there is a transmembrane migration of the signal; (3) the signal is potentiated by ATP at ARM; and (4) the amplified signal is finally transduced at the catalytic site" (Goraczniak et al., 1992).

#### *ATP allosteric effect and the ARM domain*

Within the KHD ANF-RGC contains a sequence, G503-X-G505-X-X-X-G509 (Goraczniak et al., 1992) which is a modified form of the protein kinases' nucleotide-binding consensus sequence G-X-G-X-X-G necessary for kinase activity (Wierenga and Hol, 1983; Hanks et al., 1988; Bratová et al., 2005). This motif was probed for its significance in ATP function in ANF signaling. Through analyses of a series of deletion and point mutants it was determined that the G503-X-G505-X-X-X-G509 motif is critical for the ATP function (Goraczniak et al., 1992; Duda et al., 1993a; Duda and Sharma, 1995; reviewed in Sharma, 2002; Duda et al., 2005). Although it is not involved in ATP binding it maintains the steric arrangements to fit the ATP molecule. For this reason, the motif has been named ARM (Goraczniak et al., 1992) and the KHD, where the ARM resides, was termed the ARM domain (Duda et al., 2001b; Sharma et al., 2001). To get an insight into the mechanism of ATP function, the structure of the ARM domain was simulated through computer-assisted homology based modeling (Duda et al., 2001b; Sharma et al., 2001; PDB ID 1T53). The basic structural features of the model have been experimentally validated through point mutation/expression studies (Duda et al., 2001b; Sharma et al., 2001; reviewed in Sharma, 2002; Duda et al., 2005).

*Spatial determinants of the ARM domain.* The domain comprises amino acid residues 496–771, which are arranged into two lobes: the smaller, N-terminal (91 aa residues: 496–586), and the larger, C-terminal (185 aa residues: 587–771; Duda et al., 2001b). Four antiparallel β strands and one α helix form the smaller lobe; the larger lobe is predominantly helical, composed of eight αhelices and two β strands (Duda et al., 2001b; Sharma et al., 2001). The ARM motif, G503-X-G505-X-X-X-G509, is located within the smaller lobe.

*The ATP-binding pocket.* In order to identify the ARM domain residues potentially involved in ATP binding the model of this domain in its ATP-bound form was analyzed (Duda et al., 2005). A radius of 4 Å from the ATP molecule was chosen as the limiting distance for the electrostatic, hydrogen or van der Waals' interactions. Two sets of residues were identified (Figure 4 in reference Duda et al., 2005): (1) those forming the floor of the ATP binding pocket; (2) those surrounding ATP molecule. The residues in the first set are: G503, R504, G505, S506, N507, Y508, and G509. These residues have no direct chemical interaction with ATP; they, however, provide necessary space to accommodate the ATP molecule. The residues in the second set surround the individual components of ATP, the adenine ring, the ribose ring and the triphosphate moiety. The adenine ring is surrounded by L511, T513, T514, E515, Q517, A533, T580, E581, C583, V635, and T645 and L511, T513, and C583 are within a distance shorter than 3 Å.

G503, L511, T513, T514, G580, S587, D590surround the ribose ring with L511 and T<sup>514</sup> located within 2.5 Å radius from the ribose. The phosphate groups are surrounded by R504, G505, L511, K535, N633, D646, and K535 is the nearest residue (2.6 Å) forming a hydrogen bond with the phosphate and has been shown to be critical for ATP regulation of ANF-dependent ANF-RGC activity; D646 interacts with the triphosphate group of ATP through the formation of a coordinate bond with the metal ion Mg2+.

*Interaction of ATP with the ARM domain.* Kinetics of ATP binding to theARM domain was determined through SPR spectroscopy (Burczynska et al., 2007). AMP-PNP, the non-hydrolyzable analog of ATP was used for the binding studies. Half-maximal binding (EC50) occurred when the concentration of AMP-PNP was ∼0.2 mM and the calculated*K*<sup>D</sup> value was 0.21 mM. Similar results (EC50 value of 0.15 mM and *K*<sup>D</sup> of 0.13 mM) were obtained when 8-azido-ATP was used.

The model-predicted ATP binding pocket (*vide supra*) was authenticated experimentally by cross-linking of the purified isolated ARM domain protein with 8-azido-ATP, trypsin digestion of the cross-linked product and sequencing of the resulting peptides (Burczynska et al., 2007). Three peptides were found to be photoaffinity modified. The longest modified peptide was identified as G614MLFLHNG-AICSHGNLKSSNCVVDGR639and the shortest as S631SNC634V635VDGR. In all three peptides Cys634was modified indicating that this residue was the closest to the azido group. The G614–R639 fragment contains six ATP-binding pocket-predicted residues (Duda et al., 2005): S625, K630, S631, S632, N633, and V635; and the shortest S631–R639 fragment contains three residues: S631, N633, and V635(Duda et al., 2005). The refined ARM model (Duda et al., 2005) predicts that V<sup>635</sup> is within a 4 Å radius from the adenine ring; cross-linking of the neighboring residue to C<sup>634</sup> validates that these residues are indeed the closest to the C-8 of the adenine ring.

*ATP binding dependent changes in the ARM domain.* If ATP binding to the ARM domain were to serve signaling purposes, it should induce structural changes that ultimately would signal activation of the catalytic domain. Comparative analyses of the ARM domain models in the apo and ATP-bound states identified such changes. They involve rotations of the β strands within the smaller lobe as well as movements of the β strands and α helices in the larger lobe (Duda et al., 2001b; Sharma et al., 2001; reviewed in Duda et al., 2005). Consequences of two such changes appear to be of particular importance: first, of the β1, β2 strands in the smaller lobe, and the second, of the EF and F helices in the larger lobe.

The β1 and β2 strands and the loop connecting them encompasses the G503-X-G505-X-X-X-G509 motif (Duda et al., 2001b; Sharma et al.,2001; reviewed in Duda et al.,2005), which is meshed

in and flanked by six phosphorylation sites, S497, T500, S502, S506, S510, T513 (Potter and Hunter, 1998a, 1999a). The present consensus is that "phosphorylation of KHD is absolutely required for hormone-dependent activation of NPR-A" (Potter and Hunter, 1998a,b, 1999b). In this concept, hypothetical protein kinase and protein phosphatase co-exist with ANF-RGC (Foster and Garbers, 1998). Until now, however, neither the kinase nor the phosphatase has been identified.

Comparison of the ATP-free and ATP-bound ARM domain models allows explaining how ATP binding makes it possible for the serine and threonine residues to undergo phosphorylation. After ATP binds to the ARM domain, the β1, β2 strands, and the loop between them shift by ∼3–4 Å and rotate by ∼15◦ (Duda et al., 2001b; Sharma et al., 2001; reviewed in Duda et al., 2005). This movement triggers reorientation of the serine and threonine residues (**Figure 1**, red colored residues) causing the side chains and the OH groups of T500, S502, S506, and T513 becoming directed toward the protein surface (**Figure 1**, compare the positions of the cyan- and red-colored OH groups). The change in the positions of the side chains is most drastic for the S<sup>502</sup> and S506 residues. Although upon ATP binding there is no toward the surface reorientation of the S<sup>497</sup> and S<sup>510</sup> OH groups, the entire residues are shifted toward the surface (**Figure 1**). This structural rearrangement permits the hypothetical protein kinase to access the side chains of the residues and transfer the phosphate group (Duda et al., 2011).

Another important result of ATP binding on the dynamics within the ARM domain is the translocation within the larger lobe of two helices, EF and F. Their movement was characterized by analyzing the ATP-dependent changes in the fluorescence intensity and wavelength of two tryptophan residues W<sup>601</sup> and W669 (Duda et al., 2009). Both residues reside in the larger lobe of the ARM domain, outside the ATP-binding pocket. The W<sup>601</sup> residue is a part of the helix E structure (Duda et al., 2001b, 2005), its side-chain is oriented toward the protein surface, and it is flanked by several hydrophobic residues, F640, L607, S606, and S596 (**Figure 2A**). These structural features define the fluorescence λmax of W601 at 332 nm. W669 residue is located at the end of a loop

**of the six phosphorylable residues.** The conformation of the six phosphorylated residues is shown before (cyan) and after (red) ATP binding. The ATP molecule is shown in green. The positions of the OH groups are indicated by cyan and red balls (reproduced with permission from ref. Duda et al., 2011).

connecting β8 strand and EF helix (Duda et al., 2009); it is a part of a conserved hydrophobic motif, 669WTAPELL675. The side chain of W<sup>669</sup> is flanked by the polar residues: E699, S631, K667, K630, and L<sup>696</sup> (**Figure 2B**). This environment causes red shifts in the fluorescence λmax of W<sup>669</sup> to 345 nm.

Superimposition of the ATP-free and the ATP-bound forms indicates that ATP binding induces contraction of the entire ARM domain and affects the orientation and environment of W669. The contraction results in shortening of the distance between W669and the ATP binding pocket (**Figure 2C**). ATP binding also causes reorientation of the W<sup>669</sup> side chain. It turns and becomes more shielded by the surrounding amino acids. Turning of the W<sup>669</sup> side chain pushes the remainder of the motif, 670TAPELL675, to the surface resulting in its exposure. In general, the movement of a hydrophobic motif toward the surface of the protein indicates its readiness for interaction. For the 669WTAPELL675 motif it was proposed that it interacts with subsequent transduction motif, possibly within the catalytic domain, propagates the ANF/ATP binding signal and activates the catalytic domain (Duda et al., 2009).

Based on the studies narrated above a model for ANF/ATP signaling of ANF-RGC activity was proposed: The ANF signal originates by the binding of one molecule of ANF to the extracellular dimer domain of ANF-RGC (Ogawa et al., 2004, 2009). The binding modifies the juxtamembrane region where the disulfide 423Cys-Cys432 structural motif is a key element in this modification (Ogawa et al., 2004, 2009; Duda and Sharma,2005). The signal twists the transmembrane domain (Parat et al., 2010), induces a structural change in the ARM domain, and prepares it for the ATP activation. ARM domain binds ATP to its pocket what leads to a cascade of temporal and spatial changes (Duda et al., 2001b; Sharma et al., 2001; reviewed in Duda et al., 2005). They result in (1) exposure of the hydrophobic 669WTAPELL675 motif which directly (or indirectly) interacts with the catalytic domain causing its partial activation; and (2) exposure and phosphorylation of six serine, threonine residues and full activation of ANF-RGC. Concomitantly, phosphorylation converts ATP binding site from the high to low affinity, ATP dissociates and ANF-RGC returns to its ground state (Duda et al., 2011).

#### **CATALYTIC DOMAIN**

When more then 20 years ago an ANF RGC was purified to homogeneity and shown to contain the ANF binding and cyclic GMP forming activities on the same protein chain, the authors proposed a topological model for the transmembrane receptor enzyme in which the receptor part was extracellular and the cyclase catalytic domain was intracellular (Sharma et al., 1988a,b). The cloning studies confirmed this prediction. Alignment of the deduced amino acid sequences of the cloned guanylate cyclases indicated that the catalytic domain is located at the C-terminus of ANF-RGC and comprises 239 amino acid residues (Chinkers et al., 1989; Duda et al., 1991). This prediction was tested experimentally (Thorpe and Morkin, 1990; Thorpe et al., 1996). The carboxy terminal 293 amino acids fragment of ANF-RGC was expressed as a soluble protein and shown to exhibit guanylate cyclase activity. These results were in agreement with studies that determined through radiation inactivation experiments that the ANF-RGC

**FIGURE 2 | (A)** Amino acid residues surrounding W601 and **(B)** amino acid residues surrounding W669 within the ARM domain. Amino acid residues depicted in red are located withina4Å sphere from the respective tryptophan residue (green). **(C)** Conformational changes within the 669WTAPELL675 motif induced by ATP binding to the ARM domain. The backbone structure of the ATP-bound ARM domain is shown in cyan and the ATP molecule is in green. The 669WTAPELL675 motif is shown in magenta color. Apo structure of the ARM domain was superimposed on the ATP-bound form to assess the relative, ATP binding induced, conformational changes. For clarity, only the 669WTAPELL675 motif (shown in red) of the apo-enzyme is visible. ATP binding results in a

fragment containing cyclase activity has a molecular weight of 32 ± 8 kDa (Tremblay et al., 1991). Further studies identified several residues within this region that appeared to be critical for the guanylate cyclase activity: L<sup>817</sup> (Miao et al., 1995), D877, K887, D893, G900, H909, R940, and H944 (Thompson and Garbers, 1995). Their individual mutations to Ala resulted in ANF-RGC mutants without detectable guanylate cyclase activity. In contrast to these residues, mutation of E<sup>974</sup> to Ala resulted in a hyperactive ANF-RGC mutant (Wedel et al., 1997). Based on these results it was concluded that the indicated residues are located within or close to the catalytic center or are critical for the proper folding of the catalytic center.

more compact structure of the ARM domain: the W669 side chain moves toward the ATP binding pocket while the side chains of T670, E673, L674, and L675 move toward the protein surface [compare the orientation of side chain of these amino acid residues before (in red) and after (in magenta) ATP binding; fonts for W669 and L675 residues are increased for better visibility]. This movement changes the surface properties of the ARM domain. The movement toward the surface of the protein is poised to facilitate interaction of this amino acid stretch with subsequent transduction motif, possibly within the catalytic domain, propagation of the ANF/ATP binding signal and activation of the catalytic domain (reproduced with permission from ref. Duda et al., 2009).

In the absence of a crystal structure of any membrane guanylate cyclase catalytic domain a model of the catalytic center of retGC-1, a mammalian membrane guanylate cyclase expressed in the retina, also known as ROS-GC1 was proposed (Liu et al., 1997). The model was built based on homology with the catalytic center of the adenylate cyclase. The modeling studies allowed identification of the critical residues constituting the catalytic core. Because the catalytic domains of all membrane guanylate cyclases are highly conserved (>95% of sequence identity), by homology substitution, the critical residues of the ANF-RGC catalytic center have been identified and they are described below.

The catalytic center is homodimeric. There are two GTP binding sites. Each GTP molecule interacts with both monomers. In the following description of the ANF-RGC catalytic center the monomers are labeled "A" and "B" and the location of each residue within a monomer A or B is indicated. Numbering of amino acids is according to the mature protein (Duda et al., 1991). R959(B) and C961(B) interact with guanine's O6; second NH2 group of R959(B) interacts with guanine's N1; the carboxy group of E889(B) interacts with R959(B) and through it with N1; G964(B) interacts with N7; T854(A), D893(A) and N968(B) form hydrogen bonds with ribose's OH (3- ) group; R972(B binds α phosphate group; R940(A) binds β phosphate group; E974(A) through Mg2<sup>+</sup> interacts with β phosphate group; and S971(A) binds γ phosphate group. Specificity toward GTP is determined by E889 and C961 of ANF-RGC. This conclusion is drawn based on studies showing that in ROS-GC1 mutation of the corresponding E to K and C to D converts guanylate cyclase activity into adenylate cyclase activity (Tucker et al., 1998).

## *Mechanism of ANF-RGC activation – the role of the* **<sup>669</sup>***WTAPELL***<sup>675</sup>** *motif*

An important feature of the model of ANF/ATP-dependent activation of ANF-RGC is the relay function of the 669WTAPELL675 motif. It switches between the ATP-bound ARM domain and activates the catalytic domain. It was proposed that this motif interacts with and signals the antiparallel homo-dimer of the catalytic domain to undergo conformational changes and form a functional catalytic center (Duda et al., 2009). The exact site within the catalytic domain which is targeted by the 669WTAPELL675 motif remains to be determined. Preliminary scanning experiments, however, suggest that the site may be located close to the E889, a residue critical for cyclase specificity toward GTP (Tucker et al., 1998).

If the elements of the model are correct and the 669WTAPELL675 motif is a switch, its deletion from the ANF-RGC should lead to a protein which binds ANF and ATP but is not able to respond to them with increased synthesis of cyclic GMP. This assumption was tested experimentally. An ANF-RGC mutant was created in which the 669WTAPELL675 motif was deleted by mutagenesis.Wild type ANF-RGC and the mutant were expressed individually in COS cells and their membranes were analyzed (1) by Western blot to determine the proteins' levels of expression; (2) for basal guanylate cyclase activity; (3) for ANF binding; and (4) for KM for the substrate GTP. The results demonstrated that (1) the wild type ANF-RGC and the 669WTAPELL675 deletion mutant were expressed to the same level as determined by Western blot; (2) the basal cyclase activity of the mutant, 20 pmol cGMP min−<sup>1</sup> (mg protein)−1, was virtually identical to that of the wild type ANF-RGC, 21 pmol cGMP min−<sup>1</sup> (mg protein) <sup>−</sup>1; (3) their receptor activities were equal with the respective specific binding values of 9.7 <sup>±</sup> 0.6 and 10.1 <sup>±</sup> 0.9 pmol [125I]ANF/mg protein and (4) the KM values for the substrate GTP, 614 μM for the wild type and 608 μM for the mutant protein were identical. Thus, the basal integrity of the protein remained intact despite the deletion (Duda et al., 2009).

When the mutant was exposed to increasing concentrations of ANF and ATP its cyclase activity remained practically unchanged whereas under identical conditions wild type ANF-RGC was stimulated over 6-fold (Duda et al., 2009, 2013) demonstrating that in line with the original expectations the catalytic domain does not increase cyclic GMP synthesis in the presence of ANF/ATP. This type of outcome could only be observed if the mutant protein did not bind ANF and/or ATP or the catalytic domain lost its cyclase ability. Both of these possibilities were put to rest as the mutant and the wild type cyclase did not differ in ANF and ATP binding nor in enzymatic activity (*vide supra* and ref. Duda et al., 2009). Therefore the only explanation for the unchanged level of cyclic GMP synthesis is that the 669WTAPELL675 deletion mutant is not able to respond to ANF and ATP. This leads to just one logical conclusion that in the absence of the 669WTAPELL675 motif the ANF/ATP binding information is not transmitted to the catalytic domain. Thus, the 669WTAPELL675 motif is the critical transmitter of the ATP-potentiated ANF signal to the catalytic domain where it is translated into generation of cyclic GMP.

## **Ca2<sup>+</sup> SIGNALING OF ANF-RGC ACTIVITY NEURONAL CALCIUM SENSOR NEUROCALCIN δ**

Neurocalcin δ belongs to a distinct subfamily of neuronal calcium sensor proteins (NCS) together with visinin-like proteins (VILIPs) and hippocalcin. They all are acylated at the N-terminus by myristic acid and undergo a classical calcium-myristoyl switch (Ladant, 1995), e.g., they bury the myristoyl group in a hydrophobic pocket in Ca2+−free form and expose it in Ca2+-bound form, as first observed and described for recoverin (Zozulya and Stryer, 1992). Exposure of myristoyl group enables the protein association with the cell membrane. However, once it binds, in a Ca2+-dependent fashion, to the membrane phospholipids, even after removing Ca2<sup>+</sup> by the addition of EGTA part of it remains membrane bound (Krishnan et al., 2004). Although NCδ is primarily expressed in neuronal tissues, its expression in the periphery is also observed.

Functionally, NCδ has been linked to receptor endocytosis through interaction with α- and β-clathrin and β-adaptin (Ivings et al., 2002), trafficking and membrane delivery of glutamate receptors of the kainate type (Coussen and Mulle, 2006), and due to its Ca2+-dependent affinity for S100B protein and tubulin βchain (Okazaki et al., 1995), with microtubule assembly (Iino et al., 1995). In the sensory and sensory-linked neurons, the presence of NCδ has been found in the inner plexiform layer of the retina, e.g., in the amacrine and ganglion cells (Krishnan et al., 2004), olfactory sensory neurons (Duda et al., 2001a, 2004) and recently, it has been identified in type II cells of mouse circumvallate taste papillae, indicating its possible role in gustatory transduction (Rebello et al., 2011).

A relatively newly identified function of NCδ is its Ca2+ dependent modulation of the activities of membrane guanylate cyclases ROS-GC1 in the retina and ONE-GC, in the olfactory neuroepithelium (Duda et al., 2001b, 2004; Krishnan et al., 2004). In these tissues NCδ co-localizes with its respective target cyclases. The exact physiological significance of the ROS-GC1- NCδ signaling system in the retinal neurons is not known yet. It can be, however, safely stated that the pathway is not present in the rod and cone outer segments, thus is not linked with the phototransduction machinery. The system has been localized to the lower strata

of the inner plexiform layer and to a subpopulation of ganglion cells (Krishnan et al., 2004).

In the olfactory neuroepithelium NCδ serves as a Ca2<sup>+</sup> sensor component of the two-step odorant uroguanylin signaling machinery. This signaling mechanism was proposed to be initiated by uroguanylin interaction with the extracellular receptor domain of ONE-GC (Leinders-Zufall et al., 2007; Duda and Sharma, 2008; reviewed in Zufall and Munger, 2010; Sharma and Duda, 2010). This interaction leads to partial activation of ONE-GC, generation of small amount of cyclic GMP, partial opening of cyclic GMP-gated channel and influx of Ca2<sup>+</sup> into the olfactory receptor neuron. In the next step, Ca2<sup>+</sup> binds to NCδ which, then fully activates ONE-GC (Duda and Sharma, 2009).

#### **FREE Ca2<sup>+</sup> SIGNALS ANF-RGC ACTIVATION**

In the course of mapping the NCδ targeted site on ROS-GC1 to which it binds and transmits the Ca2<sup>+</sup> signal to the catalytic domain for signal translation into the generation of cyclic GMP, our group made a remarkable observation that NCδ binds directly to the catalytic domain and, thereby, activates ROS-GC1 (Venkataraman et al., 2008). Protein database comparison shows sequence conservation of the catalytic domain in the membrane guanylate cyclase family hinting at a possibility that other membrane guanylate cyclases might be activated by Ca2<sup>+</sup> via NCδ as well. To test this possibility, ANF-RGC membrane guanylate cyclase, for which the only established signal transduction mechanism was through ANF/ATP (*vide supra*), was chosen. And indeed, in a recombinant system, myristoylated NCδ stimulatedANF-RGC activity in a dose- and Ca2+-dependent manner; 0.5 μM Ca2<sup>+</sup> and 0.5 μM NCδ triggered half-maximal activation of ANF-RGC (Duda et al., 2012a,b). These results for the first time demonstrated that ANF-RGC activity is dually regulated, by peptide hormones ANF and BNP, and by Ca2+, thus, at least *in vitro* the cyclase was deemed a bimodal signal transducer.

### *Myristoylated dimeric form of neurocalcin* **δ** *is the transmitter of the Ca***2<sup>+</sup>** *signal*

Since NCδ belongs to the family of NCS and myristoylation at N-terminus is a characteristic feature of a majority, but not all, of these proteins important for their cellular function, it was necessary to check whether myristoylation was required for NCδ to transmit the Ca2<sup>+</sup> signal for ANF-RGC activation. Reconstitution experiments of ANF-RGC with both myristoylated and non-myristoylated NCδ showed that only the myristoylated NCδ stimulated the cyclase activity whereas the non-myristoylatedform was ineffectual (Duda et al., 2012b). Importantly, both forms exhibited comparable affinity for ANF-RGC. In addition to activating ANF-RGC myristoylated NCδ also lowered the cyclase's KM for substrate GTP and increased its catalytic efficiency, *k*cat, from 6.5 ± 0.3 to 41.4 ± 0.5 pmol cyclic GMP/s.

The biochemical and homology based modeling studies indicate that the secondary structure of the functional form of all membrane guanylate cyclases is homodimeric (Rondeau et al., 1995; Liu et al., 1997; Venkataraman et al., 2008). The contact points for their homo-dimeric formation reside in their extracellular domain (Misono et al., 2011) and in the intracellular domain within the highly conserved dimerization domain (Wilson and

Chinkers, 1995) and core catalytic core domain (Venkataraman et al., 2008). The X-ray crystallographic studies have demonstrated that NCδ also exists as a dimer (Vijay-Kumar and Kumar, 1999). Thus, it was reasonable to expect that the Ca2+-modulated functional unit is NCδ dimer and ANF-RGC dimer.

This expectation was validated experimentally. When monomeric and dimeric forms of myristoylated NCδ were individually tested in reconstitution experiments for their abilities to stimulate ANF-RGC catalytic activity only the dimer was effective (Duda et al., 2012b). The stimulation by the monomeric form was only marginal, possibly resulting from spontaneous dimerization of the monomers when higher NCδ concentrations were used (Duda et al., 2012b). Thus indeed the functional Ca2<sup>+</sup> signal transduction unit is composed of one NCδ dimer and one ANF-RGC dimer.

#### *Neurocalcin* **δ** *targets directly the catalytic domain of ANF-RGC*

The ANF-RGC fragment, aa 788–1029, encompassing the core catalytic domain, I820–G1029, was expressed as a soluble protein and tested for its basal and Ca2+-dependent modulated *via* NCδ activities. The protein exhibited intrinsic guanylate cyclase activity (18 <sup>±</sup> 4 pmol cyclic GMP min−<sup>1</sup> mg protein−1). This activity increased when Ca2<sup>+</sup> and NCδ were added, and the increase was Ca2<sup>+</sup> and NCδ-dose dependent (Duda et al., 2012b). Interestingly, the estimated EC50 for NCδ was comparable to that determined for full-length ANF-RGC strongly supporting the expectation that the NCδ signaling site resides within the catalytic domain.

The NCδ targeted site on ROS-GC1 was mapped to the aa segment V837–L858. The corresponding site on ANF-RGC, 849DIVGFTALSAESTPMQVVTLLMQ871, has 70% sequence conservation in comparison with ROS-GC1. When a synthetic peptide of this sequence was used in a functional interference experiment almost complete inhibition of the NCδ-stimulated ANF-RGC activity at 200 μM with an IC50 value of 80 μM was observed (Duda et al., 2012b). Because a scrambled peptide did not exhibit any inhibitory effect it was justified to conclude that the ANF-RGC region 849DIVGFTALSAESTPMQVVTLLMQ871 mediates NCδdependent Ca2<sup>+</sup> stimulation of ANF-RGC activity. This region is a part of the core catalytic domain and common to the corresponding sites in other membrane guanylate cyclases28, it has a secondary structure of helix-loop-helix and is acidic in nature with a pI of 3.37 (Duda et al., 2012b).

## *The effects of ANF/ATP and Ca***2+***-neurocalcin* **δ** *signaling of ANF-RGC activity are multiplicative*

To determine the liaison between ANF/ATP and Ca2+-NCδ signaling modes, ANF-RGC activity was analyzed first in the presence of 1 μM Ca2<sup>+</sup> and 2 μM myristoylated NCδ and then with increasing concentrations, ranging from 10−<sup>11</sup> M to 10−<sup>6</sup> M, of ANF and constant 0.8 mM ATP. The presence of Ca2<sup>+</sup> and NCδ resulted in stimulation of the cyclase activity approx. 3.5 fold above the basal activity. Addition of ANF and ATP resulted in additional 4.5-fold stimulation, demonstrating that the Ca2+-NCδ and ANF/ATP effects are multiplicative (Duda et al., 2012b). It is noteworthy that in the absence of Ca2<sup>+</sup> in the reaction mixture only the ANF/ATP-dependent stimulation of ANF-RGC activity was observed.

The preceding narration on the mechanisms involved in the hormone- or Ca2+-NCδ-dependent stimulation of ANF-RGC activity explains the independence and multiplicativeness of the ANF and Ca2+-NCδ signals in the activation of ANF-RGC. It is based on our evolving conceptual scheme whose central idea is that the functional specificity of a guanylate cyclase is determined by the structure of its modular blocks. The structural motif, 669WTAPELL675 is involved in transmitting the ANF signal to the catalytic domain (Duda et al., 2009) but it is not involved in transmitting the Ca2<sup>+</sup> signal to the catalytic domain; instead the 849DIVGFTALSAESTPMQVVTLLMQ871 motif is involved (Duda et al., 2012b).

#### **PHYSIOLOGICAL VALIDATION OF THE TWO SIGNALING MECHANISMS OF THE ANF-RGC ACTIVATION**

The two described above signaling mechanisms of ANF-RGC activation are depicted in **Figure 3**. It is obvious from their description that they were developed based solely on biochemical observations. And although biochemically well-proven, they would have to be considered only hypothetical until their validity was confirmed *in vivo*. And this validation is narrated below.

#### **SIGNALING THROUGH THE 669WTAPELL675 MOTIF**

In an *in vitro* reconstituted system the 669WTAPELL675 motif is critical for the transmission of the ANF/ATP signal and activation of the ANF-RGC catalytic domain and its absence results in unresponsiveness of the mutated ANF-RGC to the signal so if the same happens *in vivo*, an animal model in which the sequence coding for the 669WTAPELL675 motif is deleted from the ANF-RGC gene should be unresponsive to ANF and therefore exhibit all physiological consequences of this unresponsiveness. Following this logic, the 669WTAPELL675 sequence was deleted in genetically altered mice (Duda et al., 2013). It is worth mentioning that up till now it is the only ANF-RGC domain-specific genetically modified animal model.

In agreement with the *in vitro* determinations, basal ANF-RGC activity in the membranes of the heart, adrenal gland and kidney, the primary tissues where ANF-RGC is expressed, was practically indistinguishable between the wild type control (ANF-RGC669WTAPELL675+/+), heterozygous (ANF-RGC669WTAPELL675+/−) and homozygous (ANF-RGC669WTAPELL675−/−) mice (Duda et al., 2013). However, the mutated cyclase expressed in these tissues did not respond to ANF/ATP stimulation. The cyclase activity in the membranes of the homozygous mice was the same when assayed in the presence of 10−<sup>7</sup> M ANF and 0.8 mM ATP as when assayed in their absence. Interestingly, for the heterozygous mice, the cyclase remained responsive to ANF/ATP but the maximal achieved activity oscillated around 50% of the wild type values, clearly showing that in the heterozygous mice where the product of one ANF-RGC gene copy is of the wild type and the other, of the deletion-mutated cyclase, only the wild type ANF-RGC is responsive to ANF/ATP

**FIGURE 3 | (A)** Topography of ANF-RGC. The dashed lines on the right show the boundaries of: LS, leader sequence; ExtD, extracellular domain; TM, transmembrane domain; ICD, intracellular domain. The functional domains in ICD, their names and the aa constituting their boundaries are indicated at the left: JMD, juxtamembrane domain; ARM, the ATP regulated module; SHD-signaling helix domain; CCD-core catalytic domain. The site targeted by

NCδ (encircled) is within CCD. **(B)** The signaling pathways of ANF and of NCδ are independent. The trajectory of the ANF pathway is in red dashed arrow. From the ExtD, it passes through the TM, ARM and SHD in its course to CCD. The trajectory of the NCδ pathway (in blue dashed arrow) is within the CCD. The CCD exists as an antiparallel homodimer (reproduced with permission from ref. Duda et al., 2012b).

and the mutant is not (Duda et al., 2013). Thus, the *in vitro* activity characteristics were the mirror images of these *in vivo*.

Almost identical values of basal guanylate cyclase activity in various tissues of the 669WTAPELL675 targeted mice and their isogenic controls implied that the expression of the mutant-cyclase must not differ from the expression of the wild type cyclase. This implication was confirmed by the results of immunocytochemical analyses. Side-by-side immunostaining with immunopurified anti ANF-RGC antibody resulted in identical images of the kidneys and adrenal glands of the wild type, heterozygous and homozygous 669WTAPELL675 targeted mice. Also, visual examination of the stained sections as well as of the respective differential interference contrast images indicated that the targeting of the 669WTAPELL675 motif does not change the integrity of these tissues (Duda et al., 2013).

The main function of ANF/ANF-RGC signaling system is to regulate natriuresis, diuresis, vasodilation, and prevent cardiac and renal hypertrophy. It regulates blood volume homeostasis and blood pressure by off-setting the renin-angiotensin-aldosterone system (RAAS): inhibiting renal renin secretion and aldosterone synthesis in the adrenocortical zona glomerulosa (Burnett et al., 1984; Brenner et al., 1990; Maack, 1996; Olson et al., 1998; Aoki et al., 2000; Shi et al., 2001). Therefore, elimination of ANF-RGC signaling through knocking-out either ANF-RGC or ANF gene leads to increased blood pressure, cardiac, and renal hypertrophy, fibrosis, to name some of the related pathological conditions (Dubois et al., 2000; Kishimoto et al., 2001; Ellmers et al., 2007; Das et al., 2010; Ellis et al., 2011; Pandey, 2011; Wadei and Textor, 2012). Since the 669WTAPELL675 targeted mice have partially (heterozygous) and fully (homozygous) suppressed ANF-RGC response to ANF, but the basal cyclase activity is unaffected the critical question was: does this suppression lead to increased blood pressure and other physiological alterations in the genetically modified mice?

The answer obtained was clearly "yes." In comparison with the wild type the blood pressure of the heterozygous mice increased by ∼37% and of the homozygous mice by ∼56%. The increase was statistically highly significant with *P* value < 0.005 (Duda et al., 2013). It rose from 102 ± 9 mm Hg for the wild type mice to 134 ± 17 mm Hg for the heterozygous and 159 ± 11 mm Hg for the homozygous mice. These values showed that the progressive elevation of blood pressure directly correlates with the number of mutated ANF-RGC gene copies.

Because volume homeostasis and blood pressure are influenced by the mineralocorticoid hormone aldosterone secreted by the cells of the adrenocortical zona glomerulosa and the function of the ANF/ANF-RGC system is to inhibit this synthesis the obvious next question was whether the increased blood pressure in the 669WTAPELL675 targeted mice is a consequence of the inability of the mutated ANF-RGC to transduce the ANF signal and to synthesize cyclic GMP sufficient quantities to inhibit aldosterone synthesis? A hint was provided by studies on ANF-RGC knockout mice showing that these mice had increased aldosterone level (Zhao et al., 2007).

The plasma aldosterone concentrations were measured in both types of genetically modified mice (hetero- and homozygous) and in their isogenic controls (wild type). The results demonstrated that, in comparison with the wild type mice, the plasma aldosterone concentration increased by approximately 40% (from 147 ± 12 to 204 ± 18 pg/ml) in the plasma of the heterozygous mice (*P* < 0.005) and by approximately 75% (up to 256±22 pg/ml) in the plasma of the homozygous mice (*P* <0.005) (Duda et al., 2013).

Chronic pressure overload is almost always accompanied by cardiac hypertrophy (reviewed in Katholi and Couri, 2011). The same is observed in the 669WTAPELL675 targeted mice. The ratio of the heart weight (in mg) to whole body weight (in g) measured in 12 weeks old mice was 5.6 ± 0.3 and 6.1 ± 0.5 for the 669WTAPELL675(+/−) and 669WTAPELL675(−/−) mice, respectively, whereas it was 5 ± 0.3 for the wild type mice (Duda et al., 2013). However, hypertension is not the only cause of myocardial hypertrophy. Using animal models it has been shown that antihypertensive drugs do not always ameliorate cardiac hypertrophy (Knowles et al., 2001) and that even without systemic hypertension cardiac hypertrophy may occur (Oliver et al., 1997; Holtwick et al., 2003). Because deletion of the 669WTAPELL675 motif from ANF-RGC results in decreased synthesis of cyclic GMP it was hypothesized that cardiomyocytes of the 669WTAPELL675 targeted mice succumb to the insufficient quantities of cyclic GMP and normal inhibition of myocardial proliferative responses does not occur. Our ongoing studies indicate that indeed it is the case. The mice as young as 3 weeks of age (at weaning) exhibit significant cardiac hypertrophy. Which cyclic GMPdependent signaling pathway involved in inhibition of myocardial proliferative responses is affected in these mice remains to be determined.

Taken together the biochemical and physiological evidence prove that the 669WTAPELL675 motif is critical for the ANF-RGC function as the transducer of the ANF/ATP signal.

#### **NEUROCALCIN δ Ca2+-DEPENDENT SIGNALING OF ANF-RGC**

To demonstrate that the ANF-RGC NCδ-Ca2<sup>+</sup> signal transduction system is functional *in vivo* a mouse model with a disrupted NCδ gene was constructed and analyzed. Unfortunately, disruption of both copies of NCδ gene is lethal. Although it was unexpected at the time of developing the mouse line, it can be now understood in view of the results demonstrating that NCδ may be involved in spermatogenesis. Therefore, heterozygous, NCδ+/−, line is viable.

Because earlier studies from our laboratory and others' had shown that NCδ is expressed in the adrenocortical zona glomerulosa (Nakano et al., 1993; Duda et al., 2012a) and that the adrenal gland also contains functional ANF-RGC transduction system (Takayanagi et al., 1987; Sharma et al., 1989; Duda et al., 1991; Rondeau et al., 1995) the adrenal gland appeared to be the tissue of choice for testing whether the ANF-RGC NCδ-Ca2<sup>+</sup> signal transduction system is operational and of physiological significance there.

To show that NCδ is indeed the Ca2+-sensor modulator of ANF-RGC in the adrenal gland the particulate fractions of the adrenal glands from wild type and NCδ+/<sup>−</sup> mice and their isogenic controls (NCδ+/+) were tested for guanylate cyclase activity in the presence and absence of Ca2+. The activity in membranes isolated in the absence of Ca2<sup>+</sup> was about 66 pmol cyclic GMP min−<sup>1</sup> (mg protein) <sup>−</sup><sup>1</sup> for the control and NCδ+/<sup>−</sup> mice and the activity was not affected by the presence or absence of Ca2<sup>+</sup> in the assay mixture. However, the activity in membranes isolated in the presence of Ca2<sup>+</sup> was strongly dependent on the mice genotype and Ca2<sup>+</sup> in the assay mixture. When assessed in the absence of Ca2+, the activity was <sup>∼</sup>70 pmol cyclic GMP min−<sup>1</sup> (mg protein) <sup>−</sup><sup>1</sup> for the weight and NCδ+/<sup>−</sup> mice but when assessed in the presence of 1 μM Ca2<sup>+</sup> the activity was <sup>223</sup> <sup>±</sup> 20 pmol cyclic GMP min−<sup>1</sup> (mg protein) <sup>−</sup><sup>1</sup> for the wild type mice and 135 <sup>±</sup> 10 pmol cyclic GMP min−<sup>1</sup> (mg protein) <sup>−</sup><sup>1</sup> for the NCδ+/<sup>−</sup> mice. Thus, the Ca2+-dependent NCδ-modulated ANF-RGC signaling pathway in the mice with one copy of NCδ gene deleted (NCδ+/−) is functionally half as active as in the wild type mice. To further authenticate that the lowering of the Ca2+-dependent cyclase activity in the adrenal gland membranes of NCδ+/<sup>−</sup> mice is the exclusive consequence of lower NCδ expression, 2 μM exogenous NCδ was added to the NCδ+/<sup>+</sup> and NCδ+/<sup>−</sup> membranes (isolated in the presence of Ca2+) and the cyclase activity was determined in the presence of 1 μM Ca2+. The cyclase activity in the NCδ+/<sup>+</sup> adrenal membranes increased only minimally, from 220 to 279 ± 21 pmol cyclic GMP min−<sup>1</sup> (mg protein) <sup>−</sup><sup>1</sup> but in the NCδ+/<sup>−</sup> membranes, the increase was significantly larger, from 133 to 284 ± 24 (Duda et al., 2012b). Thus, the activity achieved was practically the same for both types of membranes. Hence, addition of exogenous NCδ to the NCδ+/<sup>−</sup> adrenal gland membranes restores the guanylate cyclase activity and brings it to the level of activity in the NCδ+/<sup>+</sup> membranes. We rationalized that the slight activity increase in the NCδ+/<sup>+</sup> membranes observed upon addition of exogenous NCδ is caused by a partial loss of the native NCδ during the membrane preparation.

What is the function of this pathway in the adrenal gland? In general, the primary role of ANF-RGC in the adrenal gland is to offset the renin-angiotensin system and inhibit aldosterone synthesis, and by doing this, to lower blood pressure (Burnett et al., 1984; Aoki et al., 2000; Shi et al., 2001). Therefore, is the Ca2+-dependent ANF-RGC signal transduction machinery in the adrenal gland involved in aldosterone synthesis? Our ongoing studies indicate that indeed it is. The plasma aldosterone levels in the NCδ+/<sup>−</sup> mice are approximately 27% higher than in the plasma of the control (NCδ+/+) mice. And the effect is exclusive for the aldosterone synthesizing glomerulosa cells, because corticosterone (synthesized in fasciculate cells) levels are unaffected by the absence of NCδ gene.

The increased plasma aldosterone level in the NCδ+/<sup>−</sup> mice correlates with the increase in blood pressure. Systolic blood pressure (measured by the non-invasive tail cuff method) was determined to be 92 ± 6 mm Hg for the wild type mice and <sup>127</sup> <sup>±</sup> 9 mm Hg for the NCδ+/<sup>−</sup> mice. Therefore, the conclusion of these studies was that the ANF-RGC-Ca2+-NCδ transduction system is not only physically present but is of physiological significance at least in the adrenal gland.

In summary, this review has high-lighted studies which define the molecular and physiological mechanisms of the hormonal signal transduction of ANF-RGC. In addition, a new signal transduction mechanism and its present state of validation has been narrated. In this model Ca2<sup>+</sup> is the additional signal of ANF-RGC. By entirely different mechanism, it regulates ANF-RGC catalytic activity and controls its physiological functions.

#### **FUTURE DIRECTIONS**

There are three venues through which, in these authors' understanding, the future research will progress. The first venue is centered on the basic research to decipher (1) how the 669WTAPELL675 motif "communicates" with the catalytic domain of ANF-RGC and signals its activation and (2) the mechanism of NCδ-Ca2<sup>+</sup> signaling of ANF-RGC activity. The second venue relates to the physiology of both ANF-RGC signaling pathways (1) establishing whether the Ca2<sup>+</sup> signaling mechanism is operative in other tissues in addition to the adrenal gland and (2) to determine the interrelationship of the hormonal and Ca2<sup>+</sup> pathways. Finally, after deciphering the molecular and physiological details of the two signaling pathway the third venue will be translational, to design a molecule that can target directly the catalytic domain and bring ANF-RGC to its full activity and prevent the pandemic of hypertension, myocardial hypertrophy and obesity. The last objective is the most far-reaching, but hopefully can be achieved after the first two goals are accomplished.

#### **ACKNOWLEDGMENTS**

The numerous NIH and NSF support (Rameshwar K. Sharma) is acknowledged for the pioneering studies on ANF-RGC signal transduction field and the HL084584 and S82701 support (Teresa Duda) is acknowledged for the development of the 669WTAPELL675 and Ca2<sup>+</sup> signaling mechanisms.

#### **REFERENCES**


guanylyl cyclase (RetGC), RetGC-2. *Proc. Natl. Acad. Sci. U.S.A.* 92, 5535–5539. doi: 10.1073/pnas.92.12.5535


Zufall, F., and Munger, S. D. (2010). Receptor guanylyl cyclases in mammalian olfactory function. *Mol. Cell. Biochem.* 334, 191–197. doi: 10.1007/s11010-009- 0325-9

**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: 13 January 2014; accepted: 25 February 2014; published online: 17 March 2014.*

*Citation: Duda T, Pertzev A and Sharma RK (2014) Atrial natriuretic factor receptor guanylate cyclase, ANF-RGC, transduces two independent signals, ANF and Ca*2+*. Front. Mol. Neurosci. 7:17. doi: 10.3389/fnmol.2014.00017*

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

*Copyright © 2014 Duda, Pertzev and Sharma. 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.*

# Presynaptic [Ca2**+**] and GCAPs: aspects on the structure and function of photoreceptor ribbon synapses

## *Frank Schmitz\**

Department of Neuroanatomy, Institute for Anatomy and Cell Biology, Medical School Homburg/Saar, Saarland University, Saarland, Germany

#### *Edited by:*

Rameshwar K. Sharma, Salus University, USA

#### *Reviewed by:*

Matthias Kneussel, Zentrum für Molekulare Neurobiologie Hamburg, Germany Noga Vardi, University of Pennsylvania, USA

#### *\*Correspondence:*

Frank Schmitz, Department of Neuroanatomy, Institute for Anatomy and Cell Biology, Medical School Homburg/Saar, Saarland University, Kirrbergerstrasse, University Campus, Building 61, Homburg/Saar 66421, Saarland, Germany e-mail: frank.schmitz@uks.eu

Changes in intracellular calcium ions [Ca2+] play important roles in photoreceptor signaling. Consequently, intracellular [Ca2+] levels need to be tightly controlled. In the light-sensitive outer segments (OS) of photoreceptors, Ca2<sup>+</sup> regulates the activity of retinal guanylate cyclases thus playing a central role in phototransduction and light-adaptation by restoring light-induced decreases in cGMP. In the synaptic terminals, changes of intracellular Ca2<sup>+</sup> trigger various aspects of neurotransmission. Photoreceptors employ tonically active ribbon synapses that encode light-induced, graded changes of membrane potential into modulation of continuous synaptic vesicle exocytosis. The active zones of ribbon synapses contain large electron-dense structures, synaptic ribbons, that are associated with large numbers of synaptic vesicles. Synaptic coding at ribbon synapses differs from synaptic coding at conventional (phasic) synapses. Recent studies revealed new insights how synaptic ribbons are involved in this process. This review focuses on the regulation of [Ca2+] in presynaptic photoreceptor terminals and on the function of a particular Ca2+ regulated protein, the neuronal calcium sensor protein GCAP2 (guanylate cyclase-activating protein-2) in the photoreceptor ribbon synapse. GCAP2, an EF-hand-containing protein plays multiple roles in the OS and in the photoreceptor synapse. In the OS, GCAP2 works as a Ca2+-sensor within a Ca2+-regulated feedback loop that adjusts cGMP levels. In the photoreceptor synapse, GCAP2 binds to RIBEYE, a component of synaptic ribbons, and mediates Ca2+-dependent plasticity at that site. Possible mechanisms are discussed.

**Keywords: photoreceptor, ribbon synapse, synaptic ribbon, RIBEYE, Ca2+, GCAP2**

## **INTRODUCTION**

#### **PRESYNAPTIC [Ca2+]: ASPECTS ON THE STRUCTURE AND FUNCTION OF PHOTORECEPTOR RIBBON SYNAPSES**

Changes in presynaptic free cytosolic Ca2<sup>+</sup> ions trigger important functions in the presynaptic terminals. These include synaptic vesicles exocytosis and the mediation of various forms of synaptic plasticity (Bliss and Collingridge, 1993; Malenka and Nicoll, 1999; Neher and Sakaba, 2008; Kawamoto et al., 2012; Rizo and Südhof, 2012; Catterall et al., 2013; Südhof, 2013). Ca2<sup>+</sup> ions also have fundamental functions in photoreceptor synapses. Photoreceptors employ tonically active ribbon synapses that encode light-induced, graded changes of membrane potential into modulation of continuous synaptic vesicle exocytosis to transmit the light-induced signals to the inner retina. The active zones of ribbon synapses contain large electron-dense structures, synaptic ribbons, that are associated with large numbers of synaptic vesicles. Synaptic ribbons have a key importance for signaling

at the ribbon synapse. This review focuses on the regulation of [Ca2+] in presynaptic photoreceptor terminals and on the function of a particular Ca2+-regulated protein, the neuronal calcium sensor (NCS) protein GCAP2 (guanylate cyclase-activating protein-2). GCAP2, an EF*-*hand-containing, Ca2+-binding protein is strongly expressed in photoreceptors and plays multiple roles in the outer segments (OS) and in the photoreceptor synapse.

#### **PHOTORECEPTORS**

Photoreceptors are the principal light-sensitive neurons in the outer retina. They detect incoming light quanta (photons) and communicate the information to the inner retina for further processing. Rod photoreceptors operate at low light intensities and are capable of single-photon-detection. Cone photoreceptors operate at daylight conditions and mediate color vision (Abramov and Gordon, 1994). Both types of photoreceptors display a bipolar morphology with two processes emanating from opposite ends of the photoreceptor cell body (**Figure 1A**). (1) An outer process forms the outer segment (OS) that is responsible for phototransduction. During this process, light-sensitive photopigments in the OS transform light quanta into a change of the photoreceptor membrane potential. (2) An inner cell process emerges from the vitread portion of the photoreceptor soma. This inner process ends in a specialized presynaptic terminal that contains synaptic ribbons in their active zones (**Figures 1A** and **2**). At these

**Abbreviations:** CNG, cyclic nucleotide-gated channel; CRAC, Ca2+-releaseactivated Ca2+-channels; GCAP2, guanylate cyclase-activating protein 2; IS, inner segments; NAD(H), designates both the oxidized (NAD+) as well as the reduced form (NADH) of nicotinamide adenine dinucleotide; OS, outer segments; PMCA, plasma membrane Ca2+-ATPases; NCX, Na+/Ca2+-exchanger; RIM, rab3 interacting molecule; RIM-BP, RIM-binding protein; ret-GC, retinal guanylate cyclase; SERCA, sarco/endoplasmic reticulum Ca2+-ATPase; SOCE, store-operated Ca2+-entry; TIRF, total internal reflection fluorescence; VGCC, voltage-gated Ca2+-channels; RE(A), RIBEYE(A)-domain; RE(B), RIBEYE(B)-domain.

ribbon-type active zones, signals generated by the absorption of light quanta in the OS are continuously transmitted to the inner retina (for review, see Wässle, 2004; Nassi and Callaway, 2009; Lee et al., 2010).

### **Ca2+-DEPENDENT REGULATION OF PHOTOTRANSDUCTION IN PHOTORECEPTOR OUTER SEGMENTS**

The OS contain dense and regular intracellular staples of membrane disks (for rods, **Figure 1**; for review, see Mustafi et al., 2009). These membrane-specializations are enriched with light-sensitive opsin-containing photopigments (Burns and Baylor, 2001; Chen, 2005). As light quanta are detected by the photopigments in the OS, a signal transduction cascade is initiated that finally leads via G-protein-mediated activation of a phosphodiesterase, phosphodiesterase 6 (PDE6), to the hydrolysis of cGMP (Burns and Baylor, 2001; Chen, 2005; Dell'Orco and Koch, 2010). The drop in cGMP leads to the closure of cyclic nucleotide-gated cation channels (CNG-channels) in the plasma membrane, a non-selective cation channel through which both Na+- and Ca2+-ions can pass from extra- to intracellular space along their electrochemical gradient. The light-induced closure of the CNG-channels generates a hyperpolarization response. The membrane potential ranges from ≈−40 mV in the dark to ≈−70 mV in bright light. The phototransduction system has a high dynamic range that is adjusted – not exclusively, but to a considerable extent – by Ca2+-dependent processes (Arshavsky, 2002; Korenbrot and Rebrik, 2002; Sharma, 2002, 2010; Koch, 2006; Artemyev, 2008; Stephen et al., 2008; Dell'Orco and Koch, 2010; Sharma and Duda, 2012). The light-induced drop of Ca2<sup>+</sup> is sensed by Ca2+-binding, EF-hand-containing proteins, the guanylate cyclase-activating proteins (GCAPs; Korenbrot and Rebrik, 2002; Palczewski et al., 2004; Koch, 2006; Sakurai et al., 2011; Sharma and Duda, 2012). GCAPs are members of the family of NCS proteins (Sharma, 2002, 2010; Koch, 2006; Burgoyne, 2007; Baehr and Palczewski, 2009; Koch et al., 2010; Schmitz et al., 2012). GCAP2 (as well as its close relatives GCAP1 and GCAP3; for review, see Palczewski et al., 2004) contains four EF-hands. The first EF-hand is not able to bind Ca2<sup>+</sup> due to exchange of critical amino acids (Palczewski et al., 2004; Koch, 2006; Burgoyne, 2007; Schmitz et al., 2012). Thus, under light stimulation, when Ca2<sup>+</sup> is low, GCAPs activate the photoreceptor retinal guanylate cyclase (ret-GC) thus restoring cGMP baseline levels (Burns and Baylor, 2001; Sharma, 2002, 2010; Chen, 2005; Sharma and Duda, 2012). In the dark, at elevated Ca2+, GCAPs inhibit ret-GCs activity and prevent overproduction of cGMP. These Ca2+-/GCAP-dependent processes set the dynamic range of phototransduction and keep the OS and phototransduction operational at different levels of photon fluxes (Koch, 2006; Stephen et al., 2008; Sharma and Duda, 2012). GCAPs are myristoylated at their N-terminus. Remarkably, in contrast to other members of the NCS family, GCAPs do not perform a Ca2+-dependent myristoyl switch at the N-terminus (Stephen et al., 2007). Instead, the N-terminal myristoyl-group remains buried in the inside of the molecule both in the Ca2+-bound and Ca2+-free state where it is essential to stabilize the protein (Stephen et al., 2007). Unusually, the C-terminus of GCAP2 performs a Ca2+-dependent conformational change (Peshenko et al., 2004).

#### **THE PHOTORECEPTOR RIBBON SYNAPSE OF THE MATURE RETINA: ULTRASTRUCTURE AND MOLECULAR ANATOMY OF THE SYNAPTIC RIBBON COMPLEX**

Photoreceptors are non-spiking neurons and communicate lightinduced, graded changes of membrane potential into modulations of tonic neurotransmitter release rates at the presynaptic terminals (Heidelberger et al., 2005; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011). For this purpose, photoreceptor terminals are equipped with large active zones to which conspicuous presynaptic specializations, the synaptic ribbons, are attached (for review, see tom Dieck and Brandstätter, 2006; Schmitz, 2009; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011; Schmitz et al., 2012).

The synaptic ribbon is a large, electron-dense structure that is associated with large numbers of synaptic vesicles along its entire surface (**Figure 2**; see also Schmitz, 2009). In photoreceptor ribbon synapses, the synaptic ribbon usually appears as a bar-shaped structure of several 100 nm in length in the transmission electron microscope (Schmitz, 2009; Mercer and Thoreson, 2011; Schmitz et al., 2012). Three-dimensional reconstructions of the synaptic ribbon revealed that this bar-shaped profile is a cross-section of a plate-shaped structure of the synaptic ribbon (**Figures 2** and **3E**; Rao-Mirotznik et al., 1995; for review, see also Schmitz, 2009; Mercer and Thoreson, 2011; Schmitz et al., 2012). Furthermore, the synaptic ribbon is bended in a horseshoe-shaped manner along the invaginated presynaptic plasma membrane of the photoreceptor terminal. The horseshoe-shaped synaptic ribbon can have a depth of about 1 μm (Schmitz, 2009; Schmitz et al., 2012). Rod photoreceptor presynaptic terminals usually possess a single active zone and a single synaptic ribbon; cone presynaptic terminals possess multiple active zones and multiple, though typically smaller synaptic ribbons (**Figure 2**; Wässle, 2004; Schmitz, 2009; Mercer and Thoreson, 2011; Regus-Leidig and Brandstätter, 2012; Schmitz et al., 2012). In photoreceptor ribbon synapses, synaptic ribbons are anchored to the arciform density. The arciform density corresponds to the presynaptic projections of conventional synapses (tom Dieck and Brandstätter, 2006; Schmitz, 2009; Mercer and Thoreson, 2011; **Figure 2**). At the plasma membrane of the active zone, L-type voltage-gated Ca2+-channels (VGCC) are clustered. Ca2+-influx through these channels triggers release at the photoreceptor synapse (see below). The active zone with the attached synaptic ribbon is opposed by the tips of multiple postsynaptic dendritic endings from horizontal and bipolar cells. These dendritic tips are located in a cavity formed by an invagination of the presynaptic photoreceptor terminal. The dendritic tips contain metabotropic glutamate receptors (in case of invaginating ON bipolar cells) or ionotropic glutamate receptors (in case of horizontal cells and OFF bipolar cells; Nomura et al., 1994; De Vries and Schwartz, 1999; De Vries, 2000; Wässle, 2004; De Vries et al., 2006; for review, see Dhingra and Vardi, 2012). Retinal bipolar cells, photoreceptor-like neurons in the pineal gland, inner ear hair cells, and vestibular hair cells also build ribbon synapses in mammals (Moser et al., 2006; Schmitz, 2009; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011; Safieddine et al., 2012; Schmitz et al., 2012).

RIBEYE is a unique and major component of synaptic ribbons (Schmitz et al., 2000, 2012; Wan et al., 2005; Schmitz, 2009; Mercer

**FIGURE 1 | (A)** Photoreceptors show a bipolar morphology. A distal process forms the outer segment (OS) and the inner segment (IS). The OS is the place where phototransduction takes place. The IS represents the metabolic center of the photoreceptor. An inner, vitread process ends in the presynaptic terminal that contains a synaptic ribbon at the active zone. **(B)** Transmission electron micrograph of an outer segment (OS). Densely packed arrays of membrane disks can be observed. The arrows in **(B)** point to individual disks. In the OS, phototransduction

occurs and GCAP proteins perform a Ca2+-dependent feedback on guanylate cyclases. **(C)** Transmission electron micrograph at the junctional zone between OS and IS. OS and IS are connected by the connecting cilum (cc) that organizes mt-dependent vesicle transport into the outer segment (Sung and Chuang, 2010). OS, outer segment; IS, inner segment; m, mitochondrion; cc, connecting cilium linking OS and IS; mt, microtubule; pm, plasma membrane. Scale bars: 50 nm **(B)**, 250 nm **(C)**.

and Thoreson, 2011). RIBEYE is the only known protein specific to synaptic ribbons (Schmitz et al., 2000; Magupalli et al., 2008; for review, see Schmitz, 2009; Mercer and Thoreson, 2011; Schmitz et al., 2012). RIBEYE consists of a unique A-domain that is specific to synaptic ribbons and a B-domain that is largely identical to CtBP2. Using different promoters, RIBEYE and CtBP2 are transcribed from the same gene (Schmitz, 2009). While CtBP2 is expressed in most if not all cells; the expression of RIBEYE expression is limited to cells that form ribbon synapses. Particularly the A-domain contains many interaction sites, at which RIBEYE can

bind to neighboring RIBEYE molecules. These multiple RIBEYE– RIBEYE interactions could allow the assembly of multiple RIBEYE molecules into ribbon-like structures (Magupalli et al., 2008; for review, see Schmitz, 2009; Mercer and Thoreson, 2011; Schmitz et al., 2012). These primitive, spherical ribbon precursors could then assemble into mature, bar-/plate-shaped synaptic ribbons in mature photoreceptors (see below). RIBEYE appears to be the main building block of synaptic ribbons. Additional proteins are needed to generate mature, functional synaptic ribbons (see below).

**FIGURE 2 | Photoreceptor ribbon synapses.** In **(A,C)** rod photoreceptor ribbon synapses are shown by transmission electron microscopy; in **(B,D)** cone photoreceptor ribbon synapses. **(A,B)** The large presynaptic terminals are filled with numerous synaptic vesicles (sv). The active zone is characterized by specialized presynaptic densities, the arciform densities (ad). The synaptic ribbon (sr) is anchored to the arciform density. The synaptic ribbon is associated with large numbers of synaptic vesicles. Opposite to the active zones are the dendritic tips of horizontal cells (hc) and bipolar cells (bc)

Both RIBEYE(B)-domain and CtBP1/CtBP2 bind NAD(H) (Schmitz et al., 2000). In case of CtBP1/CtBP2, binding of NADH regulates interaction with nuclear proteins (Fjeld et al., 2003). This NAD(H)-regulated interaction can switch on or off distinct target genes relevant, e.g., for cellular migration, adhesion, and apoptosis (Paliwal et al., 2006, 2007, 2012; Chinnadurai, 2009). In the synapse, NAD(H) promotes interaction of RIBEYE with other synaptic proteins, e.g., the NCS protein GCAP2 (Venkatesan et al., 2010; López-del Hoyo et al., 2012). RIBEYE–GCAP2 interaction appears to be relevant for mediating Ca2+-dependent effects on synaptic ribbon dynamics (see below). Other protein– protein interactions of RIBEYE do not depend on NADH, e.g., RIBEYE–Munc119 interaction (Alpadi et al., 2008) or have not been analyzed for their NADH-dependence, e.g., in case of RIBEYE-Bassoon (tom Dieck et al., 2005). While RIBEYE appears to be the only unique and major protein of the synaptic ribbon, there are other ribbon proteins which are also present in the active zone of non-ribbon containing synapses (for review, see tom Dieck and Brandstätter, 2006; Schmitz, 2009; Mercer and Thoreson, 2011; Schmitz et al., 2012). Bassoon is localized in

that contain ionotropic and metabotropic glutamate receptors for signaling. The dendritic tips are located within an invagination of the presynaptic terminal. Cone synapses **(B,D)** are larger in diameter than rod synapses and contain multiple active zones (dashed circles) and multiple synaptic ribbons (sr). sr, synaptic ribbon; sv, synaptic vesicles; ad, arciform density; pre, presynaptic terminal; m, mitochondrium; hc, dendritic tips of horizontal cells; bc, dendritic tips of bipolar cells; pm, presynaptic plasma membrane. Scale bars: 400 nm **(A)**, 1 μm **(B)**, 150 nm **(C)**, 800 nm **(D)**.

the active zone of photoreceptors, bipolar cells, and hair cells (in addition to the active zone localization of conventional synapses) and is considered as an important anchor for the synaptic ribbon in ribbon synapses. Bassoon knockout mice display non-plasma membrane-anchored "floating" ribbons in photoreceptor and hair cells (Khimich et al., 2005; tom Dieck et al., 2005), though not in retinal bipolar cells. Bassoon is also present in retinal bipolar cells (Deguchi-Tawarada et al., 2006). Therefore, additional anchoring mechanisms for the synaptic ribbon might exist. Bassoon was also been reported to be linked to presynaptic Cav-channels recently (Nishimune et al.,2004,2012; see also below). Piccolo, like bassoon a very large multidomain-protein, is present both in conventional and ribbon-type synapses. Recently, a splice variant of piccolo was described, named piccolino, that is preferentially expressed at ribbon synapses (Regus-Leidig et al., 2013). Proteins known to be essential for synapse function in conventional synapses are the multi-domain rab3-interacting molecule (RIM) protein family (Wang et al., 1997; Kaeser et al., 2011, 2012; for review, see Südhof, 2013). These proteins have been localized to the synaptic ribbon. The long isoforms of RIM contain a N-terminal Zn-finger, central

PDZ-domain and two C2-domains (Südhof, 2012a,b, 2013) that enable them to perform key synaptic functions. RIM proteins control synaptic vesicle docking and vesicle priming as well as Ca2+-channel localization (for review, see Südhof, 2012b, 2013). Their function in the ribbon synapse is less clear. RIM2 appears to be localized at the base of the synaptic ribbon and RIM1 at the body of the synaptic ribbon (tom Dieck et al., 2005). RIM proteins interact with Munc13 proteins in conventional synapses to mediate vesicle priming (Deng et al., 2011). While in conventional synapses, RIMs interact with Munc13-1; Munc13-1 is missing in photoreceptor synapses (Schmitz et al., 2001; Cooper et al., 2012). Instead of Munc13-1, ubMunc13-2 is expressed in photoreceptor synapses (Cooper et al., 2012). Other active zone proteins [e.g., CAST/ELKS; RIM-binding proteins (RIM-BPs)] are also present at photoreceptor ribbon synapses partly differentially distributed in distinct sub-compartments of the synapse (for review, see tom Dieck and Brandstätter, 2006; Mercer and Thoreson, 2011; Schmitz et al., 2012).

#### **SYNAPTIC RIBBONS: FUNCTIONAL CONSIDERATIONS**

The synaptic ribbon is the most prominent structural feature of photoreceptor ribbon synapses. The function of the synaptic ribbon is still unclear. The synaptic vesicles attached to the synaptic ribbon have been functionally correlated to different synaptic vesicle pools (for review, see Heidelberger et al., 2005; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011). Total internal reflection fluorescence (TIRF) microscopy studies of individual synaptic vesicles demonstrated that the base of the ribbon is a hotspot of synaptic vesicle exocytosis (Zenisek et al., 2000; Chen et al., 2013). The basal row of synaptic vesicles attached to the synaptic ribbon represents the vesicle pool that can be immediately released (the "RRP," readily releasable pool). Therefore, one role of the ribbon might be to position this basal row of synaptic vesicles in a "strikedistance" for SNARE proteins to rapidly execute Ca2+-triggered exocytosis (Zenisek et al., 2000). The synaptic ribbon seems to be needed for both fast, synchronous release as well as for slow, asynchronous release components (Khimich et al., 2005; Zenisek, 2008; Frank et al., 2010 Snellman et al., 2011).

Exocytosis was frequently observed at ribbon-containing sites if exocytosis was induced by mild stimuli, e.g., weak depolarizations (Chen et al., 2013). If longer depolarizations were applied, exocytosis also occurred at non-ribbon-containing sites, i.e., in some distance from the presynaptic Ca2+-channels (Midorikawa et al., 2007; Zenisek, 2008; Chen et al., 2013). Exocytosis of synaptic vesicles at these non-ribbon-containing sites could contribute to the asynchronous release component (Midorikawa et al., 2007; Zenisek, 2008).

The large dimensions of the synaptic ribbon allow many vesicles to be docked at the base of the synaptic ribbon (Matthews and Fuchs, 2010; Mercer and Thoreson, 2011; Schmitz et al., 2012). This is quite likely necessary to support coordinated multivesicular vesicle fusion at the synaptic ribbon (Glowatzki and Fuchs, 2002; Singer et al., 2004; Khimich et al., 2005; Graydon et al., 2011). The synaptic vesicles attached to the synaptic ribbon in a more distant position (also denoted as "tethered vesicles"; Heidelberger et al., 2005; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011) could contribute to the synaptic vesicle pool that is released with

a slower kinetics, the slowly releasable pool ("SRP"). The tethered synaptic vesicles could possibly move down toward the base of the synaptic ribbon before they finally fuse at the active zone ("conveyor belt" hypothesis of the synaptic ribbon; Lenzi and von Gersdorff, 2001; Parsons and Sterling, 2003).

Recently, it was shown that the synaptic ribbon has also an important role in the replenishment of synaptic vesicles (Jackman et al., 2009; Snellman et al., 2011; Chen et al., 2013; see also below). Regulation of synaptic vesicle replenishment could be a major function of synaptic ribbons (Jackman et al., 2009; Chen et al., 2013). It was shown that tonic release at the photoreceptor ribbon synapse leads to vesicle depletion at the base of the ribbon. Based on that suggestion, the ribbon would work as a capacitor for synaptic vesicles that reloads the empty release sites at the base of the synaptic ribbon (Jackman et al., 2009; Snellman et al., 2011; Chen et al., 2013). The ribbon capacitor will be re-charged with synaptic vesicles in the light phase when vesicle exocytosis is low. This recharged ribbon capacitor will provide release-ready vesicles in the dark when the photoreceptor depolarizes (Jackman et al., 2009; Chen et al., 2013). According to this hypothesis, the rate of release is determined by the rate of vesicle replenishment at the synaptic ribbon. In agreement with this assumption, it was recently shown that major endocytic proteins were strongly enriched in the periactive zone around the synaptic ribbon (Wahl et al., 2013).

### **SYNAPTIC CODING AT RIBBON SYNAPSES IN COMPARISON TO CODING IN CONVENTIONAL SYNAPSES**

Synaptic coding in ribbon synapses differs from coding in conventional synapses. Conventional synapses are typically phasic synapses that initiate synaptic vesicle exocytosis only when a stimulus, usually an action potential, reaches the presynaptic terminal. The depolarization of the presynaptic terminal opens VGCC, typically P-/Q-/N-type calcium channels in phasic synapses (for review, see Catterall et al., 2005; Striessnig et al., 2010; Catterall, 2011; Südhof, 2012a,b, 2013). The subsequent Ca2+-influx initiates synaptic vesicle fusion at the active zone (for review, see Südhof, 2012a,b, 2013). In contrast, ribbon synapses are tonically active synapses that continuously fuse synaptic vesicles with the presynaptic plasma membrane. The basal rate of synaptic vesicle exocytosis is modulated in response to graded changes of the membrane potential. In photoreceptors, exocytosis is high in the dark when the photoreceptor is depolarized. The rate of exocytosis is diminished in the light when the photoreceptor hyperpolarizes (Jackman et al., 2009). Synaptic vesicle exocytosis at ribbon synapses is also initiated by influx of Ca2<sup>+</sup> through VGCC. But ribbon synapses typically use other Ca2+-channels than phasic synapses (see below). As mentioned above, exocytosis of synaptic vesicles preferentially occurs at the base of the synaptic ribbon (Zenisek et al., 2000; Chen et al., 2013). Ribbon synapses can fuse several hundreds of synaptic vesicles per second and ribbon (Heidelberger et al., 2005; Schmitz, 2009). Rod photoreceptors maintain continuous exocytosis at physiological membrane potentials indicating a high vesicle release probability in ribbon synapses (Jackman et al., 2009; Chen et al., 2013). How this is achieved is not clear but could be based on a high Ca2+-sensitivity of the release machinery and/or an

efficient coupling of the release machinery to the presynaptic L-type Ca2+-channels (see below). In contrast, release probability in conventional, phasic synapses is typically much lower than in ribbon synapses (for review, see Nicoll and Schmitz, 2005; Rusakov, 2006; Fioravante and Regehr, 2011; Regehr, 2012).

At the single vesicle level, using TIRF microscopy it was shown that synaptic vesicles did not fuse immediately after they appeared at the plasma membrane but paused at the base of the ribbon for about 60 ms before fusion finally occurred (Zenisek et al., 2000; Chen et al., 2013). Sixty-millisecond is a relatively long time delay and indicates that the speed of single vesicle exocytosis cannot be the main determinant for synaptic coding and its temporal fidelity at ribbon synapses.

The retrieval of synaptic vesicles has been recently appreciated as an important aspect of signaling at ribbon synapses (see also above). Continuous exocytosis in photoreceptor ribbon synapses caused depletion of synaptic vesicles at the base of the synaptic ribbon at physiological membrane potentials (Jackman et al., 2009; Chen et al., 2013). This depletion of presynaptic release sites during tonic exocytosis at physiological membrane potentials suggested that synaptic vesicle replenishment is ratelimiting for exocytosis and an important aspect for signaling at ribbon synapses (Jackman et al., 2009; Chen et al., 2013). The rate of vesicle replenishment would thus determine how smooth exocytosis could proceed. Gaps in continuous synaptic vesicle exocytosis at the ribbon synapse, e.g., as a result of insufficient vesicle replenishment, could contribute important signaling informations.

Despite the large physiological differences between ribbon synapses and conventional synapses, these synapses do not differ strongly in the composition of the exocytotic machinery (for review, see Schmitz, 2009; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011; Schmitz et al., 2012). In photoreceptor ribbon synapses, syntaxin 1 is replaced by syntaxin 3b (Curtis et al., 2008, 2010); complexin 1/2 by complexin 3/4 (Reim et al., 2005, 2009); Munc13-1 by ubMunc13-2 (Schmitz et al., 2001; Cooper et al., 2012); piccolo by piccolino (Regus-Leidig et al., 2013). But the key players of exocytosis are remarkably conserved.

It should be mentioned that ribbon synapses are not a uniform population of synapses but display both structural and functional differences (Heidelberger et al., 2005; Moser et al., 2006; Matthews and Fuchs, 2010; Mercer and Thoreson, 2011). Exocytosis in ribbon synapses of hair cells in the inner ear operates in a fundamentally different manner and appears to function without neuronal SNARE proteins (Nouvian et al., 2011).

#### **DEVELOPMENT OF THE SYNAPTIC RIBBON COMPLEX**

Photoreceptor ribbon synapses of the mouse retina develop postnatally (Olney, 1968; Blanks et al., 1974). The assembly of the synaptic ribbon complex in photoreceptors occurs via distinct steps. Starting around postnatal day 4 (P4) small, spherical ribbon precursors are anchored at the arciform density of the photoreceptor active zone as judged by electron microscopy and immunofluorescence microscopy with antibodies against RIBEYE (Olney, 1968; Hermes et al., 1992; Regus-Leidig et al., 2010a,b; Liu et al., 2013; Zabouri and Haverkamp, 2013). Starting around P10 and proceeding with the time of eye opening, the immature, short spherical ribbons develop into the mature form of the synaptic ribbon which is bar-shaped in cross sections and plate-shaped in three-dimensional representations (**Figures 3C,F**; Hermes et al., 1992;Vollrath and Spiwoks-Becker,1996;Adly et al.,1999; Schmitz et al., 2006; Regus-Leidig et al., 2010a,b; Liu et al., 2013).

Recent studies demonstrated that VGCC have a key role for the developmental organization of the presynaptic cytoskeleton including the synaptic ribbons.

#### **SYNAPTIC RIBBONS AND VOLTAGE-GATED Ca2<sup>+</sup> CHANNELS**

L-type VGCC are clustered at the basal end of the synaptic ribbon (Heidelberger et al., 2005; Doering et al., 2007; Striessnig et al., 2010; Mercer and Thoreson, 2011; Mercer et al., 2011a; Schmitz et al., 2012). The site is a hotspot of exocytosis (Zenisek et al., 2000). Entry of Ca2<sup>+</sup> through L-typeVGCC is essential for triggering exocytosis at the photoreceptor synapse (Dacheux and Miller, 1976; Schmitz andWitkovsky,1997;Witkovsky et al.,1997; Heidelberger et al., 2005). The VGCC at the active zone of photoreceptors consists of a pore-forming transmembrane α1 subunit (Cav1.4, in photoreceptors; Cav1.3 in inner hair cells), an auxiliary β2-subunit and the α2δ4-subunit (Ball et al., 2002, 2011; Catterall et al., 2005; Morgans et al., 2005; Wycisk et al., 2006; Buraei and Yang, 2010; Striessnig et al., 2010; Catterall, 2011; Mercer et al., 2011a; Schmitz et al., 2012; Liu et al., 2013). Few Cav-channels at the base of the synaptic ribbons are sufficient to drive exocytosis (Brandt et al., 2005; Bartoletti et al., 2011). L-type VGCC in photoreceptors are well suited to support tonical exocytosis at ribbon synapses because these channels do not show Ca2+-dependent inactivation (CDI) and little voltage-dependent inactivation (VDI; Singh et al., 2006; Striessnig et al., 2010; Catterall, 2011; Shaltiel et al., 2012). Furthermore, the Ca2+-sensitivity of release is high and exocytosis is triggered by low, submicromolar Ca2+-concentrations (Rieke and Schwartz, 1996; Thoreson et al., 2004; Heidelberger et al., 2005; Sheng et al., 2007; Choi et al., 2008; Jarsky et al., 2010). In addition to L-type channels, retinal bipolar cells also express transient, T-type VGCC (Pan et al., 2001; Ma and Pan, 2003; Hu et al., 2009).

In the development of the outer retina, Cav-channels associate early with synaptic ribbon precursors and synaptic ribbons (Liu et al., 2013; Zabouri and Haverkamp, 2013). Synaptic ribbons do not form properly, if the L-type calcium channels are missing, e.g., in Cav1.4 mouse knockout models (Mansergh et al., 2005; Chang et al., 2006; Raven et al., 2007; Lodha et al., 2010). In the different available Cav1.4 mouse knockout models, the Cav1.4-channel was shown to exert a strong, severe influence on synaptic ribbon localization and assembly. Some quantitative differences were observed in the different available Cav1.4 mouse knockout models (Mansergh et al., 2005; Chang et al., 2006; Raven et al., 2007; Lodha et al., 2010).

Two recent studies (Liu et al., 2013; Zabouri and Haverkamp, 2013) analyzed in detail the relevance of the Cav-channels for distinct steps of synaptic ribbon assembly at the active zone and for the subsequent maturation of the synaptic ribbon complex (Liu et al., 2013; Zabouri and Haverkamp, 2013). Initially, at early stages of development, immature ribbon precursors are clustered in the

**FIGURE 3 |The synaptic ribbon is a dynamic structure, both in the mature retina (A,B,D,E) as well as in the immature retina during early postnatal retinal development (C,F).** Panels **(A,B)** show disassemblying synaptic ribbons in the mature retina. The mature, bar-shaped synaptic ribbon disassembles via spherical structures, the synaptic spheres (arrowheads in **A,B**). Disassembly into synaptic spheres occurs at the cytosolic, non-plasma membrane-anchored end of the synaptic ribbon (arrowheads in **A,B**). At this end, the synaptic ribbon appears enlarged before the spherical synaptic spheres detach from it **(A,B)**. The spherical synaptic ribbons are still associated with synaptic vesicles via thin connections. But they are no longer anchored to the presynaptic plasma membrane. Panel **(A)** is modified from Schmitz and Drenckhahn (1993) with permission from Springer-Verlag, Berlin, Heidelberg, Germany. **(C)** Assembly of the mature, bar-shaped synaptic ribbon during postnatal development. The arciform density (ad) is clearly visible at the active zone (dashed arrowheads in **C**). In the synapse shown in **(C)**, the assembly of the synaptic ribbon just started. A small piece of the ribbon is already assembled (small black arrow in **C**) and anchored at the arciform density. Most of the ribbon is still present as

immature synaptic spheres (arrowheads) located in vicinity to the small anchored ribbon primer (arrow in **C**). Later in development, the spherical synaptic ribbons (arrowheads) coalesce with the anchored ribbon primer (small black arrow in **C**) to form the mature, bar-shaped synaptic ribbon. RIBEYE–RIBEYE interactions might mediate this process. Panel **(C)** is taken from Schmitz et al. (2006) with permission from PNAS (copyright (2006) National Academy of Sciences, USA). Panels **(D,E)** schematically depict the illumination-induced disassembly of the synaptic ribbon (see also **A,B**). Panel **(F)** summarizes the possible sequence of synaptic ribbon assembly that occurs during postnatal development. sr, synaptic ribbon; ss, synaptic sphere; pre, presynaptic, sv, synaptic vesicle; bc, bipolar cell dendritic tip; hc, horizontal cell dendritic tip; m, mitochondrium; P4, postnatal day 4. Arrowheads in **(A,B)** point to synaptic spheres forming at the distal, non-membrane-anchored end of the synaptic ribbon. Dashed arrows in **(C)** point to the arciform density, which is the presynaptic density in ribbon synapses. The spheres colored in red in **(D–F)** represent synaptic vesicles. Panels **(D–F)** were re-drawn based on figures by Adly et al. (1999) and Spiwoks-Becker et al. (2004). Scale bars: 100 nm **(A–C)**.

terminal and probably anchored to the active zone (Zabouri and Haverkamp, 2013) even if the Cav-channels are missing. But – in contrast to the wildtype – these immature ribbon precursors are not aligned and thus functionally linked with postsynaptic glutamate receptors in the Cav1.4 knockout (Liu et al., 2013). Furthermore, without proper channel activity, these immature ribbon precursors do not mature into the bar-shaped synaptic ribbons but remain immature (Liu et al.,2013; Zabouri and Haverkamp,2013). Interestingly, normal physiological activity of the Cav-channel is needed to transform immature ribbons into mature ribbons. Both loss-of-function mutants as well as gain-of-function mutants (Knoflach et al., 2013; Liu et al., 2013) of the Cav-channels prevented proper maturation of the synaptic ribbon complex (Liu et al., 2013).

Similarly, also knockout models of various proteins linked to the Cav1.4-channel (i.e., β2-subunit of Cav1.4 channels, CaBP4, bassoon, and laminin β2 (for review, see Nishimune, 2012) displayed absent, smaller or non-attached "floating" synaptic ribbons (Libby et al., 1999; Ball et al., 2002; Dick et al., 2003; Haeseleer et al., 2004; Nishimune et al., 2004; Nishimune, 2012). In agreement with these findings, voltage-gated Cav-channels in the inner ear were shown to regulate synaptic ribbon size (Frank et al., 2009, 2010; Sheets et al., 2011, 2012).

Zabouri and Haverkamp (2013) demonstrated that Cavchannels are not only important for the proper development of the synaptic ribbon complex but also for the development of the presynaptic cytoskeleton in photoreceptors. Deletion of Cav1.4 lead to misassembly of many components of the presynaptic photoreceptor cytoskeleton including PSD95, the plasma membrane Ca2+-ATPase (PMCA),Veli3, β-dystroglycan, bassoon, and RIM2. As consequence, the architecture of the synaptic terminal was profoundly altered.

These data argue that the Cav1.4 channel in photoreceptors is a central organizer of the photoreceptor ribbon synapse active zone in general and for the synaptic ribbon complex in particular. Several molecules are candidates that could provide a link between synaptic ribbons and voltage-gated Cav1.4 calcium channels. The RIM multi-domain family of proteins are essential components of the active zones in conventional and ribbon synapses (Wang et al., 1997; Südhof, 2013). In conventional active zones, they are found at presynaptic densities; in ribbon synapses both at the active zone (arciform densities; RIM2; tom Dieck et al., 2005) and also at the proper synaptic ribbon (RIM1, tom Dieck et al., 2005). RIMs are important for various central aspects of synapse function (Deng et al., 2011; Han et al., 2011; Kaeser et al., 2011; for review, see Südhof, 2013). RIM proteins are crucial for anchoring P/Q/N-type VGCC at the active zone in conventional synapses via a direct PDZ-domain interaction (Deng et al., 2011; Han et al., 2011; Kaeser et al., 2011; for review, see Südhof, 2013). RIMs indirectly interact with L-type Ca2+-channels via RIM-BPs (Hibino et al., 2002). Besides immobilizing VGCC, RIMs are essential for vesicle priming via interaction with Munc13 and for synaptic vesicle docking via interaction with rab3/rab27 (Deng et al., 2011; Han et al., 2011; Kaeser et al., 2011; for review, see Südhof, 2013).

RIM1 has also been reported to be associated with the β-subunits (β4, β2a) of VGCC (Kiyonaka et al., 2007; Miki et al.,

2007). This interaction was suggested to suppress inactivation of VGCC (Kiyonaka et al., 2007; Miki et al., 2007). Interestingly, a point mutation in the C2A-domain of RIM1 is responsible for a cone rod-dystrophy (CORD7; Johnson et al., 2003). CAST/ELKS proteins that are also present at the synaptic ribbon complex have been recently reported to be also associated with β-subunits of VGCC (Chen et al., 2011; Kiyonaka et al., 2012). Similarly, bassoon has been described to be associated with β-subunits of VGCC (Chen et al., 2011; Nishimune et al., 2012). The binding of bassoon to β-subunits of VGCC is compatible with the finding that the organization of VGCC was disturbed in inner ear ribbon synapses of bassoon-deficient mice (Frank et al., 2010). Thus, the active zone with the synaptic ribbons appears to be intimately associated with VGCC and organized by these channels by a multitude of different protein–protein interactions.

#### **PRESYNAPTIC Ca2+, GCAP2, AND SYNAPTIC RIBBON DYNAMICS**

In the mature adult mouse retina, synaptic ribbons also underlie changes in size. These changes are illumination-dependent and appear to be regulated by Ca2<sup>+</sup> (Vollrath and Spiwoks-Becker, 1996; Adly et al., 1999; Spiwoks-Becker et al., 2004; Schmitz, 2009; Regus-Leidig et al., 2010a). Several studies observed a lightinduced reduction in the size of synaptic ribbons in the mouse retina (Adly et al., 1999; Spiwoks-Becker et al., 2004; Regus-Leidig et al., 2010a; Fuchs et al., 2013). The bar-shaped ribbons appear to disassemble (and re-assemble) from the free cytosolic, non-membrane-anchored end (**Figure 3**). From this end of the bar-shaped ribbon, spherical ribbons, the synaptic spheres, detach (**Figure 3**). Thus, this illumination-dependent growth and disassembly of the synaptic ribbon appears to employ similar intermediate structures as in development (**Figure 3C**). These spherical intermediates are similar in appearance as the spherical synaptic ribbons of hair cells in the inner ear (for review, see Moser et al., 2006). The synaptic spheres are still associated with numerous synaptic vesicles (**Figures 3A–C**). Spherical ribbons were added to remaining ribbons during darkness (Spiwoks-Becker et al., 2004; Regus-Leidig et al., 2010a).

The light-induced reduction of synaptic ribbon size could be mimicked by lowering intracellular Ca2<sup>+</sup> (Adly et al., 1999; Spiwoks-Becker et al., 2004) suggesting that Ca2<sup>+</sup> is an important factor that regulates ribbon size. In fact, while lowering intracellular Ca2<sup>+</sup> lead to a reduction of synaptic ribbon size, increases of intracellular Ca2<sup>+</sup> resulted in increased ribbon sizes (Spiwoks-Becker et al., 2004; Regus-Leidig et al., 2010a). The genetic background of animals also influences synaptic ribbon dynamics. Synaptic ribbon size plasticity has been reported to be stronger in albinotic mice than in pigmented mice (Fuchs et al., 2013).

Mechanisms of assembly and disassembly of synaptic ribbons in the mature retina are still largely unknown. The NCS protein GCAP2 is involved in the Ca2+-dependent disassembly of synaptic ribbons (Venkatesan et al., 2010; López-del Hoyo et al., 2012). RIBEYE, the main component of synaptic ribbons, binds the NCS protein GCAP2 (Venkatesan et al., 2010). GCAP2 is the only known Ca2+-binding protein of synaptic ribbons. Interaction between RIBEYE and GCAP2 occurs via the B-domain of RIBEYE and the C-terminal region (CTR) of GCAP2 (Venkatesan et al., 2010; for review, see Schmitz et al., 2012). The CTR of GCAP2 is divergent in sequence between GCAP1 and GCAP2. In agreement, GCAP1 does not bind to RIBEYE while GCAP2 does (Venkatesan et al., 2010). Recent ultrastructural analyses, using postembedding immunogold electron microscopy demonstrated that GCAP2 is a component of synaptic ribbons that is directly localized at the synaptic ribbon body (López-del Hoyo et al.,2012). Viral overexpression of GCAP2 in organotypical retina cultures leads to a disassembly of synaptic ribbons (Venkatesan et al., 2010). Similarly, in GCAP2 – overexpressing transgenic mice, synaptic ribbons also tend to disassemble more easily than in the wildtype retinas (López-del Hoyo et al., 2012) although the synaptic ribbon complex as such is normally formed in GCAP2 knockout mice.

Low intracellular Ca2+-concentrations lead to dimerization and also to conformational changes in the CTR of GCAP2 (Olshevskaya et al., 1999; Peshenko et al., 2004). In particular, low intracellular Ca2<sup>+</sup> leads to exposure of the CTR of GCAP2 (Peshenko et al., 2004). In the Ca2+-free form of GCAP2, the CTR is exposed while in the Ca2+-bound state it is not (or much less; Peshenko et al., 2004). Therefore, RIBEYE–GCAP2 interaction is most likely favored by conditions of low intracellular Ca2+, e.g., at illumination. Furthermore, while RIBEYE–GCAP2 interaction is promoted by NAD(H) (Venkatesan et al., 2010), NADH destabilizes the RIBEYE–RIBEYE interaction network (i.e., RE(A)-RE(B) interaction ; Magupalli et al., 2008) which would lead to an overall tendency of reduced RIBEYE–RIBEYE interaction. By this mechanism, depletion of Ca2<sup>+</sup> from GCAP2 could contribute to the disassembly of synaptic ribbons in photoreceptor synapses. By this way of thinking, GCAP2 binding to the synaptic ribbon could thus be involved in the regulation of the Ca2+-dependent dynamics of synaptic ribbons.

The conformational change of GCAP2 induced by low Ca2<sup>+</sup> could thus lead to a severing and subsequent disassembly of the ribbon body into spherical structures, the synaptic spheres (**Figure 4**). The immunogold electron microscopic studies by López-del Hoyo et al. (2012) showed that GCAP2 is not homogeneously distributed along the entire ribbon body (as is RIBEYE; Schmitz et al., 2000), but is clustered at hotspots at the synaptic ribbon surface. These hot spots of GCAP2 immunoreactivity on the synaptic ribbon could represent sites of ribbon disassembly (**Figure 4**).

#### **INTRACELLULAR MODULATION OF Ca2+-SIGNALS IN PHOTORECEPTOR TERMINALS**

Ca2<sup>+</sup> signals generated by influx through VGCC are modulated by various intracellular mechanisms that can subsequently increase or decrease cytosolic Ca2+-concentration. The final Ca2+-signal in the presynaptic terminal is shaped by the temporal and spatial interactions of these mechanisms (Krizaj and Copenhagen, 2002; Szikra and Krizaj, 2006).

The smooth endoplasmic reticulum (ER) is a main intracellular Ca2+-source (Parekh, 2011) and located close to the synaptic ribbon in photoreceptor ribbon synapses as judged by immunolabeling with antibodies against ER components, e.g., the sarco(endo)plasmic reticulum Ca2+-ATPase (SERCA)-2 protein (Johnson et al., 2007; Babai et al., 2010). Ca2+-induced Ca2+- release (CICR) from the ER can amplify Ca2+-signals and enhance signal transmission (Krizaj et al., 1999; Krizaj, 2003; Cadetti et al., 2006; Suryanarayanan and Slaughter, 2006; Szikra and Krizaj, 2007; Szikra et al., 2008, 2009; Babai et al., 2010). Several studies showed that CICR contributes to tonic release at the photoreceptor synapse (Krizaj et al., 1999; Cadetti et al., 2006; Suryanarayanan and Slaughter, 2006; Babai et al., 2010). Both classes of ER-localized receptors responsible for mobilization of Ca2<sup>+</sup> from the ER, i.e., ryanodine receptors (RyR) and IP3 receptors (IP3R) have been localized to photoreceptor terminals (Peng et al., 1991; Krizaj et al., 1999; Krizaj, 2003). CICR can also activate Ca2+ activated Cl−-channels at the presynaptic plasma membrane. TMEM16B-type of Ca2+-activated Cl−-channels have been localized to the photoreceptor terminal (Lalonde et al., 2009; Stöhr et al., 2009; Mercer et al., 2011b). Activation of Ca2+-activated Cl−-channels in photoreceptor terminals should lead to depolarization because the membrane potential of about ≈–40 mV at darkness is more negative than the Cl−-equilibrium potential of −20 mV in rod photoreceptor synapses (Thoreson et al., 2002; Babai et al., 2010).

If intracellular Ca2+-stores are exhausted, store-operated calcium entry (SOCE) mechanisms can replenish them (for review, see Parekh and Putney, 2005; Lewis, 2011; Soboloff et al., 2012; Gudlur et al., 2013). The ER-localized EF-hand-containing Ca2+ sensor protein STIM activates Ca2+-selective, plasma membrane located Ca2+-release-activated Ca2<sup>+</sup> (CRAC)-channels in case of store-depletion. CRAC channels contain Orai-proteins as major components that allow Ca2<sup>+</sup> influx through the plasma membrane and subsequent Ca2<sup>+</sup> replenishment of the emptied ER store (Lewis, 2011; Soboloff et al., 2012; Gudlur et al., 2013). TRPC1 proteins also contribute to CRAC channel activity (Szikra et al., 2009; Molnar et al., 2012). Silencing of TRPC1 by RNA interference abolished SOCE in rods but not in cones indicating that TRPC1 could be an important component of SOCE in rods (Szikra et al., 2008). STIM1 proteins have been recently localized to the photoreceptor terminal (Szikra et al., 2009). The SERCA2 ATPase (see below) has been suggested to be the "third element" of SOCE that mediates influx of Ca2<sup>+</sup> into the ER after it has entered the cytosol via the CRAC channels (Manjarrés et al., 2010, 2011; Alonso et al., 2012). Store-operated Ca2+-channels could contribute to sustained Ca2<sup>+</sup> signals and extend the dynamic range of signaling in the tonically active photoreceptor ribbon synapse.

The PMCAs represent a family of transmembrane proteins (Carafoli, 1999; Tempel and Shilling, 2007) that extrudes Ca2<sup>+</sup> from the cytosol into the extracellular space in an ATP-dependent manner (Carafoli, 1999; Tempel and Shilling, 2007). The PMCA is immobilized in the plasma membrane of the presynaptic terminal that is scaffolded by PSD95, MPP4, and VELI3 (Aartsen et al., 2006, 2009; Yang et al., 2007; Förster et al., 2009). Proper localization of PMCA in the presynaptic plasma membrane also depends on VGCC (Xing et al., 2012). The PMCA has a strong influence on presynaptic Ca2+-signaling (Duncan et al., 2006; Szikra and Krizaj, 2006) and also on synaptic ribbon length dynamics (Yang et al., 2007; Aartsen et al., 2009). Deletion of the MPP4 gene leads to a loss of PMCA from the plasma membrane of the presynaptic photoreceptor terminals (Yang et al.,2007;

intracellular Ca2<sup>+</sup> is buffered away by chelators. As a consequence, ribbons

photoreceptor ribbon synapses); EF, EF-hand.

Aartsen et al., 2009). The subsequent loss of extrusion activity leads to increased intracellular Ca2<sup>+</sup> and enlarged synaptic ribbons (Yang et al., 2007; Aartsen et al., 2009). Na+/Ca2<sup>+</sup> exchangers (NCXs) extrude Ca2<sup>+</sup> from the cytosol into the extracellular space under physiological conditions (Lytton, 2007). NCX proteins are closely localized to the synaptic ribbon and thus can be expected to strongly influence synaptic transmission and ribbon dynamics (Duncan et al., 2006; Johnson et al., 2007). PMCA appears to be predominantly expressed in rod spherules; NCX is predominantly expressed in cone pedicles (Johnson et al., 2007). The SERCA can lower intracellular Ca2<sup>+</sup> by pumping Ca2<sup>+</sup> from the cytosol into the ER (Bublitz et al., 2013). In photoreceptor synapses, the SERCA pump has been localized in close proximity to the synaptic ribbon (Yang et al., 2007; Babai et al., 2010). The described Ca2+-extruding systems are important to restore resting levels of Ca2<sup>+</sup> after depolarization-induced increases (Duncan et al., 2006; Szikra and Krizaj, 2006; Johnson et al., 2007; Mercer and Thoreson, 2011; Wan et al., 2012; Bublitz et al., 2013). Photoreceptor terminals often contain also mitochondria in proximity to the synaptic ribbon (Johnson et al., 2007). Mitochondria have been suggested to be able to take up large amounts of Ca2<sup>+</sup> (Rizzuto et al., 2004; Parekh and Putney, 2005; Rizzuto and Pozzan, 2006). In ribbon terminals, Ca2+ uptake by mitochondria was found to be only functionally relevant if the Ca2+-load was particularly high (Zenisek and Matthews, 2000).

#### **THE SYNAPTIC RIBBON SIZE IN PHOTORECEPTOR SYNAPSES: A PROXY FOR PRESYNAPTIC Ca2+-CONCENTRATION?**

As summarized above, there are multiple links between Cavchannels, presynaptic Ca2+-levels, synaptic signaling and the synaptic ribbons. Ca2+-flux through VGCC as well as the subsequent cross-talk to other Ca2+-regulating systems in the presynaptic terminal affect structural plasticity and developmental maturation of the synaptic ribbon in photoreceptor synapses. In the light, when Ca2<sup>+</sup> is low in presynaptic photoreceptor terminals (Thoreson et al., 2004; Choi et al., 2005, 2008; Sheng et al., 2007; Jackman et al., 2009), synaptic ribbons appear to be smaller than in the dark (Adly et al., 1999; Balkema et al., 2001; Spiwoks-Becker et al., 2004; Hull et al., 2006; Regus-Leidig et al., 2010a). Similarly, in knockout mouse models in which extrusion of Ca2<sup>+</sup> is disturbed and presynaptic Ca2<sup>+</sup> increased, synaptic ribbons were longer than in control animals (Yang et al., 2007; Aartsen et al., 2009). On the other hand, in mouse knockout models where Cav-channels were deleted, synaptic ribbons were smaller and did not develop into mature, large synaptic ribbons. Therefore, the size of the synaptic ribbon could appear as a readout of presynaptic Ca2+. A recent publication (Liu et al., 2013) indicated, that the Ca2+-mediated dynamics of synaptic ribbons – at least in development – is more complex. Also gain-of-function mutations of the Cav1.4 channel, which lead to an activation of the Cav1.4 channel already at more negative membrane potentials, displayed shorter than normal synaptic ribbons (Liu et al., 2013). Future investigation need to provide a more detailed picture about the assembly and disassembly of the synaptic ribbon, its regulation and its consequences for signaling at the photoreceptor synapse.

#### **SUMMARY AND OUTLOOK**

Biochemical and molecular analyses identified the protein composition of synaptic ribbons to a large extent. Surprisingly, most of the ribbon proteins are not only present in ribbon synapses but also in conventional synapses. RIBEYE is the only major building block unique to synaptic ribbons. Still, many questions concerning the structure, assembly, and function of synaptic ribbons remain to be elucidated. How is the assembly of synaptic ribbons regulated? How do immature ribbons develop into mature ribbons? How does Ca2<sup>+</sup> exerts its influence on synaptic ribbons? Does Ca2+ dependent phosphorylation of ribbon proteins play a role? Why are there so many links between presynaptic Ca2+-channels and synaptic ribbons? How can very few ribbon synapse-specific components build such a fundamentally different synapse? How are the synaptic ribbon components mechanistically linked to the proposed functions of synaptic ribbons? The application of innovative physiological tools and the analysis of mouse knockout models are needed to answer these questions.

## **ACKNOWLEDGMENTS**

Work of the author was supported by research grants from the German Research Community (DFG) SFB894, GRK1326 and by the Human Frontiers Science Organization HFSP. Thanks to Dr. Jutta Schmitz-Kraemer and Prof. Dr. Doris M. Kraemer (University of Oldenburg) for critically reading the manuscript. The author apologizes that not all relevant original papers could be cited due to limitations in space.

#### **REFERENCES**


Matthews, G., and Fuchs, P. (2010). The diverse roles of ribbon synapses in sensory neurotransmission. *Nat. Rev. Neurosci.* 11, 812–822. doi: 10.1038/nrn2924


and a dihydropyridine-sensitive calcium current. *J. Neurosci.* 17, 7297– 7306.


**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.

*Received: 30 October 2013; accepted: 15 January 2014; published online: 06 February 2014.*

*Citation: Schmitz F (2014) Presynaptic [Ca*2+*] and GCAPs: aspects on the structure and function of photoreceptor ribbon synapses. Front. Mol. Neurosci. 7:3. doi: 10.3389/fnmol.2014.00003*

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

*Copyright © 2014 Schmitz. 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.*

## Insights into the role of RD3 in guanylate cyclase trafficking, photoreceptor degeneration, and Leber congenital amaurosis

## *Laurie L. Molday,Thomas Jefferies and Robert S. Molday\**

Department of Biochemistry and Molecular Biology, Center for Macular Research, University of British Columbia, Vancouver, BC, USA

#### *Edited by:*

Wolfgang Baehr, University of Utah, USA

#### *Reviewed by:*

John Oberdick, The Ohio State University, USA Johann H. Brandstätter, Friedrich-Alexander-Universität Erlangen-Nuremberg, Germany

#### *\*Correspondence:*

Robert S. Molday, Department of Biochemistry and Molecular Biology, Center for Macular Research, University of British Columbia, 2350 Health Sciences Mall, Vancouver, BC V6T 1Z3, Canada e-mail: molday@mail.ubc.ca

Retinal degeneration 3 (RD3) is an evolutionarily conserved 23 kDa protein expressed in rod and cone photoreceptor cells. Mutations in the gene encoding RD3 resulting in unstable non-functional C-terminal truncated proteins are responsible for early onset photoreceptor degeneration in Leber Congenital Amaurosis 12 patients, the rd3 mice, and the rcd2 collies. Recent studies have shown that RD3 interacts with guanylate cyclases GC1 and GC2 in retinal cell extracts and HEK293 cells co-expressing GC and RD3. This interaction inhibits GC catalytic activity and promotes the exit of GC1 and GC2 from the endoplasmic reticulum and their trafficking to photoreceptor outer segments. Adeno-associated viral vector delivery of the normal RD3 gene to photoreceptors of the rd3 mouse restores GC1 and GC2 expression and outer segment localization and leads to the long-term recovery of visual function and photoreceptor cell survival. This review focuses on the genetic and biochemical studies that have provided insight into the role of RD3 in photoreceptor function and survival.

**Keywords: retinal degeneration 3, guanylate cyclase, photoreceptor cells, protein trafficking, photoreceptor degeneration, Leber congenital amaurosis, gene therapy**

#### **INTRODUCTION**

Guanylate cyclase (GC) plays a key role in phototransduction by catalyzing the synthesis of cGMP. In the dark the basal catalytic activity of GC in photoreceptor outer segments (OS) is balanced by the basal activity of phosphodiesterase (PDE) to maintain cGMP at a level sufficient for maintaining a significant fraction of cGMPgated channels in their open state. This allows the influx of Ca2<sup>+</sup> and Na<sup>+</sup> into the OS with the intracellular Ca2<sup>+</sup> level reaching 250–500 nM (Gray-Keller and Detwiler, 1994; Woodruff et al., 2002). At this concentration, guanylate cyclase activating proteins (GCAPs) with bound Ca2<sup>+</sup> maintain GC at its basal level (Gorczyca et al., 1994; Dizhoor et al., 1995; Polans et al., 1996). Following photoexcitation the Ca2<sup>+</sup> level decreases to less than 25 nM through the combined closure of the cGMP-gated channels and continued efflux of Ca2<sup>+</sup> by the Na/Ca-K exchanger (Woodruff et al., 2002). The reduced Ca2<sup>+</sup> concentration triggers the exchange of Ca2<sup>+</sup> for Mg2<sup>+</sup> on GCAPs resulting in the activation GC (Peshenko and Dizhoor, 2006; Dizhoor et al., 2010). The increase in cGMP reopens the cGMP-gated channels and restores Ca2<sup>+</sup> and cGMP to their dark state levels.

Vertebrate photoreceptors contain two membrane-bound GCs, generally called GC1 and GC2, that share a high degree of sequence identity (∼49%) and structural organization (Shyjan et al., 1992; Lowe et al., 1995; Baehr et al., 2007; Karan et al., 2011). They consist of an N-terminal signal sequence followed by an extracellular domain, a transmembrane segment, a kinase homology domain, a dimerization domain, a cyclase catalytic domain, and a carboxyl-terminal extension. GC1 and GC2 exist as homodimers which localize to disk membranes of photoreceptor OS (Yang and

Garbers, 1997; Ramamurthy et al., 2001). GC1 is present at relatively high concentrations in rod and cone OS, whereas GC2 is restricted to rod OS in mice and humans (Baehr et al., 2007; Azadi et al., 2010). The relative amounts of GC1 and GC2 vary with species with a GC1 to GC2 ratio of 4:1 in mouse (Peshenko et al., 2011b) and 30:1 in bovine photoreceptors (Helten et al., 2007; Kwok et al., 2008). The importance of GC1 and GC2 in the visual response and photoreceptor survival has been determined in knockout mice. In GC1 knockout mice, the cone photoresponse is absent and these cells undergo rapid degeneration (Baehr et al., 2007; Karan et al., 2010). Rod photoreceptors remain viable and functional due to the presence of GC2. In the GC1/GC2 double knockout mice, the rod and cone photoresponse is undetectable and both cell types undergo degeneration. The importance of GC1 in rod and cone function and survival is further highlighted by the finding that mutations in the *GUCY2D* gene encoding human GC1 cause Leber Congenital Amaurosis (LCA) Type 1 (LCA1), a severe early onset retinal dystrophy, and cone-rod dystrophy (Perrault et al., 1996; Kelsell et al., 1998; Karan et al., 2010).

Photoreceptor cells are highly polarized neurons with the OS segregated from the rest of the cell by a thin cilium. OS proteins must be efficiently transported from the endoplasmic reticulum (ER) in the inner segment through the cilium since OS turnover every 10 days through the phagocytosis of aged OS membranes by retinal pigment epithelial cells and the addition of new membrane at the base of the OS (Sung and Chuang, 2010). The molecular machinery which orchestrates the transport of proteins to the OS is highly complex involving vesicle

budding, fusion, and motorized transport along microtubules and other cytoskeletal elements in the inner segment and cilium. Molecular-based studies indicate that the C-terminal targeting segment of several OS membrane proteins including rhodopsin and perpherin-2 play an essential role in vesicle trafficking through protein–protein interactions (Tam et al., 2000, 2004; Deretic et al., 2005; Chuang et al., 2007; Mazelova et al., 2009). The vesicle transport pathways, however, appear to be distinct for various OS membrane proteins (Fariss et al., 1997). Efforts to define a sequence required to transport GC to outer segments have been inconclusive (Karan et al., 2011). Recently, retinal degeneration 3 (RD3), a photoreceptor protein encoded by the gene associated with retinal degeneration in the *rd3* mouse, *rcd2* collie, and LCA12 patients has been shown to play a crucial role in the trafficking of GC1 and GC2 in photoreceptors (Azadi et al., 2010). In this review, we focus on the molecular properties of RD3 and its role in GC trafficking, catalytic activity, and photoreceptor degeneration.

## **GENETIC AND PROTEIN ANALYSIS OF THE RD3**

The gene responsible for retinal degeneration in the *rd3* mouse was first identified as an uncharacterized transcript in an *in silico* search of retina-specific transcripts and called *C1orf36* for Chromosome 1 open reading frame 36 (Lavorgna et al., 2003). The gene now known as *RD3* consists of 3 exons spanning the 5 and 3 untranslated region with the open reading frame comprised of part of exon 2 (amino acids 1–99) and exon 3 (100–195). RT-PCR and *in situ* hybridization confirmed the presence of *RD3* in the retina with high expression in the photoreceptors.

The RD3 protein encoded by the *RD3* gene is highly conserved across vertebrates with the human protein consisting of 195 amino acids and sharing 95% sequence identity with other primates, 86% with mouse and rat, 83% with bovine, 67% with chicken, and 50–60% with lower vertebrates including *Danio rerio* (Zebrafish) and *Xenopus tropicalis* (Western clawed frog). Computer algorithms indicate that RD3 lacks any known homology domains or transmembrane segments. RD3 is predicted to have a high α-helix content (∼43%), no β-sheets and considerable disordered regions (**Figure 1**). There are four conserved stretches of predicted α-helices (H1–H4) with the first helix consisting of 34 amino acids and the last helix having 39 amino acids. Additional conserved features include a putative coil–coil region between amino acids 22–54 and several predicted phosphorylation sites.

RD3 has been isolated from retinal homogenates and HEK293 cells and bacteria expressing the recombinant protein (Azadi et al., 2010; Peshenko et al., 2011a). On SDS gels RD3 migrates as a 23 kDa protein consistent with its amino acid sequence. Dynamic light scattering and gel filtration chromatography, however, indicate that purified RD3 in solution exists as a polydispersed oligomeric protein.

### **TRUNCATION MUTATIONS IN RD3 CAUSE PHOTORECEPTOR DEGENERATION**

The *rd3* mouse was one of the first naturally occurring murine strains found to display early onset rod and cone degeneration (Chang et al., 1993). The rate of degeneration varies with the background strain (Linberg et al., 2005; Friedman et al., 2006; Danciger et al., 2008; Molday et al., 2013). The albino RBF/DnJ strain shows the fastest rate of degeneration with only a monolayer of photoreceptor nuclei present 8 weeks after birth. The pigmented In(5) 30RK/J strain displays the slowest rate with some photoreceptors still present after 30 weeks. The Rb(11.13)4Bnr/J strain is intermediate in its rate of degeneration. Rod-derived scotopic electroretinograms (ERG) are measurable in all strains at an early age prior to significant photoreceptor degeneration (∼24– 35 days postnatal), but a cone response has only been reported in the In(5) 30RK/J strain (Friedman et al., 2006; Molday et al., 2013). The mutation which causes photoreceptor degeneration in all strains was first reported by Friedman et al. (2006). They identified a homozygous C→T substitution (c.319C→T) in the *Rd3* gene which causes a premature stop codon resulting in a protein lacking the last 99 amino acids.

An extensive human *RD3* mutational screen was carried out on individuals with autosomal dominant and recessive retinal degenerative diseases. A homozygous G → A transition in the donor splice site at the end of exon 2 was found in two siblings from a family with LCA (Friedman et al., 2006). This mutation causes a premature stop codon following codon 99 of the *RD3* gene. The *RD3* gene was the twelfth gene associated with LCA and accordingly this subclass is called LCA12. Subsequent genetic screens revealed homozygous mutations in *RD3* with severely truncated proteins in other LCA families (Preising et al., 2012; Perrault et al., 2013). World-wide genetic screens indicate that mutations in the *RD3* gene causing LCA are rare. In addition to truncation mutations, missense mutations in *RD3* have been observed in individuals with other retinopathies (Friedman et al., 2006). However, additional studies are needed to determine if these mutations are directly responsible for the disease.

A mutation in the canine *Rd3* gene causes rod-cone dysplasia 2 (*rcd2*) in a strain of collies (Kukekova et al., 2009). An insertion in the gene leads to an alteration in the last 61 codons and a further extension of the open reading frame.

Biochemical studies indicate that the truncated RD3 proteins associated with photoreceptor degeneration in the *rd3* mouse and LCA12 patients are highly unstable and non-functional (Friedman et al., 2006; Peshenko et al., 2011a). These studies support a crucial role of the C-terminal segment in the stability and function of RD3.

## **INTERACTION OF RD3 WITH GC1 AND GC2**

RD3 was first detected in a mass-spectrometry-based proteomic analysis of bovine photoreceptor OS (Kwok et al., 2008). Monoclonal and polyclonal antibodies against RD3 confirmed the presence of the 23 kDa RD3 protein in mouse retinal extracts by western blotting (Azadi et al., 2010). These antibodies, however, proved to be problematic for immunolocalizing RD3 in photoreceptors as they failed to consistently label retinal cryosections above background levels due to either the inaccessibility of the epitopes or the low level of RD3 expression. One polyclonal antibody showed some immunoreactivity in outer and inner segment layers above control retinas, but in subsequent studies this observation could not be reproduced (Azadi et al., 2010). Additional studies

are needed to definitively determine the subcellular localization of RD3 in photoreceptor cells.

The first insight into the possible role of RD3 in photoreceptors was obtained from co-immunoprecipitation experiments. A monoclonal antibody to murine RD3 coupled to Sepharose immunoprecipitated GC1 and GC2 together with RD3 from mouse retinal extracts (Azadi et al., 2010). The interaction of RD3 and GC1 was confirmed in co-expression and co-immunoprecipitation studies. When RD3 and GC1 were co-expressed in HEK293 cells, an anti-RD3 antibody co-precipitated GC1 with RD3 and in a reverse experiment an anti-GC1 antibody co-immunoprecipitated RD3 with GC1 confirming the direct interaction between these proteins (Azadi et al., 2010). Analysis of GC1 deletion mutants further indicated that the C-terminus of GC1 is required for binding of RD3.

## **RD3 IS IMPORTANT FOR THE EXPRESSION AND LOCALIZATION OF GUANYLATE CYCLASES**

To determine the significance of the RD3-GC interaction, the expression and localization of GC1 and GC2 in the *rd3* mouse was compared with age-matched wild-type (WT) mice by western blotting and confocal microscopy (Azadi et al., 2010). GC1 and GC2 from WT mouse retina migrated as 120 and 125 kDa proteins on SDS gels, but were absent or present in reduced amounts in extracts from *rd3* mice depending on the strain used for the analysis (Azadi et al., 2010; Molday et al., 2013). At a subcellular level, GC1 and GC2 was localized to the photoreceptor OS of WT mice with GC1 present in rod and cones and GC2 restricted to rod cells as reported previously (Karan

et al., 2010). In contrast, GC1 and GC2 immunolabeling was absent in 21-day old Rb(11.13)4Bnr/J *rd3* mice and mislocalized to the inner segment of the In(5)30RK/J strain (Azadi et al., 2010; Cheng and Molday, 2013; Molday et al., 2013). GCAP1 and GCAP2 were detected at a reduced level in *rd3* mice and primarily confined to the inner segments as observed in GC1/GC2 knockout mice (Azadi et al., 2010; Karan et al., 2010). Other photoreceptor OS proteins including the cyclic GMP-gated channel, PDE6, peripherin-2, rhodopsin, and cone arrestin targeted normally to the OS of the *rd3* mice. Gene expression observed by microarray analysis indicated that the expression levels of genes encoding RD3, GC1 and GC2 in 21 day old (Rb(11.13)4Bnr *rd3* mice were similar to age-matched WT mice. Up-regulation of endothelin 2 (*EDn2*), glial fibrillary acid protein (*Gfap*), and complement component factor 1 (*Cf1*) and down-regulation of phosducin (*Pdc*), gap junction protein α5 (*Gja5*) and retinal G protein – coupled receptor (*Rgr*) genes, however, were observed for the *rd3* retina (Cheng and Molday, 2013).

The expression and distribution of GC1 was also examined in the In(5)30RK/J strain of *rd3* mice by immunofluorescence microscopy (Molday et al., 2013). GC1 was detected at reduced levels in this strain and confined primarily to the inner segments where it co-localized with the anti-KDEL ER marker. A small amount of GC1 was detected in the OS layer which may accountfor the attenuated scotopic and photopic ERGs observed in this strain. Similar mislocalization of GC in the *rcd2* collie has been observed. When the normal murine *Rd3* gene was delivered to photoreceptor cells of either strain of *rd3* mice using adeno-associated virus

(AAV), GC1 and GC2 expressed at levels comparable to that of WT mice and localized normally to the photoreceptor OS layer (Molday et al., 2013). Collectively, these studies indicate that RD3 is required for efficient expression and localization of GCs to the photoreceptor OS.

The effect of RD3 expression on the localization of GC was also examined in transfected culture cells (Azadi et al., 2010). GC1 primarily localized to the ER of transfected cells in the absence of RD3, whereas RD3 co-localized with Rab11 in endosomes. Coexpression of GC1 and RD3 resulted in the exit of GC1 from the ER and colocalization with RD3 in Rab11 containing vesicles. These studies are consistent with the observed effect of RD3 in promoting the exit of GC from the ER in photoreceptor cells.

#### **RD3 INHIBITS GUANYLATE CYCLASE ACTIVITY**

In addition to altering the expression level and trafficking of GC, RD3 inhibits its catalytic activity (Peshenko et al., 2011a). In the absence of GCAPs, RD3 expressed in either bacteria or HEK293 cells inhibited the basal and GCAP-activated GC catalytic activity at submicromolar concentrations, but did not alter the Ca2<sup>+</sup> sensitivity of GCAPs. It remains to be determined if RD3 plays a significant role in modulating the activity of GCs during phototransduction. It is possible that RD3 only plays an important role in inhibiting cyclase activity during the trafficking of GCs within the inner segments. At present the quantity of RD3 in photoreceptors remains to be determined, but in *in vitro* measurements, RD3 at nanomolar concentrations competitively inhibited the activation of GC by GCAP (Peshenko et al., 2011a).

The effect of the RD3 mutation (F100ter) associated with LCA12 and other possible disease-linked missense mutations on the interaction of RD3 with GC was studied (Peshenko et al., 2011a). The F100ter and the G57V mutants were undetectable when expressed in HEK293 cells, but were expressed at significant levels in *E. Coli*. Other missense mutants (K130M, G35R, R68W, W6R/E23D) expressed at levels comparable to WT RD3 and bound GC1. The F100ter mutant had no effect of the cyclase activity, whereas the missense mutations inhibited GC activity to varying degrees. The W6R/E23D and G35R mutations inhibited the cyclase activity at a similar concentration as WT RD3, whereas the R68W, K130M and G57V were less effective. It remains to be determined if these missense mutations in RD3 alter the ability of RD3 to function in GC trafficking in photoreceptors.

#### **MODELS FOR THE ROLE OF RD3 IN PHOTORECEPTORS**

Studies carried out to date suggest that RD3 serves two roles in photoreceptor cells. A primary function of RD3 is to facilitate the exit of GC from the ER and its trafficking to the photoreceptor cilium (**Figure 2**). RD3 may also be involved in the trafficking of GC within the cilium although this remains to be determined.

A second function of RD3 is to inhibit GC catalytic activity (Peshenko et al., 2011a). The inhibition of GC activity by RD3 may be crucial for insuring that cGMP is not produced in the inner segment of photoreceptors cells during the trafficking of GC to the outer segment. In the absence of RD3 inhibition, GC

#### **FIGURE 2 | Simplified diagram showing the role of RD3 in GC trafficking.** GC synthesized in the endoplasmic reticulum (ER) of

photoreceptor inner segments binds to RD3 on the cytoplasmic side of the ER membrane. This interaction may induce membrane curvature required for the budding of GC-RD3 from the ER in the form of small vesicles. RD3 also inhibits the cyclase activity of GC. The RD3-GC complex together with other adaptor proteins is translocated through the Golgi to the base of the connecting cilium where GC perhaps in the absence of RD3 is transported through the cilium to the base of the outer segments where it is incorporated into disk membranes. At some point RD3 dissociates from GC possibly through phosphorylation or other post-translational modifications or protein–protein interaction. RD3 associated with endosomes is returned to the ER to initiate another cycle of RD3-mediated GC transport.

could either use up the supply of GTP required for the function of other proteins including small G-proteins such as Rabs required for protein vesicle trafficking or overproduction of cGMP in the inner segment could be toxic to photoreceptors. In this regard it is interesting to note that an early study reported that retinal extracts from the *rcd2* collie showed a 10 times higher level of cGMP than control retinas during a 2–8 week period with only 25% reduction in PDE activity (Woodford et al., 1982).

## **ROLE OF RD3 IN PHOTORECEPTOR CELL SURVIVAL**

The mechanism by which loss in RD3 causes photoreceptor degeneration remains to be determined. Here we discuss two possible mechanisms. The first mechanism is centered on the role of GC and cGMP in maintaining calcium homeostasis in photoreceptors. It is generally known that too high or too low Ca2<sup>+</sup> levels can be toxic to cells (Fain, 2006). Calcium homeostasis in OS is maintained by the influx of Ca2<sup>+</sup> through the cGMP-gated channels and the efflux through the Na/Ca-K exchanger. In the absence of GC in the OS, cGMP will not be produced resulting in permanent closure of cGMP-gated channels and a reduction in intracellular Ca2<sup>+</sup> below a threshold required for photoreceptor survival. In this case the mechanism of photoreceptor degeneration in the *rd3* mouse and LCA12 patients would be similar to photoreceptor degeneration in the GC1/GC2 knockout mouse and LCA1 patients. Direct comparison of the rate of photoreceptor degeneration in the *rd3* mouse and GC1/GC2 knockout mouse is complicated by the large variation in the rate of degeneration observed for different strains of *rd3* mice, although in general degeneration in the *rd3* mouse appears to be more rapid. Comparison of LCA12 with LCA1 patients is complicated by the presence of functional GC2 in rod photoreceptors of LCA1 patients. However, limited clinical assessment of these patients suggests that LCA12 is more severe (Preising et al., 2012; Perrault et al., 2013).

A second mechanism involves the role of RD3 in the inhibition of GC activity (Peshenko et al., 2011a). In the absence of RD3, GC in photoreceptor inner segments is likely to be active in catalyzing the production of cGMP (Woodford et al., 1982). Elevated cGMP levels can be toxic through the unregulated activation of cGMP-dependent enzymes. High cGMP levels have been implicated in photoreceptor cell death in the *rd1* and *rd10* mice linked to mutations in the β-subunit of PDE (Bowes et al., 1990; Chang et al., 2007), mice deficient in the A3 subunit of the cone cGMP-gated channel (Xu et al., 2013), and transgenic mice harboring an Y99C mutation in GCAP1 (Olshevskaya et al., 2004). Both the loss in RD3 chaperone activity and inhibition of GC activity may contribute to photoreceptor degeneration.

#### **ACKNOWLEDGMENTS**

This work was supported by grants from the National Institutes of Health (EY02422-33) and the Macular Vision Research Foundation.

#### **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 February 2014; accepted: 29 April 2014; published online: 26 May 2014. Citation: Molday LL, Jefferies T and Molday RS (2014) Insights into the role of RD3 in guanylate cyclase trafficking, photoreceptor degeneration, and Leber congenital amaurosis. Front. Mol. Neurosci. 7:44. doi: 10.3389/fnmol.2014.00044*

*This article was submitted to the journal Frontiers in Molecular Neuroscience. Copyright © 2014 Molday, Jefferies and Molday. 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 andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Membrane guanylyl cyclase complexes shape the photoresponses of retinal rods and cones

## *Xiao-HongWen1\*, Alexander M Dizhoor <sup>2</sup> and Clint L Makino1*

<sup>1</sup> Department of Ophthalmology, Massachusetts Eye and Ear Infirmary and Harvard Medical School, Boston, MA, USA <sup>2</sup> Department of Basic Sciences Research and Pennsylvania College of Optometry, Salus University, Elkins Park, PA, USA

#### *Edited by:*

Wolfgang Baehr, University of Utah, USA

#### *Reviewed by:*

Daniele Bottai, University of Milan, Italy Trevor Lamb, Australian National University, Australia

#### *\*Correspondence:*

Xiao-Hong Wen, Department of Ophthalmology, Massachusetts Eye and Ear Infirmary and Harvard Medical School, Boston, MA 02114, USA e-mail: wenx@meei.harvard.edu

In vertebrate rods and cones, photon capture by rhodopsin leads to the destruction of cyclic GMP (cGMP) and the subsequent closure of cyclic nucleotide gated ion channels in the outer segment plasma membrane. Replenishment of cGMP and reopening of the channels limit the growth of the photon response and are requisite for its recovery. In different vertebrate retinas, there may be as many as four types of membrane guanylyl cyclases (GCs) for cGMP synthesis. Ten neuronal Ca2<sup>+</sup> sensor proteins could potentially modulate their activities. The mouse is proving to be an effective model for characterizing the roles of individual components because its relative simplicity can be reduced further by genetic engineering. There are two types of GC activating proteins (GCAPs) and two types of GCs in mouse rods, whereas cones express one type of GCAP and one type of GC. Mutant mouse rods and cones bereft of both GCAPs have large, long lasting photon responses. Thus, GCAPs normally mediate negative feedback tied to the light-induced decline in intracellular Ca2<sup>+</sup> that accelerates GC activity to curtail the growth and duration of the photon response. Rods from other mutant mice that express a single GCAP type reveal how the two GCAPs normally work together as a team. Because of its lower Ca2<sup>+</sup> affinity, GCAP1 is the first responder that senses the initial decrease in Ca2<sup>+</sup> following photon absorption and acts to limit response amplitude. GCAP2, with a higher Ca2<sup>+</sup> affinity, is recruited later during the course of the photon response as Ca2<sup>+</sup> levels continue to decline further. The main role of GCAP2 is to provide for a timely response recovery and it is particularly important after exposure to very bright light. The multiplicity of GC isozymes and GCAP homologs in the retinas of other vertebrates confers greater flexibility in shaping the photon responses in order to tune visual sensitivity, dynamic range and frequency response.

**Keywords: guanylate cyclase, guanylyl cyclase activating protein, Ca2<sup>+</sup> feedback, Ca2+-binding protein, phototransduction, rod and cone photoreceptors, knockout mouse, neuronal calcium sensors**

### **INTRODUCTION**

Unlike most neurons, retinal rods and cones are partially depolarized while at rest (in darkness). Cation channels in the open state permit an influx of Na<sup>+</sup> and also some Ca2<sup>+</sup> that is termed the "dark" current. When rods and cones receive light, photon capture by visual pigment within the outer segment is amplified by the subsequent activation of many transducin (G-protein) molecules, each of which stimulate phosphodiesterase (PDE) enzymatic activity. Cyclic GMP (cGMP) hydrolysis leads to closure of cyclic nucleotide-gated (CNG) channels thereby blocking the dark current. The ensuing hyperpolarization of membrane potential spreads through the photoreceptor to reduce neurotransmitter release at the synaptic terminal. To recover quickly from stimulation by light, cGMP, the second messenger that links photon capture to CNG channel opening, must be restored (for reviews on phototransduction, see Luo et al., 2008; Wensel, 2008; Gross and Wensel, 2011; Korenbrot, 2012).

Membrane guanylyl cyclases (GCs) catalyze the synthesis of cGMP. As opposed to other membrane GCs, those of photoreceptors do not respond to extracellular ligands, but instead

are subject to regulation by Ca2+-binding, EF-hand bearing, GC activating protein (GCAP) subunits (reviewed in Sharma and Duda, 2012; Koch and Dell'Orco, 2013; Lim et al., 2014). GCAP suppresses synthesis of cGMP by GC in darkness, when intracellular Ca2<sup>+</sup> is relatively high. After light closes the CNG channels and blocks Ca2<sup>+</sup> influx, continued extrusion by Na+/K+, Ca2<sup>+</sup> exchangers lowers intracellular Ca2+. Ca2<sup>+</sup> dissociates from GCAP allowing it to bind Mg2<sup>+</sup> instead, whereupon it stimulates GC activity.

Negative feedback onto cGMP synthesis provided by a Ca2<sup>+</sup> sensing GCAP is likely to have already been in place in very ancient ciliary photoreceptors (reviewed in Lamb, 2013). Two rounds of whole-genome duplications predating the origin of vertebrates could have generated multiple isoforms. A third genome duplication in fish could explain why they have at least four GCs and eight GCAPs (Imanishi et al., 2004), although it is not clear whether all are expressed in photoreceptors (**Figure 1**). The existence of multiple isoforms of GC and GCAP provided a substrate for natural selection to optimize the photon response for particular visual ecologies. Where an overly complex system of GCs and GCAPs was not needed, components were eventually lost. For example in mouse, three frame deletions degraded the GUCA1C gene for GCAP3 into a pseudogene (Imanishi et al., 2002). But multiple GCs and GCAPs were retained in some species and were possibly even supplemented with GC inhibitory protein (GCIP; **Figure 1**), that inhibits GC at high Ca2<sup>+</sup> but does not stimulate it at low Ca2<sup>+</sup> (Li et al., 1998), and unrelated S100 proteins that stimulate GC1 at high Ca2<sup>+</sup> (reviewed by Sharma et al., 2014), to tune GC synthesis in order to meet more challenging demands on vision. Over 400 million years of evolution, GCAP1 and GCAP2 genes are ubiquitously preserved in mammals in a tail-to-tail array (Surguchov et al., 1997). In human, there are two isoforms of GCs, RetGC1, and RetGC2 (Lowe et al., 1995), and two isoforms of GCAPs (GCAP1 and GCAP2) in rods (Dizhoor et al., 1994; Palczewski et al., 1994), while at least some cones express an additional GCAP3 (Haeseleer et al., 1999) and possibly only one type of GC (Karan et al., 2010). Three types of GCs and six types of GCAPs are expressed in zebrafish UV cones (Imanishi et al., 2002, 2004; Rätscho et al., 2009), although it remains to be seen whether all are present in the outer segment (**Table 1**). GCs and GCAPs have responsibilities in the inner segment (reviewed in Karan et al., 2010) and synapse of photoreceptors (reviewed by Schmitz, 2014), both of which constitute separate Ca2<sup>+</sup> compartments. Here, we review the progress that has been made in understanding why so many types of GCs


**Table 1 | Distribution of GCs and GCAPs in photoreceptors of mouse, carp, and zebrafish.**

Results on mouse and carp refer to photoreceptor outer segments, while results on zebrafish were obtained from an analysis of mRNA so the proteins may not necessarily localize to the outer segments. No distinctions are made for cones of different spectral type in mouse or carp. Where the concentrations have not been determined, positive expression is marked with (+), lack of expression is marked with (−), unknown status is marked with (?). Expression of GC2 and GCAP2 is negligible in mouse cones [neither GC2 nor GCAP2 is expressed in the all-cone retina of Nrl−/− mouse Xu et al., 2013]. Expression of GCAPs 4 and 7 is presumed to be absent in carp retina based on exceedingly low mRNA levels. Expression of GCAP4 is moderate in zebrafish UV cones. GCAP7 is present in zebrafish cones but at very low levels. ∗Total concentration of GCAPs 1 and 2 in mouse rods is likely between 3 and 9 μM. From Howes et al., 1998; Imanishi et al., 2002, 2004; Baehr et al., 2007; Rätscho et al., 2009; Takemoto et al., 2009; Peshenko et al., 2011; Scholten and Koch, 2011.

and GCAPs are involved in phototransduction. First, we will consider a relatively simple system that utilizes predominantly two components, GC1 and GCAP1. The impact of genetically deleting the GCAP revealed its role in shaping the response to light. Next, we will describe how this approach was extended to a more complex system that utilizes two additional components. Finally, an extrapolation will be made to other systems in which genetic studies have not yet been carried out or are still in early stages.

#### **A SIMPLE SYSTEM IN MOUSE CONES**

Mouse cones use a simple system for cGMP synthesis with a single type of GC, GC1, coupled almost exclusively to GCAP1 (**Table 1**). To understand how the Ca2<sup>+</sup> feedback onto cGMP synthesis influences the cone photoresponse, it is useful to simplify the system further by genetic engineering. The tacit assumption is that the genetic manipulation "cleanly" removes a targeted protein(s) in an otherwise undisturbed system. Caution is necessary because full verification is impossible and there may be unpredictable consequences. As an illustration, knockout of GC1 in mice interferes with the expression of many cone phototransduction proteins, including transducin, PDE, GCAP1, and arrestin, and the cones eventually degenerate. Surprisingly, small cone ERG responses to light were detected at 1 month of age (Yang et al., 1999), but they soon vanished over the next few weeks as the cones degenerated (Yang et al., 1999; Baehr et al., 2007).

Genetic deletion of GCAPs (GCAPs−/−) may be a more specific experimental perturbation and cones retain good health. **Figure 2A** shows that the single photon response of a GCAPs−/<sup>−</sup> cone rises normally, but continues to do so for twice as long to finally reach a peak after 220 ms that is 2.7-fold higher than normal (Sakurai et al., 2011). Integration time, calculated as the integral of the response divided by response amplitude, increases by 2.3-fold. Loading a WT cone with a Ca2<sup>+</sup> buffer to delay the onset of Ca2+ dependent feedback also prolongs its photon response but such treatment does not affect the flash response of a GCAPs−/<sup>−</sup> cone because GCAP1, the main mediator of the Ca2+-dependent negative feedback onto the phototransduction cascade that dynamically shapes the photon response in mouse cones, is not present.

Since responses to dim steady light summate the responses to individual photons, the large, slow photon responses in GCAPs−/<sup>−</sup> cones should shift the response-intensity relation for steps of light to light intensities (2.7 × 2.3) = 6.2 times lower. There is such a shift for very weak responses, however, GCAPs−/<sup>−</sup> cones are about 20-fold more sensitive than wild type (WT) cones at the half maximal response level (**Figure 2B**; Sakurai et al., 2011). The extra shift suggests that the feedback regulation provided by GCAP1 becomes more vigorous when extended exposures to brighter light cause Ca2<sup>+</sup> to drop lower than the minimum reached during a single photon response. The steeper GCAPs−/<sup>−</sup> response-intensity relation for steps of light shows that GCAP1 normally enables cones to operate over a wider range of intensities.

Thus GCAP1 restricts the growth and quickens the recovery of the single photon response in mouse cones, making them less sensitive but improving their temporal resolution. GCAP1 also extends the operating range of cones by preventing saturation when cones are exposed to bright, steady illumination.

**ranges of cones and rods. (A)** Enlarged, slowed photon response of cones from mice lacking the α-subunit of rod transducin, to render rods unresponsive to light, and both GCAPs (Gnat1−/−, GCAPs−/−). Control response is from Gnat1−/− cones. Mean responses to flashes given at time zero are normalized by the peak amplitude of the control cone response, 0.01 pA. For reasons that are not known, cone response kinetics of this study were slower than those reported by Cao et al. (2014) in **Table 2**. **(B)** Shift in the operating range of GCAPs−/− cones to lower intensities. Fractional current is that present 2 s after light onset. Step intensities are normalized to the mean intensity suppressing half of the dark current for controls, <sup>I</sup>1/<sup>2</sup> <sup>=</sup> 94,300 h<sup>υ</sup> <sup>μ</sup>m−<sup>2</sup> <sup>s</sup>−1. Continuous lines are

#### **GREATER COMPLEXITY IN MOUSE RODS**

In mouse rod outer segments, two GCs coupled to two GCAPs are responsible for cGMP synthesis (**Table 1**). As with mouse cones, a genetic approach is useful for trying to understand how the system works, but with rods, it is necessary to knock out components individually and in combinations to isolate the physiological function of each component (Yang et al., 1999; Baehr et al., 2007; Makino et al., 2008, 2012b). Mutant mice induced to express transgenes for either GCAP1 or GCAP2 on the GCAPs−/<sup>−</sup> background are also informative (Mendez et al., 2001; Howes et al., 2002; Pennesi et al., 2003).

Rods are tasked with counting single photons and are therefore designed to be far more sensitive than cones. Major modifications in rods include slowing of the shutoff and recovery phases of the cascade to grant the photon response time to rise to a large size (reviewed in Luo et al., 2008; Gross and Wensel, 2011; Korenbrot, Sakurai et al. (2011) with permission from Society for Neuroscience. **(C)** Mean single photon responses of WT, GCAP1−/−, GCAP2−/−, and GCAPs−/− rods with amplitudes of 0.47, 1.02, 0.43, and 2.34 pA, respectively, before normalization, includes results from Makino et al. (2008, 2012b). **(D)** Progressive shift in the rod operating range to lower intensities as one or both GCAPs are knocked out. Fractional current is that remaining after 10 s of illumination. The mean I1/<sup>2</sup> for WT rods was 480 hυ μm−<sup>2</sup> s−1. Traces show fits with Hill equation with Hill coefficients of: 1.1, 0.8, 1.0, and 1.0 for WT, GCAP1−/−, GCAP2−/− and GCAPs−/−. **(B,D)** Error bars show SEM.

2012). In mouse, the rod photon response peaks after 140 ms at an amplitude of 0.6 pA, compared to 80 ms and 0.012 pA in cones (**Table 2**). Negative feedback provided by GCAPs is a key factor. The feedback during the single photon response is somewhat more powerful in mouse rods than in mouse cones (**Figure 2**). In the absence of both GCAPs the rod response rises for three times longer than normal to reach an amplitude that is four times larger (Mendez et al., 2001; Burns et al., 2002). The shift in sensitivity for the GCAPs−/<sup>−</sup> rod is to flashes 6–8 times dimmer compared to WT rods, greater than the shift of 2–3 times for cones after knockout of both GCAPs (**Figure 2A**). When GCAP2 alone is missing (GCAP2−/−), the size of the single photon response in rods is normal and the only change is a slightly slower recovery (Makino et al., 2008). Knockout of GCAP1 alone in rods (GCAP1−/−) produces a single photon response roughly twice the normal size (Makino et al., 2012b). So between the two GCAPs, GCAP1 must be the



T = 35–38◦C for mammalian cells, T = 20–22◦C for salamander cells. Amplitude (a), time to peak (tp) and integration time (ti) are for the single photon response, typically obtained by analysis of dim flash responses in the linear range. Values for i1/<sup>2</sup> give the flash strengths at λmax that elicit a half maximal response and are thus inversely proportional to relative sensitivity. Flash responses of different spectral types of mammalian cones do not vary so they were averaged: L, M, S cones for primates; M, S for mouse and squirrel. From Baylor et al., 1984; Schnapf et al., 1990; Perry and McNaughton, 1991; Zhang et al., 2003; Isayama et al., 2006, 2014; Cao et al., 2014. The single photon response amplitude from chipmunk rods studied in Zhang et al. (2003) was provided by T. Kraft, personal communication. Results from mouse were selected from the most recent reports where all parameters were available and in which recording methods were similar:Yang et al., 1999; Mendez et al., 2001; Makino et al., 2004, 2012a,b; Luo and Yau, 2005; Tsang et al., 2006; Krispel et al., 2007; Woodruff et al., 2007; Herrmann et al., 2010.

primary determinant for setting single photon response amplitude. GCAP2 assumes greater importance with stronger flashes (Mendez et al., 2001; Makino et al., 2008). Bright flashes saturate the rod response; the brighter the flash, the longer it takes for the response to emerge from saturation. GCAP2 acts to reduce the saturation time to an extent roughly equivalent to lowering flash strength 3.8-fold (Makino et al., 2008). Loading WT rods with a Ca2<sup>+</sup> buffer increases the size of the photon response and alters its kinetics but has no effect on GCAPs−/<sup>−</sup> rods, indicating that as with cones, no other major Ca2<sup>+</sup> feedback besides GC/GCAP comes into play during the flash response (Burns et al., 2002).

It might seem that enlarged single photon responses should improve the ability of GCAPs−/<sup>−</sup> and GCAP1−/<sup>−</sup> rods to count photons. But when GCAPs are not present, there is a 40-fold greater variance of the dark noise (Burns et al., 2002), which would instead diminish the accuracy of detecting photons. Phototransduction noise stems from two sources: discrete noise due to the thermal activations of rhodopsin (Baylor et al., 1980) and continuous noise due to spontaneous PDE activation (Rieke and Baylor, 1996). GCAPs counteract the cGMP depletion caused by both components (Burns et al., 2002). GCAP1−/<sup>−</sup> rods are also noisy, albeit somewhat less so, whereas GCAP2−/<sup>−</sup> rods appear to be normal, indicating that GCAP1 is more important for dampening the noise although GCAP2 likely comes into play when GCAP1 is not present (**Figure 3A**).

In steady light, given that stimulation of GC activity by GCAPs increases with light intensity as Ca2<sup>+</sup> is driven down to lower and lower levels, one might expect GCAPs to extend the operating range of rods by keeping them out of saturation, just as GCAP1 does in cones (**Figure 2B**). But that does not happen; the relation between fractional current and light intensity does not fall more gradually in WT rods compared to GCAPs−/<sup>−</sup> rods (**Figure 2D**). One explanation might be that something happens at higher light intensities, such as the summation of long lasting, aberrant photon responses (see below), which undermines the

efforts of GCAPs in their attempt to help rods evade saturation. Interestingly, knockout of GCAP1−/<sup>−</sup> does extend the operating range of rods, perhaps compensatory overexpression of GCAP2 boosts the maximal GC activity and the stimulation of GC activity is concentrated at higher intensities (Makino et al., 2012b).

Mammalian rods possess an odd characteristic: about once per several hundred events, rods generate an aberrant single photon response that climbs to an amplitude nearly twofold larger than normal and then remains at that amplitude for an average of 3–6 s, before recovering. In individual trials, the duration is unpredictable and some aberrant responses last for tens of seconds (Baylor et al., 1984; Chen et al., 1995; Kraft and Schnapf, 1998). Apparently, rhodopsin excitations are not always shut off properly by phosphorylation and arrestin binding (Chen et al., 1995, 1999). The aberrant response is enhanced in the absence of GCAP1, to a greater extent than the normal single photon response (Makino et al., 2012b). Even though aberrant responses are relatively rare, they have a significant physiological impact, because they produce a "tail" that delays the recovery after bright flashes. Tails are especially prominent when GCAP1 is missing (**Figure 3B**).

The basis for the different roles of GCAP1 and GCAP2 lies in their Ca2<sup>+</sup> sensitivities. GCAP1 has a lower affinity for Ca2<sup>+</sup> than GCAP2 (**Figure 4A**). The K1/<sup>2</sup> for Ca2<sup>+</sup> of GCAP1 in the mouse rod is 130–140 nM, while that of GCAP2 is 50–60 nM (Makino et al., 2008; Peshenko et al., 2011). Levels of free Ca2<sup>+</sup> inside a mouse rod range from 250 nM in darkness to 23 nM in saturating light (**Table 3**). So during the initial fall in Ca2<sup>+</sup> incurred during the photon response, many more GCAP1 molecules release their Ca2<sup>+</sup> than GCAP2. The majority of GCAP2 molecules only release their bound Ca2<sup>+</sup> when intracellular Ca2<sup>+</sup> plummets, during the saturating response to very bright light. A model illustrating the actions of GCAP1 and GCAP2 is shown in **Figure 4B**. A similar "relay" model was proposed earlier (Koch and Dell'Orco, 2013) but we here prefer the term "recruitment" model to avoid

conveying inadvertently the impression that GCAP2 stimulation substitutes for GCAP1 stimulation at low Ca2+. "Recruitment" emphasizes that GCAP2 adds to stimulation of GC by GCAP1.

The timing of Ca2<sup>+</sup> feedback onto GC activity during the photon response is shown in **Figure 5A**. GC activity ascends at the same rate as the electrical response with a ∼40 ms delay, then declines sharply and falls briefly to a level less than the dark value. GCAP1's lower affinity to Ca2<sup>+</sup> makes it the logical "first responder" to the initial decrease in Ca2<sup>+</sup> (**Figure 4**). Although it could have been the case that GCAP2 responds later because it has a slower release of Ca2<sup>+</sup> than GCAP1 after a sudden drop in intracellular Ca2+, it will be argued below that both GCAPs are rapid Ca2<sup>+</sup> sensors. **Figure 5B** shows the rising phases of the photon responses of mouse rods lacking GCAP1, GCAP2, or both. The GCAPs−/<sup>−</sup> response is the first to diverge at <sup>∼</sup>80 ms after the flash, while the GCAP2−/<sup>−</sup> response is the last to diverge. Based on this pattern, it may be inferred that GCAP1 senses the fall in Ca2<sup>+</sup> and begins to stimulate GC by 80 ms post-flash. GCAP1−/<sup>−</sup> rods are highly variable in their response waveforms apparently because they upregulate GCAP2 expression and the degree of compensation differs from rod to rod. Rods with the least compensation show response divergence from the WT response nearly as early as the GCAPs−/<sup>−</sup> response. Accordingly, GCAP2 must respond rapidly to the initial fall in Ca2+. At normal expression levels, the paltry number of GCAP2 molecules responding to the initial fall in Ca2<sup>+</sup> is unable to make its presence felt. But the mutant rods show that given enough GCAP2, its feedback onto GC does restrain the early rising phase of the response.

As one might expect, knockout of both GCs in rods eliminates the dark current, sabotaging their capacity for phototransduction (Baehr et al., 2007). Deleting GC2 alone seems to have little effect on the dark current or on flash response kinetics and sensitivity in recordings from individual mouse rods (Baehr et al., 2007). In contrast, GC1 deletion compromises the expression of both GCAPs revealing differences between the duties of the two GCs not directly related to phototransduction in the outer segment. There is an increase in flash sensitivity and a delayed time to peak of the photon response (Yang et al., 1999; Baehr et al., 2007), as predicted from the reduced levels of GCAPs (Mendez et al., 2001). Paradoxically, elimination of GC1 also accelerated the rod response recovery (Yang et al., 1999).

In summary, the functional significance to phototransduction of expressing two types of GCs in rods is not yet clear. For GCAPs, there is a division of labor. GCAP1 helps to achieve the proper balance of keeping noise to a minimum while restricting photon response size to optimize the range of flash strengths and light intensities over which the rod operates. Contrary to the situation in cones, the shutoffs of photoexcited visual pigment and activated transducin are slow in rods to allow the photon response to grow. The slow shutoffs pose a problem with bright flashes because the time that the rod stays in saturation gets prolonged. GCAP2 addresses that problem by boosting GC activity to bring the rod out of saturation sooner.

## **TUNING Ca2+ FEEDBACK WITH ADDITIONAL GCs AND GCAPs BETWEEN DIFFERENT ANIMAL SPECIES**

While it is generally true that the photon response of rods is larger and slower than that of cones, the response can vary across rods and across cones of different species. The preceding sections described how Ca2<sup>+</sup> feedback onto cGMP synthesis reduces the size of the single photon response and quickens its kinetics in mouse rods and cones. The extent to which that occurs is dependent upon the selective expressions of GC(s) and GCAP(s). Therefore, variation that exists in the sensitivity and flash response kinetics of rods and cones from

maxima and follow the Hill equation: <sup>A</sup> <sup>=</sup> (Amax–Amin)/(1 <sup>+</sup> ([Ca2+]/ K1/2) nH)<sup>+</sup> Amin with nH values of 1.8, 1.6, and 2.1 for WT, GCAP1−/<sup>−</sup> and GCAP2−/−, respectively. The physiological range of internal free Ca2<sup>+</sup> is 23 nM in the light to 250 nM in the dark, as demarcated by the gradient background. Modified from Makino et al. (2008, 2012b). **(B)** Recruitment model. The physiological free Ca2<sup>+</sup> concentration range shown by the black progression at the bottom of the figure. In darkness (right), nearly all of the GCAP1 and 2 "players" bind Ca2<sup>+</sup> and suppress GC activity (only a single representative player of each type is shown). Soon after illumination, both GCAPs respond to the light-induced Ca2<sup>+</sup>

on GCAP, prompting it to raise GC activity (Peshenko and Dizhoor, 2004). As intracellular Ca2<sup>+</sup> falls further, the rest of the GCAP2 players exchange Mg2<sup>+</sup> for bound Ca2<sup>+</sup> and they join the GCAP1 players in lifting GC activity. Very few GCAP2s exchange Ca2<sup>+</sup> for Mg2<sup>+</sup> during the single photon response, but with bright light, all GCAP1 and GCAP2 players get involved (far left). During the recovery of the response to light, cyclic nucleotide gated (CNG) channels reopen to allow Ca2<sup>+</sup> back in. GCAP2 players regain their Ca2<sup>+</sup> first and begin to restrain GC. Gradually, as intracellular Ca2<sup>+</sup> levels return to baseline, GCAP1 players also regain Ca2<sup>+</sup> and turn down GC activity.



Intracellular free Ca2<sup>+</sup> declines after sudden closure of CNG channels by bright light as the sum of two exponential functions. The two time constants listed are from measurements of fluorescent Ca2<sup>+</sup> indicators. Electrophysiological measurements of Ca2<sup>+</sup> exchange are not included because they often do not resolve the second, slower component. T = 37◦C for mouse rod, and T = 16–23◦C for lizard and salamander cells. From Woodruff et al., 2002; Gray-Keller and Detwiler, 1994; Lagnado et al., 1992, Sampath et al., 1998, 1999.

other vertebrates is likely to arise at least in part from differences in their cGMP synthetic machinery. This section will present a few examples of the variation between species and lay out three mechanisms by which cGMP production could be altered to tune the photon response: by augmenting the amounts of GC and GCAP expressed, by a change in intracellular Ca2<sup>+</sup> dynamics or by a switch in the types of GCs or GCAPs expressed.

In comparison to mouse rods, primate rods have a single photon response that peaks later whereas the response of chipmunk

**FIGURE 5 |Timeline for GCAP regulated GC activity during the photon response in mouse rods. (A)** The calculated time course of GC activity α normalized to the basal value in darkness (broken trace) overlaid on the single photon response (solid trace). The two curves rise with similar slopes to reach their peaks 40 ms apart. Thereafter, GC activity declines sharply. Reproduced from Burns et al. (2002) with permission from Elsevier. **(B)** Divergent rising phases of the photon responses of WT rods and rods lacking either or both

rods peaks and recovers earlier (**Table 2**). In general, the single photon response of cones is far smaller and peaks much sooner than the response of rods in the same animal. Across species, ground squirrel cones generate smaller, faster photon responses than mouse cones (**Figure 6**; **Table 2**). At present, primate cones are something of an enigma. Their single photon response kinetics are faster than those of mouse yet the single photon response is larger and relative sensitivity is greater. In initial studies on isolated primate retina, cone flash responses exhibited a prominent and reproducible undershoot to the recovery (Baylor et al., 1987; Schnapf et al., 1987; Schnapf et al., 1990). Under the same conditions, cones of other mammals did not generally have undershoots, with the possible exception of chipmunk cones (Zhang et al., 2003). However, in a later study on primate cones, undershoots were absent or only seen occasionally (**Figure 6**) and kinetics inferred for cones in the intact human eye lack the undershoot (van Hateren and Lamb, 2006). Although flash response kinetics are invariant across cone spectral type in mammal (e.g., Baylor et al., 1987 and Cao et al., 2014), there are distinct differences across cone type in fish and amphibians where faster time to peak corresponds with reduced sensitivity (Perry and McNaughton, 1991; Miller and Korenbrot, 1993; Isayama et al., 2006, 2014). After a partial bleach of the visual pigment content, flash responses of salamander L cones develop an undershoot but flash responses of S cones do not (cf. Figures 4–6 of Isayama et al., 2006).

It has been proposed that an enhanced expression of GC in cones accounts, in part, for their faster flash response kinetics and their ability to signal over an enormous range of light intensities (Takemoto et al.,2009). Expression of GC is 17 times higher in carp cones than in rods (**Table 1**) and the basal rate of cGMP synthesis in cones is an order of magnitude higher. The differential appears to be lower in salamander to account for a basal rate that is only ∼3-fold higher in their cones compared to their rods (Cornwall and Fain, 1994; Cornwall et al., 1995). But despite the higher basal activity in cones, the fold change in stimulation of GC activity in the light is greater in rods than in cones for both carp and

GCAPs. The GCAPs−/− response separates from the WT response at 70–80 ms after the flash. Separation of the larger response of those GCAP1−/− rods presumed to have little GCAP2 overexpression (dashed blue trace) is next, at 90–100 ms, followed by the smaller response of those GCAP1−/− rods presumed to have the most GCAP2 overexpression (solid blue trace) at 100–110 ms. The response of GCAP2−/− rods overlays the WT until well after the peak. Includes results from Makino et al. (2008, 2012b).

salamander. GC will function in the absence of GCAP, so a lower fold change in GC synthesis at low Ca2<sup>+</sup> could be achieved by reducing the ratio of GCAP to GC. That turns out not to be the case in carp, indicating that there are quantitative differences in the way that the cGMP synthetic machineries of rods and cones respond to low Ca2<sup>+</sup> and/or differences in their Ca2<sup>+</sup> dynamics (Takemoto et al., 2009).

Dissimilarities in the operation of the same cGMP synthetic machinery may exist across photoreceptors because of how they handle Ca2+. Ca2<sup>+</sup> dynamics in the outer segment depend upon the rates of Ca2<sup>+</sup> entry through the CNG channel and removal by a Na+/K+, Ca2<sup>+</sup> exchanger, as well as Ca2<sup>+</sup> buffering and outer segment surface to volume ratio. In cones, the fraction of circulating current into the outer segment carried by Ca2<sup>+</sup> is roughly twice that in rods, due to a greater permeability of the cone CNG channel for Ca2<sup>+</sup> (Nakatani and Yau, 1988; Perry and McNaughton, 1991; Ohyama et al., 2000). It is not clear whether the cone exchanger operates at a faster rate than the rod exchanger, but surface to volume is greater in cone outer segments and their time constant for Ca2<sup>+</sup> extrusion is faster (Sampath et al., 1999; Paillart et al., 2007). As a result, [Ca2+] changes more rapidly and the fold change in Ca2<sup>+</sup> levels is bigger in cones than in rods (**Table 3**). In salamander, a faster rate of Ca2<sup>+</sup> extrusion in L cones than in S cones may contribute to faster response kinetics of L cones (Perry and McNaughton, 1991). In both salamander rods and cones, the dark level of Ca2<sup>+</sup> is higher than in mouse rods, and in salamander cone, the light induced fall in Ca2<sup>+</sup> is to 5 nM, a lower level than that in mouse rod, 23 nM (Sampath et al., 1999; Woodruff et al., 2002). For feedback on cGMP synthesis to operate over the full range of Ca2<sup>+</sup> concentrations, the choice of GCAP(s) expressed will need to be based in part on Ca2<sup>+</sup> affinity.

The occurrence of isoforms of GC and GCAP allows for adjustments to be made in their biochemical properties: K1/<sup>2</sup> for Ca2+, cooperativity for Ca2<sup>+</sup> binding, maximal and basal rates of cGMP synthetic activity. Further flexibility stems from having maximal stimulation of GC depend on the pairing of a GC with specific GCAP isoforms. The unstimulated rate of cGMP synthesis by mouse GC1 is nearly 10-fold less than that of GC2. But stimulation of GC1 activity by GCAP1 and GCAP2 is 28-fold and 13-fold, respectively, whereas for GC2 activity, it is only sixfold and fivefold, respectively (Peshenko et al., 2011). In some cases, the effects of GCAPs on GC are species dependent. Zebrafish GCAP5 will stimulate carp rod GCs severalfold at low Ca2<sup>+</sup> concentration but has little effect on carp cone GC (Takemoto et al., 2009). The 6 zebrafish GCAPs differ in the degree to which they will stimulate bovine rod GC activity, ranging from 1.5- to 13-fold (**Figure 7**), but stimulation of zebrafish GC3 by GCAPs 3–5 and 7 is more uniform at ∼2-fold (Fries et al., 2013). Carp GCAP1 stimulates carp GC preparations consisting mostly of GC1 to a greater extent than GCAP2 (Takemoto et al., 2009), which is just the opposite of the effect of these zebrafish GCAPs on bovine GC (Scholten and Koch, 2011). These comparisons do not carry physiological significance, nonetheless, it is clear that switching GC and GCAP types is important for setting the basal rate of cGMP synthesis at high Ca2+, the maximal rate at low Ca2+, and the fold activation. In contrast to mouse cones that express GC1 alone, mouse rods add a second GC with a ratio of GC1/GC2 of (4:1) (Peshenko et al., 2011), similar to the (3:1) ratio in bovine rods (Skiba et al., 2013). It should be noted that these recent determinations indicate that the fraction of RetGC2 in mammalian rods is much higher than previously thought (Hwang et al., 2003; Helten et al., 2007).

GCAPs also vary in their K1/<sup>2</sup> for Ca2<sup>+</sup> (Hwang et al., 2003; Peshenko et al., 2011), a property that would appear to be important for optimizing feedback onto cGMP synthesis to the

**GCAPs.** Reconstituted zebrafish GCAPs were assayed in vitro against bovine rod outer segment membranes that were stripped of native GCAPs but contained a mixture of GC1 and GC2, with GC1 predominating. Activity is plotted relative to the maximum rate for each isoform. K1/<sup>2</sup> values are 30, 35, 25, 520, 440 and 180 nM, and the fold activation values are 3, 10, 6, 13, 1.5, and 2 for zGCAP1-5 and 7, respectively. From Scholten and Koch (2011).

light induced fall in Ca2+. For example, with K1/<sup>2</sup> values of 130–140 nM and 50–60 nM (Makino et al., 2008; Peshenko et al., 2011), mouse GCAPs 1 and 2 provide little or no incremental activation of GC below ∼10 nM (**Figure 4A**), but intracellular [Ca2+] in salamander cones falls to 5 nM upon exposure to bright light. So another GCAP with lower K1/<sup>2</sup> for Ca2<sup>+</sup> is needed if cGMP synthetic rate is to increase over the entire Ca2<sup>+</sup> range. K1/<sup>2</sup> values for zebrafish GCAPs 1– 3 cluster at about 30 nM (**Figure 7**). It remains to be seen whether lower values are possible. Certainly GCAPs, e.g., zGCAPs 4,5, can have K1/<sup>2</sup> values higher than those of mouse GCAPs (**Figure 7**), but while such GCAPs may be useful for photoreceptors of lizards and salamanders (**Table 3**), they cause degenerative retinal disease in mammals (reviewed in Behnen et al., 2010). Ca2+-dependent activation of GC by GCAPs is cooperative (reviewed in Luo et al., 2008; Gross and Wensel, 2011; Korenbrot, 2012) but so far, there is no evidence for differences in their cooperativity (e.g., Peshenko and Dizhoor, 2004) or in the rapidity of their response to low Ca2<sup>+</sup> (see above). Given the differences in the composition of their cGMP machineries (**Table 1**), it may be predicted that the physiological properties of zebrafish cones will vary with spectral type.

The selectivity rules for forming GC complexes are not completely understood. The apparent affinity of mouse GC2 is slightly higher for GCAP2 than for GCAP1 *in vitro* (Peshenko et al., 2011), yet in living mouse rods, regulation of RetGC2 is heavily dominated by GCAP2 with very little, if any, contribution from GCAP1 (Olshevskaya et al., 2012). Apparently, the selectivity mechanism taking place *in vivo* is not governed strictly by binding affinity observed *in vitro*. Thus while the Ca2<sup>+</sup> sensitivity of the regulation is determined by the type of GCAP rather than isozyme of the cyclase (Peshenko et al., 2011), the type of GC expressed could shift the GCAP dominance and the Ca2<sup>+</sup> sensitivity.

It may be argued that Nature could have given mouse rods a single GCAP having a K1/<sup>2</sup> for Ca2<sup>+</sup> that matches the value obtained by their mixture of GCAPs 1 and 2. This path was not chosen, perhaps because the flexibility of a mixture of components allowed for adjustments to be made on a shorter evolutionary time scale to accommodate other changes in the phototransduction cascade, as rods evolved as photoreceptors (Lamb, 2013).

On a much shorter time scale, relevant to the lifetime of an individual, the complexity of a mixture of components leaves open many options for plasticity. The absence of GCAP5 transcripts in zebrafish until 15 dpf (Rätscho et al., 2009), even though at least some cones are present by 3 dpf suggests a changeover in the type of GCAP expressed with age. Since each GC has a rank order of preference for GCAPs with which it will form a complex (Takemoto et al., 2009; Peshenko et al., 2011; Scholten and Koch, 2011; Olshevskaya et al., 2012), relative expression levels of each GCAP may be a factor in determining which GC/GCAP complexes predominate. In zebrafish cones, GCAP3 expression exceeds GC3 expression by an order of magnitude, yet there is no change in optomotor behavior upon knockdown of GCAP3 (Fries et al., 2013). Presumably, GCAP3 is replaced with a GCAP having similar properties. However, it is possible that reduced expression of a different GCAP would result in greater incorporation of GCAP3 with a change in visual function. In mice, knockout of GCAP2 has no effect on GCAP1 expression, but knockout of GCAP1 drives an increase in GCAP2 expression with a greater fractional incorporation of GCAP2 into GC complexes that yields a higher maximal GC activity at low Ca2<sup>+</sup> (Makino et al., 2012b). It is tempting to speculate that GCAP1 expression may fall under transcriptional control by some conditions. Exposure to bright light reduced PDEα mRNA levels in mice by twofold (Krishnan et al., 2008) and PDEα and PDEβ protein levels in rat by threefold (Hajkova et al., 2010). In rat, there were no changes in GC1 (GC-E) expression; other cGMP synthetic machinery components were not reported. Then unless another GC activity went down, cGMP levels would have risen. GCAPs are sensitive to the [Mg2+], which allows them to maintain their cyclase activating conformation in the light and affects the [Ca2+] at which they de-activate GC (Peshenko and Dizhoor, 2004). However, evidence so far indicates that internal [Mg2+] does not change in rods or in cones upon exposure to light and remains in the vicinity of 1 mM (Chen et al., 2003). So while the potential exists for changes in the choice of GC or GCAP expressed and in their relative levels of expression during development or in response to environmental conditions, the issue of plasticity is at present, unresolved.

Evolution has favored the occurrence of multiple isoforms of GCAPs that vary in their Ca2<sup>+</sup> affinity and GCs with different activities and preferences for the type of GCAP bound. By expressing selected components and regulating their levels of expression, vertebrates "design" the cGMP synthetic machinery in their photoreceptors in order to adjust their physiological responses to light stimuli. The possibility exists for plasticity during development or on a shorter time scale to meet changing visual demands.

### **ACKNOWLEDGMENTS**

This work was supported by the Howe Laboratory Endowment of the Massachusetts Eye and Ear Infirmary, Research to Prevent Blindness, National Eye Institute EY014104, EY011522. Alexander M. Dizhoor is the Martin and Florence Hafter Chair Professor of Pharmacology. The authors bear sole responsibility for the contents herein. The views expressed do not necessarily convey those of the National Eye Institute.

#### **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: 05 February 2014; accepted: 01 May 2014; published online: 02 June 2014. Citation: Wen X-H, Dizhoor AM and Makino CL (2014) Membrane guanylyl cyclase complexes shape the photoresponses of retinal rods and cones. Front. Mol. Neurosci. 7:45. doi: 10.3389/fnmol.2014.00045*

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

*Copyright © 2014 Wen, Dizhoor and Makino. 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.*

## PDE9A is expressed in the inner retina and contributes to the normal shape of the photopic ERG waveform

#### *Anuradha Dhingra1†, Shanti R. Tummala1†, Arkady Lyubarsky2 and Noga Vardi <sup>1</sup> \**

*<sup>1</sup> Retina Lab, Department of Neuroscience, University of Pennsylvania, Philadelphia, PA, USA*

*<sup>2</sup> Department of Ophthalmology, University of Pennsylvania, Philadelphia, PA, USA*

#### *Edited by:*

*Rameshwar K. Sharma, Salus University, USA*

#### *Reviewed by:*

*Laura J. Frishman, University of Houston, USA Edward N. Pugh, University of California, Davis, USA*

#### *\*Correspondence:*

*Noga Vardi, Retina Lab, Department of Neuroscience, University of Pennsylvania, 123 Anat-Chem Bldg., Philadelphia, PA 19104, USA e-mail: noga@mail.med.upenn.edu*

*†These authors have contributed equally to this work.*

The ubiquitous second messenger cGMP is synthesized by guanylyl cyclase and hydrolyzed by phosphodiesterase (PDE). cGMP mediates numerous signaling pathways in multiple tissues. In the retina, cGMP regulates signaling in nearly every cell class including photoreceptors, bipolar cells, amacrine cells, and ganglion cells. In order to understand the specific role of cGMP and its regulating enzymes in different cell types, it is first necessary to localize these components and dissect their influence on the circuits. Here we tested the contribution of PDE9A to retinal processing by recording the electroretinograms (ERG) of *PDE9A*−/<sup>−</sup> (KO) mice and by localizing the enzyme. We found that while the scotopic ERG of KO was the same as that of wild type (WT) in both amplitude and kinetics, the photopic ERG was greatly affected. The greatest effect was on the recovery of the b-wave; the falling phase and the b-wave duration were significantly longer in the KO mice for all photopic stimuli (UV, green, or saturating white flashes). The rising phase was slower in KO than in WT for UV and green stimuli. For certain stimuli, amplitudes of both the a- and b-waves were smaller than in WT. Using *Lac-Z* expression in KO retinas as a reporter for PDE9A expression pattern, we found that PDE9A is localized to GABA-positive and GABA-negative amacrine cells, and likely also to certain types of ganglion cells. Our results indicate that PDE9A, by controlling the level of cGMP, modulates inhibitory processes within the cone pathway. We speculate that these circuits involve NO/cGMP signaling pathways.

**Keywords: ERG, serial inhibition, amacrine cells, cyclic GMP, cone pathways, ciliary body**

## **INTRODUCTION**

The ubiquitous second messenger cGMP, which is synthesized by guanylyl cyclases (GC) and hydrolyzed by phospodiesterases (PDE), controls a variety of processes from smooth muscle cell contraction in the vasculature to signaling in the central nervous system (Beavo, 1995; Polson and Strada, 1996; Juilfs et al., 1999). cGMP typically accomplishes these functions either by gating cyclic nucleotide-gated (CNG) channels or by activating cGMP-dependent kinases and phosphatases. In the retina, cGMP regulates nearly every step of signal processing. At the first step of retinal processing, rod and cone photoreceptors use cGMP to gate CNG channels. In these cells, light activates PDE6, which hydrolyzes cGMP; this closes the CNG channels and the cells hyperpolarize (Lamb and Pugh, 1992; Hsu and Molday, 1993; Yau, 1994). At the next step of retinal processing, ON bipolar cells (rod bipolar that mediate night vision and ON cone bipolar cells that signal light increments in daylight vision) appear to use cGMP, but its function is still unclear. cGMP was initially thought to gate a CNG channel similar to photoreceptors (Nawy and Jahr, 1990; Shiells and Falk, 1990, 1992; de la Villa et al., 1995). It is now known that ON bipolar cells use the nonselective cation channel TRPM1, which is not gated by cGMP (Morgans et al., 2009; Koike et al., 2010). Instead, some evidence suggests that cGMP potentiates the light response at low light intensities (Nawy, 1999; Shiells and Falk, 2002; Snellman and Nawy, 2004). At the next stage of visual processing, some third order neurons (including certain types of amacrine and ganglion cells) modulate cGMP, often in response to NO (Ahmad et al., 1994; Gotzes et al., 1998; Chun et al., 1999; Blute et al., 2003). The NO/cGMP pathway is also critical for modulating coupling conductance between the AII amacrine cells and ON cone bipolar cells (Mills and Massey, 1995). The exact role and mechanism of cGMP regulation in second and third order neurons is not clear. In order to better understand the role of cGMP in different steps of processing, and especially in ON bipolar cells, it is important to localize PDE enzymes in retina. To this end, using our ON bipolar cell cDNA library, we identified a PDE isoform *PDE9A* and determined that these cells express the transcript for this isoform (Dhingra et al., 2008). Moreover analysis of the retinal transcriptome database (Siegert et al., 2009) showed that several amacrine and ganglion cell types also express *PDE9A* transcripts. Because PDE9A is expressed predominately in neurons (van Staveren and Markerink-van Ittersum, 2005), and hydrolyses cGMP with a very low Km (Fisher et al., 1998), it likely contributes to retinal function. Here, using *PDE9A*−/<sup>−</sup> (KO) mouse, we report that PDE9A contributes to the normal kinetics of the photopic electroretinogram (ERG). PDE9A is localized to amacrine cells, ganglion cells, and to the ciliary body, but not ON bipolar cells. We therefore conclude that inner retinal expression of PDE9A in amacrine cells contributes to response kinetics of the ERG b-wave.

### **MATERIALS AND METHODS**

#### **GENERATION OF** *PDE9A* **KNOCKOUT MICE**

*PDE9A*−/<sup>−</sup> (KO) mice were obtained from Pfizer (Menniti et al., 2008) and were initially generated by Deltagen (San Mateo, CA). Briefly, a targeting construct with neomycin resistance (neo) and bacterial *LacZ* (coding for β-galactosidase or β-gal) genes was designed to disrupt the *PDE9A* gene (Accession # AF031147) in ES cells by homologous recombination. The resultant allele had a replacement of part of exon 11/12 of *PDE9A* with neo-*LacZ* cassette that shifted the downstream sequence out of the open reading frame. The ES cells were injected into blastocysts and the resulting chimeras were tested for germ line transmission. The following primers were used for genotyping:

GS (E) (CACAGATGATGTACAGTATGGTCTGG);

GS (T,E) (TGCAGTCATCAGGACCAAGATGTCC); and Neo (T) (GACGAGTTCTTCTGAGGGGATCGATC).

Genotyping was performed on tail DNA using a multiplex PCR reaction: GS (T, E) + Neo (T) for targeted allele (expected size 599 bp); GS (E) + GS (T, E) for endogenous allele (expected size 256 bp). KO mice were bred to homogenicity by multiple back-crossings (>6) with C57BL6/J mates. Mouse colonies were maintained by breeding KO mice; no specific behavioral phenotypes were observed in these mice. Age matched C57BL6/J (WT) were purchased from Jackson Labs (Bar Harbor, Maine). All experiments were done in compliance with federal regulations and the protocol was reviewed and approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Both male and female mice were used for all experiments.

#### **ELECTRORETINOGRAM**

The ERG recording setup and methods have been described elaborately (Lyubarsky et al., 1999, 2000; Ng et al., 2010). In short, mice were dark-adapted, deeply anesthetized intraperitoneally under dim light with ketamine/xylazine/urethane (20, 8, and 800μg/gm bodyweight, respectively) and were placed on a platform maintained at 37–38◦C. Pupils were dilated with 1% tropicamide saline solution (Mydriacyl, Alconox). A platinum electrode was inserted into the mouth to serve as a reference and ground electrode, and another platinum electrode was placed on the cornea. Mice were placed inside a light-proof Ganzfield Faraday cage. ERG recordings from KO and age matched WT mice were performed on the same day under identical settings and conditions. Light stimuli were either 4 ms flashes produced by a light-emitting diode (LED) stimulator or <1 ms flashes produced by a Xenon tube delivered in the Ganzfeld (Espion Electrophysiology System; Diagnosys). For scotopic ERG, intensities of light flashes were converted to the estimated number of photoisomerizations (*R*∗) per rod as described previously (Lyubarsky et al., 2004). For photopic ERG, light intensity was converted to number of photons per μm<sup>2</sup> at the cornea as described previously (Lyubarsky et al., 1999, 2000). For photopic stimuli, a rod-suppressing step of light (30 scot cd m−2) was given every 5.5 s and a light flash was superimposed on it 2 s after the onset of the step. ERGs were recorded from both eyes using differential amplifiers with a bandwidth of 0.1 Hz–1 kHz, and a sampling interval of 1 ms. Depending on the signalto-noise ratio, each record was an average of 3–25 individual trials.

The various parameters characterizing the ERG waveform were determined using a user-defined MATLAB® (Mathworks, MA) program. The amplitude of the b-wave was quantified by subtracting the peak of the a-wave from the peak of the b-wave. The functional characteristics of phototransduction in photoreceptors in WT and KO animals were determined from the ERG a-wave by fitting the rising phase of the a-wave with the transduction cascade activation model (Breton et al., 1994; Lyubarsky and Pugh, 1996) as follows:

$$F(t) = \exp[-0.5\Phi A \left(t - t\_{\rm eff}\right)^2],\tag{1}$$

where *F(t)* is the a-wave amplitude at time *t* after a brief flash normalized to the saturated amplitude; is the number of photoisomerizations per rod; *teff* a brief delay; *A* is the amplification constant whose value is estimated by the best fit. ERG data presented in this paper were performed on 12 WT and 14 KO mice, but occasional records were too noisy and were not considered. Values of the measured response parameters for the left and right eyes were averaged and treated as a single data point. The number of animals or data points acquired for each stimulus condition is specified in the Results section. All data values in the figures are presented as mean ± s.e.m. Unless stated otherwise, statistical comparison of a parameter between groups of WT and KO was performed using two-tailed Student's *t*-test. A *p*-value of less than 0.05 (*p* < 0.05) was considered significant and *p* < 0.01 was considered highly significant.

#### **β-GALACTOSIDASE ASSAY**

Mice were deeply anesthetized with an intraperitoneal injection of ketamine and xylazine (100 and 10μg/gm bodyweight). Eyes were enucleated and incised just below the limbus. Eyeballs were fixed in 4% paraformaldehyde for 10 min, and rinsed in 0.1 M phosphate buffer. For cryosections, eyeballs were cryoprotected by soaking overnight at 4◦C in 0.1 M phosphate buffer containing 30% sucrose and embedded in a mixture of two parts 20% sucrose in phosphate buffer and one part optimal cutting temperature compound (Tissue Tek, Electron Microscopy Sciences, Hatfield, PA, USA). Radial sections (15– 20μm) were cut on a cryostat (Leica) at −20◦C and collected on superfrost plus slides (Fisher Scientific, Pittsburgh, PA, USA). β-galactosidase expression was detected enzymatically using Xgal (5-bromo-3-indoyl-b-D-galactopyranoside stock solution) as substrate. Briefly, whole mount or retinal sections were preincubated in a detergent buffer (0.02% Nonidet P-40, 0.01% Sodium Deoxycholate, and 2 mM MgCl2 in 0.1 M phosphate buffer, pH 7.3) for about 10 mins at room temperature (RT), incubated for 2–4 days in staining buffer (0.02% Nonidet P-40, 0.01% Sodium Deoxycholate, 5 mM Potassium Ferricyanide, 5 mM Potassium Ferrocyanide, and 2 mM MgCl2 in 0.1 M phosphate buffer, pH 7.3 containing 1–1.5 mg/ml X-gal) and finally rinsed twice in detergent buffer (5 min each).

#### **IMMUNOFLUORESCENCE**

Sections were permeabilized and blocked with 10% normal goat serum, 5% sucrose and 0.5% Triton X-100 in phosphate buffer for 1 h at RT. Sections were incubated in primary antibodies (diluted in blocking solution) at 4◦C overnight or occasionally for 3 days. Primary antibodies were washed 3 times in 5% sucrose in phosphate buffer at RT. Sections were then incubated in secondary antibodies at RT for 3 h, washed, mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA) and coverslipped. To identify X-gal-stained cells in the inner retina, some retinal sections were first stained for GABA and then processed for X-gal staining, and others were first processed for X-gal staining, and then immunostained for GABA. For double staining on the whole mount, we first stained for GABA overnight and then with x-gal for at least 3 days; after imaging, the retinas were also stained with DAPI. To avoid collapse of retinal layers, retinas were mounted between number 0 cover slips, which served to support the coverslip. The blue staining corresponding to the sites of β-galactosidase activity and the fluorescent stainings were examined with a conventional microscope (Nikon Eclipse TE-300 microscope; Nikon Inc., Melville, NY, USA).

#### *Antibodies*

Rabbit anti-PDE9A (1:100–500, GTX14625 from Genetex, Irvine, CA); rabbit anti-PDE9A (1:100, PD9A-101AP from FabGennix, Frisco, TX); guinea-pig anti-GABA (1:1000, AB175, Chemicon International, Temecula, CA); mouse anti-β-gal antibodies (a monoclonal; 1:200, Promega, Madison, WI); and chicken anti-βgal (1:500, Aves labs Inc., Tigard, Oregon).

## **RESULTS**

## **ABSENCE OF PDE9A DOES NOT AFFECT THE SCOTOPIC ERG, BUT DECREASES THE AMPLITUDE OF THE ROD/CONE-GENERATED ERG a-WAVE**

To determine if PDE9A contributes to visual processing, we recorded ERGs from WT and *PDE9A* KO mice. We first presented a series of scotopic flash intensities under dark-adapted conditions and measured the amplitude of the b-wave. There was no significant difference in the ERG b-wave amplitude between KO and WT (**Figures 1A–C**). To assess the response kinetics, we measured the time it took the b-wave to rise to 66% of its peak amplitude, the time it took to fall to 66% of its peak amplitude, and the 66% maximum width (as in **Figure 1D**). We also analyzed the half width; however, because responses of dark-adapted mice at high intensities often do not return to baseline in the sampled duration of 500 ms, this measure is less reliable and is not reported here. Under scotopic conditions, the kinetic parameters of KO and WT ERGs were similar (**Figure 1E**).

Next, to determine the possible contribution of PDE9A to the a-wave, we presented strong saturating flashes that stimulate both rods and cones (**Figure 2A**). In response to the strongest flash (*<sup>I</sup>* <sup>=</sup> <sup>9</sup>.<sup>3</sup> <sup>×</sup> <sup>10</sup><sup>5</sup> *<sup>R</sup>*∗/rod), both the a- and b-wave amplitudes were significantly lower in KO than in WT. The a-wave saturating amplitude in KO (227μV) was 36% lower than in WT (357μV; *p* = 0.002), and the b-wave saturating amplitude of KO was 30% lower (411μV in KO vs. 601 in WT; *p* = 0.009; **Figure 2B**). The

**conditions.** For all figures, WT is drawn in black and KO in red. Also for all figures, data are plotted as average ±s.e.m. When the s.e.m. cannot be seen, it is smaller than the symbol. **(A,B)** Representative responses to increasing light intensities (0.2– 600 *R*∗/rod) from age matched WT **(A)** and *PDE9A* KO **(B)** retinas. **(C)** Average b-wave amplitudes vs. flash intensity for WT (black squares; 16–22 records) and KO (red diamonds; 18–28 records) mice. **(D)** Illustration of the parameters used to characterize the kinetics of the b-wave. Stimulus onset is designated by an upward black arrow, 66% of the peak amplitude is indicated by the blue line and 50% by the green line. Time taken by the b-wave to rise from the stimulus onset to 66% of its peak amplitude (left dashed blue arrow) or to 50% of its amplitude (left dashed green arrow) were used for comparison of its kinetics. Time taken by the b-wave to fall to 66 or 50% of its peak amplitude is indicated by the right blue and green arrows. **(E)** Times taken by the b-wave to rise (triangles) and fall (squares) to 66% of its amplitude, and the duration between them (width, circles) at the measured intensities.

ratios of b-wave to a-wave amplitude for the three intensities were similar in KO and WT (ranged from 1.7 in WT and 1.8 in KO for the lower intensity to 3.6 and 3.7 for the higher intensity).

To determine if the reduced a-wave was due to compromised photoreceptor function, we computed the Lamb and Pugh amplification constant "*A.*" This constant provides a measure of the gain of the photoreceptor transduction cascade that is independent of light intensity for low-to-moderate intensities (Lyubarsky and Pugh, 1996). The amplification constant *A* was computed by fitting the a-wave component of two flash intensities (460 and 1860 *R*∗/rod) to Equation 1 (**Figure 2C**; the fit is shown with dashed lines). The amplification constant *A* in WT (3.5) was not statistically different from that in KO mice (4.3; **Figure 2D**; *p* = 0.11), indicating that deletion of *PDE9A* does not affect phototransduction. Next, we looked at the shape of the rod/cone mixed response and computed the 66% maximum width. As expected, in both WT and KO, the rise time of the b-wave decreased slightly and the fall time increased substantially with increasing intensity. The curves of the rise time as a function of intensity for KO and WT were very similar. The curves of the fall time and 66%

maximum duration for KO crossed those of WT: at lower intensities KOs were slightly faster, and at the highest intensity tested, the KOs took longer to recover (**Figure 2E**). However, neither of these differences was significantly different between WT and KO.

#### **ABSENCE OF PDE9A SLOWS DOWN THE PHOTOPIC RESPONSE**

To test the effect of deleting PDE9A on the photopic ERG wave, we suppressed rod responses using a light step of 30 scotopic cd m−<sup>2</sup> and superimposed (in sequence, with 5.5 s delay between trials) a saturating white flash of 1000 scot cd s m−2, UV flashes (<sup>λ</sup> = 365 nm; 4 <sup>×</sup> <sup>10</sup>3, 8 <sup>×</sup> <sup>10</sup>3, and 1.<sup>65</sup> <sup>×</sup> <sup>10</sup><sup>4</sup> photon/μm2), green flashes (<sup>λ</sup> <sup>=</sup> 513 nm; 8 <sup>×</sup> <sup>10</sup>3, 1.<sup>6</sup> <sup>×</sup> 104, and 3.<sup>2</sup> <sup>×</sup> 104 photon/μm2) and another saturating white flash of 1000 scot cd s m−2. In light-adapted rodents, the ERG a-wave is small and it reflects the activity of cones and the OFF pathway; the b-wave predominantly reflects the activity of ON cone bipolar cells and to a smaller extent that of the inner retina (Xu et al., 2003; Sharma et al., 2005; Shirato et al., 2008). For the saturated flashes, the average WT a-wave (combining the amplitude of both the first and the last flash) was 37.3μV and in KO mice, it was 23.5μV; significantly smaller than in WT (*p* = 0.004). The WT b-wave (170μV) was larger than that of KO (137μV) by 20% (albeit not significantly, *p* = 0.09; **Figures 3A,B**). Thus, absence of PDE9A greatly reduced the photopic a-wave amplitude, and only mildly effected the b-wave. Examination of the wave shape (**Figure 3C**) showed that while the times for the b-wave to rise to 66% of the peak in KO (40 ms) were similar to those in WT (41 ms) (*p* = 0.56), the time it took the b-wave to fall to 66% of its peak amplitude was highly significantly longer in KO (227 in KO vs. 156 ms in WT; *p* = 0.0001). Consequently, the 66% maximum width was also highly significantly longer in KO (201 in KO vs. 116 ms in WT; *p* = 0.0001).

Because photopic ERG responses were slower in KO than in WT, we wondered if light adaptation contributed to this difference. We therefore compared the first saturating response to the last saturating response in each experiment. Recall that the last saturating flash was presented after the UV and green flashes, so the eyes were exposed to the adapting light step for 12 min longer (**Figures 3D–F**). In WT, the kinetics of the b-wave in the first and last saturating flashes were similar, while in KO, the second presentation of the saturating flash resulted in a slower decay. The difference in the 66% maximum width was not significant (paired *t*-test; *p* = 0.1), but the 50% maximum width in KO was significantly longer in the second saturating flash (*p* = 0.007; **Figure 3F**). This suggests that PDE9A contributes to the kinetics of the cone-generated response, and that this contribution may be affected by the state of light adaptation.

To see if the effect of PDE9A's absence is specific to high intensities, or perhaps to different cone pathways, we examined the responses to low intensities of UV and green flashes. In response to UV flashes, both a- and b-waves amplitudes were similar in KO and WT (**Figures 4A,B**). However, in response to green flashes, both a- and b-wave amplitudes were significantly smaller in KO (**Figures 4D,E**). Like the saturating response, the time for the bwave to decay to 66% and the 66% maximum duration were significantly longer for both wavelengths at all intensities (*p* < 0.01; **Figures 4C,F**). In addition, at these wavelengths, the time to rise to 66% of the peak amplitude was also significantly longer (*p* < 0.001). Consequently, the time to peak in KO for all intensities at both wavelengths was also significantly longer. Taken together, our ERG results suggest that the main effect of deleting PDE9A is on the photopic ERG which reflects activity in the cone pathways.

### **PDE9A IS LOCALIZED TO CELLS IN THE GANGLION AND AMACRINE CELL LAYERS**

To understand the mechanism by which PDE9A affects retinal processing, it is essential to determine its cellular localization. We initially attempted to localize PDE9A using two commercial antibodies; both gave intense staining in several layers of the retina. However, when the antibodies were applied to KO retinas, the staining patterns were similar to those of WT, suggesting that the antibodies cross-react with other proteins (not shown). We similarly obtained non-specific bands using Western blots with these antibodies. To get a reliable localization pattern, we stained retinas for the reporter *Lac-Z* that replaced

*PDE9A* in KO. We first attempted to immunostain for β-gal, but there was no specific staining; instead we used the X-gal assay. In whole mount KO retinas, staining was present throughout the ganglion cell layer (GCL) (**Figures 5A,B**). No staining was found in WT, suggesting that this staining is specific and is present only in cells that normally express PDE9A. We determined the percentage of PDE9A-expressing cells within the GCL by co-staining the retina with DAPI. In a total GCL area of 150,000μm<sup>2</sup> (3 retinas combined) we found 1844 somas of which 28% were stained for X-gal. We also performed X-gal staining on vertical cryosections of the retina. Most staining was observed in cells of the GCL, but some staining was also observed in the somas of amacrine cells in the inner nuclear layer (**Figure 5C**).

Based on transcript presence and pharmacological blocking of certain PDE isoforms, it was suggested that retinal pigment epithelium (RPE) expresses PDE9A (Diederen et al., 2007). If so, absence of PDE9A from these cells may be able to explain the slow b-wave. We have therefore examined retinal cryosections that had RPE and choroid layers intact. There was no X-gal staining in the RPE; however we observed staining in the non-pigmented epithelium of the ciliary body (**Figure 5D**), in certain unidentified cells in the sclera, and in the lens (not shown). We further examined whole mounts of the pigment epithelium layer, but did not observe any staining.

To determine whether the cells in the GCL were amacrine or ganglion cells, we immunostained the X-gal stained retina for the amacrine cell markers GABA and glycine (**Figure 6**). Our antibody against glycine gave a high general background without specific staining, so the analysis relied on staining for GABA. We found that most X-gal-stained cells in the GCL were negative for GABA (e.g., cells 2, 6, 8 in **Figures 6A–C**), but many were positive (e.g., cells 1, 7). Similarly, in the inner nuclear layer, most Xgal-stained cells were negative for GABA but some were positive. To determine the percent of X-gal-stained cells that use GABA and the percent of GABA-immunoreactive cells that express βgal, we co-stained whole mount retinas with X-gal and for GABA (**Figure 6D**). We found that in the GCL, 20% of X-gal-stained cells also stained for GABA, and 20% of GABAergic cells also stained with X-gal (3 retinas, **Table 1**). In the INL, 27% of X-galstained cells stained for GABA, but only 7% of the GABAergic cells stained with X-gal. This experiment suggests that PDE9A is present in one or more types of GABAergic amacrine cells, but mostly in GABA-negative cells.

### **DISCUSSION**

We report here on two main findings. First, within the retina, PDE9A is restricted to the inner layers: it is widely distributed within the GCL and less widely within the amacrine tier of the inner nuclear layer. Second, PDE9A contributes to the amplitude and kinetics of the photopic ERG.

**FIGURE 4 | PDE9A accelerates the kinetics of the photopic b-wave. (A)** Average responses from WT and KO retinas to ultra-violet (UV) light of intensity 1.<sup>65</sup> <sup>×</sup> 104 photonsμm−<sup>2</sup> (17 records were averaged for WT and 24 for KO). **(B)** Amplitudes of the a- and b-waves from WT (*n* = 9–12 mice) and KO (*n* = 12–13) retinas at increasing intensities of UV light. **(C)** The time taken by the b-wave to rise and fall to 66% of its amplitude at different intensities of stimulation. The time taken by the b-wave to rise

and decay to 66% of its amplitude was significantly longer in KO at all intensities. **(D)** Average response of WT and KO retinas to green light of intensity 3.<sup>2</sup> <sup>×</sup> 104 photonsμm−<sup>2</sup> (*<sup>n</sup>* <sup>=</sup> 16 for WT and 19 for KO). **(E)** Amplitudes of the a- and b-waves from WT (*n* = 7–11) and KO (*n* = 2–11). Mice at increasing intensities of green light. **(F)** The times taken by the b-wave to rise and fall to 66% of its amplitude: both parameters are significantly longer in KO.

**FIGURE 5 | PDE9A is localized in the inner retina. (A)** Low magnification (4×) image of a whole mount *PDE9A* KO retina stained for X-gal (blue); on, optic nerve head. **(B)** Higher magnification (40×) image of the ganglion cell layer (GCL) in the same retina as in **(A)**: the stained cells occupy less than

50% of the area. **(C)** A cryosection of a KO retina stained for X-gal. X-gal staining is present in both the inner nuclear layer (INL) and GCL. IPL: inner plexiform layer. **(D)** Cryosection of KO retina showing X-gal staining in the ciliary body.

**FIGURE 6 | A population of X-gal positive cells is positive for GABA. (A,B)** Cryosections from KO retinas double stained with X-gal (**A1,B1;** blue, visualized with DIC microscopy) and for GABA (**A2,B2**, visualized with fluorescence microscopy). Dotted circles indicate the location of the somas of the cells that are positive for X-gal. The merged images **(A3,B3)** show that some cells positive for X-gal are also positive for GABA and others are negative for GABA. **(C)** Magnified images of the X-gal positive cells

(numbers on the side correspond to numbers on the low magnification images) that are also positive for GABA. Cells 1, 7 are located in the GCL and cell 4 is located in the INL. **(D)** Whole mount preparations stained with X-gal reagent and for GABA show that some cells in both GCL and INL stained for both (circles), while several X-gal stained cells were negative for GABA and numerous GABA-immunoreactive cells were not stained with X-gal.

#### **PDE9A IS LOCALIZED TO SEVERAL RETINAL CELL TYPES**

Our experiments identify at least three retinal cell types that express PDE9A: GABAergic amacrine cells, non-GABAergic amacrine cells (hence, glycinergic), and ganglion cells. The GABAergic amacrine cell type was identified by GABA-positive staining in a subset of X-gal-stained cells. We know that our immunostaining for GABA is genuine because we have previously shown that all cells that immunostain for GABA also express GAD65 or GAD67 (Vardi and Auerbach, 1995). In the inner retina, cGMP is predominantly synthesized by the soluble isoform of guanylate cyclase (sGC), which is activated by nitric oxide (NO). Immunostaining for the β1 subunit of sGC (the obligatory subunit of the heterodimer in neurons) in rat retina shows that sGC is expressed in a subset of amacrine cells, some of which are GABAergic (Ding and Weinberg, 2007). Given that the retinas of rats and mice are sufficiently similar, it is likely that these cells may

**Table 1 | Percentage of PDE9A-expressing cells in retina (averages of 3 retinas).**


be the GABAergic PDE9A-expressing amacrine cells. It remains to be determined whether the GABA-stained amacrine cells in the GCL are of the same type as those in the INL (but displaced), or if they represent a different amacrine cell type. The glycinergic type is inferred because nearly all amacrine cells in the INL stain either for GABA or for glycine (Marc et al., 1995; Vardi and Auerbach, 1995; Kalloniatis et al., 1996), thus it is highly likely that the GABA-negative X-gal stained amacrine cells are glycinergic. This inference agrees with a previous report that NOstimulated cGMP production is localized to glycinergic amacrine cells (Yu and Eldred, 2005). We think that one or more ganglion cell types express PDE9A because of the following facts: (1) about 40% of the GCL cells are ganglion cells (Jeon et al., 1998); (2) only few glycinergic amacrine cells are present in the GCL (Pourcho and Goebel, 1985, 1987; Marc et al., 1995); and (3) the PDE9Aexpressing cells in the GCL are more numerous than those in the INL, in fact about 30% of cells in GCL express PDE9A (this study). This agrees with presence of *PDE9A* transcript in ganglion cells (Siegert et al., 2009) and with reports that NO stimulates production of cGMP in certain ganglion cells (Ahmad et al., 1994; Blute et al., 2003).

#### **PDE9A ACCELERATES THE PHOTOPIC ERG b-WAVE BY ENHANCING SERIAL INHIBITORY RESPONSES**

The most striking phenotype of deleting PDE9A is the change in the kinetics of the photopic b-wave. Under light-adapted conditions, at the tested UV, green, and saturating white light intensities, the falling phase of the b-wave in KO was slower, and the wave was wider than in WT. For UV and green flashes, the rising phase of the b-wave was also slower. Slower ERG b-wave in light-adapted conditions had been reported following intravitreal injection of TTX in mouse (Miura et al., 2009) and rat (Bui and Fortune, 2004; Mojumder et al., 2008). Similar effects of TTX on ERG kinetics were also reported in rabbit after 5 min of dark adaptation (Dong and Hare, 2000). Even a greater effect on the photpic ERG kinetics was found following injection of PDA (cis-2, 3-piperidinedicarboxylic acid which blocks transmission to hyperpolarizing 2nd order and all 3rd order neurons), or CNQX (Sharma et al., 2005; Miura et al., 2009). Another common qualitative outcome to the *PDE9A* KO and to applications of TTX, PDA, or CNQX is the reduced b-wave amplitude. While some of the effect of TTX in the light adapted retina is due to sodium channels in certain ON cone bipolar cells (Mojumder et al., 2008), an additional effect must be due to contribution from the inner retina, suggesting that spiking amacrine cells enhance and speed up the ON cone bipolar cells' responses. The precise mechanisms by which amacrine cells affect the b-wave are not known. However, it is known that cone bipolar cells receive different types of inhibitory input from amacrine cells and that these amacrine cells themselves receive inhibitory input (serial inhibition). Furthermore, it has been shown by whole cell recordings and ERG recordings that these inhibitory circuits accelerate and often enhance bipolar cells' responses (e.g., Dong and Hare, 2000, 2002; Molnar and Werblin, 2007; Eggers and Lukasiewicz, 2011; Eggers et al., 2013).

How can deletion of PDE9A mimic this effect? It is well known that many amacrine cells use cGMP to modulate their activity. For example, light stimulation increases NO levels in the IPL (Eldred and Blute, 2005) and this increases the levels of cGMP. NOS1 (the enzyme that produces NO in neurons) is expressed in certain types of GABAergic amacrine cells, and NO stimulates the production of cGMP in neighboring sGC expressing cells (Ding and Weinberg, 2007). It is possible that these sGC expressing amacrine cells are those that express PDE9A, and that these two enzymes regulate cGMP in a light-dependent manner. Since cGMP activates cGMP-dependent kinases (such as PKG), and since phosphorylation crucially controls exocytosis and endocytosis (Liu, 1997), absence of PDE9A can alter neurotransmiter release. Thus, if PDE9A-expressing cells inhibit bipolar cells in a time-dependent manner to regulate their response size and kinetics; absence of PDE9A, which likely causes an accumulation of cGMP, would render bipolar cells' responses slower. Based on the characteristics of the b-wave seen in the absence of PDE9A and their resemblance to the pharmacological results described above, we predict that these cells spike at photopic intensities. Finally, although PDE9A is likely present also in ganglion cells, we do not think that they contribute to reducing the ERG b-wave and slowing down its kinetics because in WT, ganglion cells do not contribute to the ERG a- and b- waves (Bui and Fortune, 2004; Li et al., 2005; Mojumder et al., 2008). Ganglion cells, however, do contribute to the photopic negative response (PhNR) (Li et al., 2005), and although not quantified here, this wave might have been affected.

## **PDE9A LIKELY REDUCES THE CONE-GENERATED a-WAVE DUE TO AN EFFECT ON THE OFF PATHWAY**

We found that in the absence of PDE9A, the photopic a-wave elicited by both green and saturating flashes was smaller. In rodents, the photopic a-wave is generated by the activity of cones and post-receptoral neurons, which includes a small contribution from OFF bipolar cells and horizontal, and perhaps a larger contribution from amacrine cells (Xu et al., 2003; Sharma et al., 2005; Mojumder et al., 2008; Shirato et al., 2008). We think that the reduced a-wave in KO represents reduced current contribution from the inner retinal cells and not reduced cone activity because cones do not express PDE9A. Thus, the same disrupted signal processing that explains the reduced ON bipolar cells' activity, as reflected by the ERG b-wave in KO, could explain the reduced inner retina component of the photopic a-wave (either an indirect influence on OFF bipolar cells' activity or reduced current from amacrine cells). This interpretation also explains why the awave of the mix rod-cone ERG was smaller in KO than in WT retina. With such bright flashes, both cones and inner retinal cells likely contribute to the negative a-wave. In contrast, rod function has not been altered, as can be inferred from both the similar Lamb and Pugh amplification constants and the similar scotopic responses in KO and WT.

#### **CONCLUSION**

The results of our study indicate that PDE9A, by hydrolyzing cGMP, controls its levels and thereby modulates inhibitory processes in the retina. We speculate that light increases nitric oxide levels in the cone pathway of the inner retina; this stimulates sGCs, increases cGMP levels, and modulates inhibition. By hydrolyzing cGMP, PDE9A restricts the duration of the inhibitory processes, thus sharpening and accelerating retinal signaling.

#### **ACKNOWLEDGMENTS**

This work was supported by National Institutes of Health Grants EY11105 (Noga Vardi), EY013333 (Michael A. Freed), and NEI P30 EY01583 (Vision Research Core of the University of Pennsylvania). We thank Pfizer and especially Dr. Rita Balice Gordon and Dr. Frank S. Menniti for providing the PDE9A KO mice. We also thank Drs. Sergei Nikonov, Michael A. Freed, and Robert Smith for multiple useful discussions, Dr. Narender Dhingra for reading the manuscript, and Dr. Sergei Nikonov and Richard Zorger for writing the MATLAB program to analyze ERG.

#### **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: 26 March 2014; accepted: 09 June 2014; published online: 27 June 2014. Citation: Dhingra A, Tummala SR, Lyubarsky A and Vardi N (2014) PDE9A is expressed in the inner retina and contributes to the normal shape of the photopic ERG waveform. Front. Mol. Neurosci. 7:60. doi: 10.3389/fnmol.2014.00060*

*This article was submitted to the journal Frontiers in Molecular Neuroscience. Copyright © 2014 Dhingra, Tummala, Lyubarsky and Vardi. 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.*

## Guanylyl cyclase/natriuretic peptide receptor-A signaling antagonizes phosphoinositide hydrolysis, Ca2**<sup>+</sup>** release, and activation of protein kinase C

## *Kailash N. Pandey\**

Department of Physiology, School of Medicine, Tulane University Health Sciences Center, New Orleans, LA, USA

#### *Edited by:*

Teresa Duda, Salus University, USA

#### *Reviewed by:*

Sandhya Srikant Visweswariah, Indian Institute of Science, India Teresa Duda, Salus University, USA

#### *\*Correspondence:*

Kailash N. Pandey, Department of Physiology, School of Medicine, Tulane University Health Sciences Center, 1430 Tulane Avenue, New Orleans, LA 70112, USA e-mail: kpandey@tulane.edu

Thus far, three related natriuretic peptides (NPs) and three distinct sub-types of cognate NP receptors have been identified and characterized based on the specific ligand binding affinities, guanylyl cyclase activity, and generation of intracellular cGMP. Atrial and brain natriuretic peptides (ANP and BNP) specifically bind and activate guanylyl cyclase/natriuretic peptide receptor-A (GC-A/NPRA), and C-type natriuretic peptide (CNP) shows specificity to activate guanylyl cyclase/natriuretic peptide receptor-B (GC-B/NPRB). All three NPs bind to natriuretic peptide receptor-C (NPRC), which is also known as clearance or silent receptor. The NPRA is considered the principal biologically active receptor of NP family; however, the molecular signaling mechanisms of NP receptors are not well understood.The activation of NPRA and NPRB produces the intracellular second messenger cGMP, which serves as the major signaling molecule of all three NPs.The activation of NPRB in response to CNP also produces the intracellular cGMP; however, at lower magnitude than that of NPRA, which is activated by ANP and BNP. In addition to enhanced accumulation of intracellular cGMP in response to all three NPs, the levels of cAMP, Ca2<sup>+</sup> and inositol triphosphate (IP3) have also been reported to be altered in different cells and tissue types. Interestingly, ANP has been found to lower the concentrations of cAMP, Ca2+, and IP3; however, NPRC has been proposed to increase the levels of these metabolic signaling molecules. The mechanistic studies of decreased and/or increased levels of cAMP, Ca2+, and IP3 in response to NPs and their receptors have not yet been clearly established. This review focuses on the signaling mechanisms of ANP/NPRA and their biological effects involving an increased level of intracellular accumulation of cGMP and a decreased level of cAMP, Ca2+, and IP3 in different cells and tissue systems.

**Keywords: natriuretic peptides, natriuretic peptide receptors, membrane guanylyl cyclases, cGMP, cAMP, Ca2+, inositol triphosphate**

## **INTRODUCTION**

Atrial natriuretic factor/peptide (ANF/ANP) is produced and secreted in the specific granules of cardiac atrial myocytes, which participates in the control of extracellular fluid volume, electrolyte balance, and mean arterial pressure, thus, it plays a central role in the maintenance and regulation of cardiovascular homeostasis (de Bold et al., 1981; de Bold, 1985; Brenner et al., 1990; Anand-Srivastava and Trachte, 1993; Pandey, 2005, 2011). In addition to its natriuretic, diuretic, vasorelaxant, antimitogenic, antihypertrophic, and anti-inflammatory activities, ANP inhibits the release of renin from the kidneys, aldosterone from the adrenal glands, vasopressin from posterior pituitary, and progesterone from Leydig tumor (MA-10) cells, while stimulating the synthesis and release of testosterone from normal Leydig cells in the testes, progesterone from granulosa-leuteal cells, and luteinizing hormone from anterior pituitary gland (Inagami, 1989; Brenner et al., 1990; Levin et al., 1998; Pandey, 2005). A number of studies have documented that ANP has always been found to increase the intracellular accumulation of cGMP, however, to decrease the levels of cAMP; Ca2+, and inositol triphosphate (IP3) in agonist

hormone-treated cells and tissues (Waldman et al., 1984; Pandey et al., 1985, 1988; Khurana and Pandey, 1993, 1996; Pandey, 2005; Turovsky et al., 2013). It has also been suggested that ANP decreases the cAMP levels by stimulating the cGMP-specific phosphodiesterases; however, in certain cells and tissue types, ANP did not decrease or change the cAMP concentrations. Several studies have indicated that ANP diminishes the Ca2<sup>+</sup> signals probably by activating the Ca2<sup>+</sup> extrusion processes by protein kinase-G (PKG) specifically in endothelial and vascular smooth muscle cells (VSMCs; Rashatwar et al., 1987; Zolle et al., 2000; Pandey, 2005).

Among the natriuretic peptides (NPs) hormone family, ANP is the first described member, later, two other members of NP family; brain natriuretic peptide (BNP) and C-type natriuretic peptide (CNP) were identified and characterized, which also exhibited biochemical and structural properties similar to ANP; however, each prepro NP hormone is encoded from a separate gene (Rosenzweig and Seidman, 1991; Levin et al., 1998). Although, all three NPs (ANP, BNP, and CNP) have highly homologous structure, they bind to specific NP receptors and elicit discrete biological and physiological functions (Brenner et al., 1990). There have been three subtypes of NP receptors, namely; NP receptor-A, NP receptor-B, and NP peptide receptor-C (NPRA, NPRB, and NPRC, respectively), which were identified and characterized by molecular cloning (Pandey and Singh, 1990; Garbers, 1992). Interestingly, both ANP and BNP bind to NPRA, which produces the intracellular second messenger cGMP; however, CNP binds NPRB, which also generates cGMP, but all three NPs indiscriminately bind and activate NPRC, which lacks GC catalytic domain (Fuller et al., 1988; Drewett and Garbers, 1994; Lucas et al., 2000; Sharma, 2002; Pandey, 2005). The cellular, biochemical, and pharmacological aspects of NPs and their cognate receptors have revealed classic hallmark of physiological and pathophysiological functional significance, including; renal, cardiac, vascular, neuronal, and immunological aspects in health and disease (Pandey, 2005, 2011; Kishimoto et al., 2011). It has been suggested that ANP suppresses Na+-reabsorption at the collecting duct of the kidneys, inhibits renin synthesis and release, and stimulates natriuresis and diuresis, thereby, lowers blood pressure and blood volume and maintains cardiovascular homeostasis (de Bold, 1985; Brenner et al., 1990; Levin et al., 1998; Pandey, 2005). In the vasculature, ANP relaxes VSMCs thus causing the immediate vasorelaxant effect in the vascular bed (Levin et al., 1998; Pandey, 2005).

The expression and activity of NPRA is regulated by various hormonal agents, including its ligand ANP (Pandey, 1993, 2005; Cao et al., 1995; Pandey et al., 2002). The studies detailing the *Npr1* (coding for NPRA) gene-disruption in mice have revealed the functional significance of NPRA in the control of blood pressure and cardiovascular disease states (Oliver et al., 1997; Shi et al., 2001; Holtwick et al., 2002; Vellaichamy et al., 2005; Kishimoto et al., 2011; Pandey, 2011; Yoshihara et al., 2014). Mice lacking NPRA develop high blood pressure and severe cardiac hypertrophy, fibrosis, and disorders that are reminiscent of heart disease as seen in untreated human hypertensive patients (Vellaichamy et al., 2007, 2014; Zhao et al., 2013). The regulated expression of CNP is derived from endothelial cells, which targets NPRB on the adjacent smooth muscle cells (Suga et al., 1992). Thus, the principal role of CNP is considered as a direct vasodilator involved in the regulation of vascular tone through activation of GC-B/NPRB on smooth muscle cells in the vascular beds (Hama et al., 1994). The objective of this current review is to summarize and document the findings and discoveries with particular emphasis of cellular signaling and physiological and pathological significance of ANP/NPRA in relation to the increased production of intracellular second messenger cGMP and inhibition of the phosphoinositide (IP3) hydrolysis, Ca<sup>2</sup><sup>+</sup> release, and protein kinase C (PKC) activity in target cells.

## **HISTORICAL BACKGROUND**

Thirty-three years ago, the pioneer discovery by de Bold and his coworkers established that atrial extracts contained natriuretic and diuretic activity which led to the isolation and nomenclature of ANF, usually referred to as ANP (de Bold et al., 1981; de Bold, 1985). Now, it is considered that ANP is primarily synthesized and secreted in the granules of heart atrium and BNP is largely synthesized in the heart ventricle and displays most variability

in the primary structure. Although, the atrium is the primary site of synthesis for ANP, however, ventricle also produces ANP but at the levels of 100-fold to 1000-fold lower than that of the atrium, respectively (Kojima et al., 1989). CNP was isolated from the porcine brain, however, is mostly present in the endothelial cells of the vasculature and is highly conserved among the mammalian species (Rosenzweig and Seidman, 1991). The primary structure deduced from the cDNA synthesis, suggested that ANP is synthesized first as the 152-amino acid prepro-ANP molecule that contains sequences of active peptide in its carboxyl-terminal region (Maki et al., 1984). The biologically active ANP is released by proteolytic cleavage of pro-ANP molecule into predominantly 28-amino acid active (residues 99–126) and the 98 amino acid inactive (residues 1–98) molecules. The active form of ANP has a disulfide-bonded loop between cysteine 105 and 121, which seems to be essential for the biological activity (Brenner et al., 1990). Initially, different lengths of sequences of ANP were identified and synthesized for the studies of structure-activity relationship, and it was suggested that the ring structure of ANP with a disulfidebonded loop is essential for its biological activities (Rosenzweig and Seidman, 1991). All three NPs contain highly conserved amino acid sequences with a 17-residues disulfide-bonded ring but deviate from each other in the N-terminal and C-terminal flanking amino acid sequences. Furthermore, the C-terminal sequence extending from the ring structure to Asn-Phe-Arg-Tyr is essential for the biological activity of ANP. The amino acid sequence of ANP is almost identical across the mammalian species, except at the position 10, which is substituted with isoleucine in rat, mouse, and rabbit, however, in human, dog, and bovine, ANPs have methionine in this position (Misono et al., 1984). Subsequently, BNP and CNP were both isolated and characterized from the porcine brain extracts (Sudoh et al., 1988, 1990). BNP is predominantly synthesized and secreted from the heart ventricle (Phillips et al., 1991). Similarly, CNP is predominantly localized in the central nervous system and endothelial cells and is considered as a non-circulatory peptide hormone (Suga et al., 1992).

## **NATRIURETIC PEPTIDES SYNTHESIS AND SECRETION**

It has been suggested that the processing of preprohormone to prohormone molecule and the cleavage and secretion of biologically active mature 28-residue ANP molecule occurs predominantly in response to atrial distension (de Bold, 1985; Brenner et al., 1990). Usually, ANP concentration ranges from 50 to 100-fold higher than BNP; however, the expression of both ANP and BNP increases dramatically in the atrium and ventricle in the condition of cardiac disorders and heart failure (Mukoyama et al., 1991). During the disease states, the ventricle becomes the primary site of synthesis and release for BNP. In congestive heart failure (CHF) patients, the concentrations of both ANP and BNP increase greater than the control values, however, the BNP concentration increases 10-fold to 50-fold higher than a comparative increases in the ANP levels (Mukoyama et al., 1991). Those previous findings indicated that ANP and BNP elicit distinct physiological and pathophysiological effects, nevertheless, both hormones show similar hemodynamic responses, but BNP exerts a longer duration of action and causes enhanced natriuretic responses as compared

with ANP (Yoshimura et al., 1991; Omland et al., 1996). It has been suggested that the cardiac atrium expresses almost 50-fold to 100 fold or even higher levels of ANP mRNA as compared with extra -cardiac tissues (Gardner et al.,1986). Interestingly, higher ventricular ANP levels have been found in the developing embryos and fetuses; however, both mRNA and peptide levels of ANP decline rapidly during the prenatal period (Cameron et al., 1996). On the other hand, CNP does not seem to behave as a cardiac hormone and its concentration is extremely low in the circulation (Igaki et al., 1996). It is believed that CNP is largely localized in the central nervous system and in the vascular endothelial cells (Ogawa et al., 1992; Suga et al., 1992, 1993; Tamura et al., 1996; Chen and Burnett, 1998). Another class of NPs is the D-type natriuretic peptide (DNP) that represents an additional member in the NP hormone family and is largely present in the venom of the green mamba (*Dendroaspis angusticeps*) as a 38-amino acid peptide molecule (Schweitz et al., 1992; Lisy et al., 1999). In addition, a 32-amino acid peptide termed as urodilatin (URO) is identical to C-terminal sequence of pro-ANP, which is largely present only in the urine (Schulz-Knappe et al., 1988). Initially, URO was purified from the human urine and is considered to be synthesized only in the kidneys (Saxenhofer et al., 1990). The immunohistochemical staining indicated that URO is largely present in the cortical tubules around the collecting ducts of the kidneys (Meyer et al., 1996; Doust et al., 2005).

In the circulation, the half-life of BNP is greater than ANP, thus the evaluation of the diagnostic importance of the NPs have mostly favored BNP. The inactive N-terminal fragment of BNP (NT-proBNP) has even a greater half-life than the BNP. The plasma levels of both BNP and NT-proBNP are markedly elevated under the pathophysiological conditions of cardiac dysfunction, including diastolic dysfunction, CHF, and pulmonary embolism (Felker et al., 2006; Jaffe et al., 2006). The basal plasma levels of BNP vary from 5 to 50 pg/ml and NT-proBNP levels range from 10 to 150 pg/ml. An abnormal range is considered as 100 pg/ml for BNP and 125 pg/ml for NT-proBNP (Felker et al., 2006). Nevertheless, the secretion of both ANP and BNP from the ventricular myocytes increases proportionally in relation to the magnitude of cardiac

dysfunction or heart failure condition (Yoshimura et al., 1993). It has been suggested that BNP acts as an important prognostic indicator in the CHF patients, however, NT-proBNP is considered to be a stronger risk bio-indicator for cardiovascular events (Doust et al., 2005). Both BNP and NT-proBNP seem to provide an ideal tool to be utilized as blood tests to diagnose cardiac disorders in patients with high risk of heart failure, diabetes, chronic kidney disease, and coronary artery disease (Khan et al., 2006; Freestone et al., 2008; Czucz et al., 2011; Ganem et al., 2011). The BNP level is increased to almost 200 pg/ml and NT-proBNP levels reaches to approximately 1200 pg/ml in patients with reduced creatinin clearance (McCullough et al., 2003; Anwaruddin et al., 2006).

### **IDENTIFICATION AND CHARACTERIZATION OF RECEPTOR MEMBRANE GUANYLYL CYCLASE**

The previous studies using cross-linking and photoaffinity labeling procedures, have shown the existence of NP receptors with a wide range of molecular weight (M*r*) of the 60–180 kDa (Misono et al., 1985; Schenk et al., 1985; Vandlen et al., 1985; Meloche et al., 1986; Pandey et al., 1986). Initially, NP receptors were identified with varying receptor density in different cells and tissue types (**Table 1**). Subsequently, high affinity ANP binding sites were with GC activity were co-purified (Kuno et al., 1986; Paul et al., 1987; Takayanagi et al., 1987; Meloche et al., 1988) On the basis of biological activity of different ANP analogs, NP receptors were classified and characterized as biologically active and clearance or silent receptors (Maack et al., 1987). Subsequently, three distinct subtypes of NP receptors were identified, which appeared to be specific to different cells and tissues (Pandey et al., 1988). Based on the cellular, biochemical, and molecular biological studies, the NPs and their receptors are quite widespread in cell and tissue distributions (Leitman et al., 1988; Pandey et al., 1988; Brenner et al., 1990; Marala et al., 1992; Levin et al., 1998; Pandey, 2005). Molecular cloning and expression of cDNA from mouse, rate, and human, led to identify and characterize the primary structure of three distinct subtypes of NP receptors, which are currently designated as GC-A/NPRA, GC-B/NPRB, and NPRC (Fuller et al.,


**Table 1 | ANP-dependent binding parameters of GC-A/NPRA and intracellular accumulation of cGMP in different cell types.**

HEK-293 cells, human embryonic Kidney-293 cells; kd, dissociation constant; Bmax , receptor density; MA-10 cells, Leydig tumor cells; MDCK cells, Maiden-Darby kidney epithelial cells; N4TG1 cells, neuroblastoma cells; RTASM, rat thoracic aortic smooth muscle cells.

1988; Chinkers et al., 1989; Schulz et al., 1989; Pandey and Singh, 1990; Duda et al., 1991). The general topological structures of GC-A/NPRA and GC-B/NPRB are consistent with at least four distinct domains, including extracellular ligand-binding domain, a single transmembrane spanning region, and intracellular protein kinase-like homology domain (protein-KHD), and GC catalytic domain. The transmembrane GC receptors contain a single cyclase catalytic site per protein molecule, however, based on the structural modeling data two polypeptide chains seem to be required to activate GC-A/NPRA (Wilson and Chinkers, 1995; Labrecque et al., 1999; van den Akker et al., 2000). It has been indicated that the dimerization region of the receptor is located between the KHD and GC catalytic domain that has been predicted to form an amphipathic alpha helix structure. The GC-B/NPRB has the overall domain structure similar to that of GC-A/NPRA with binding affinity to CNP also produces the intracellular second messenger cGMP (Schulz et al., 1989; Koller et al., 1991; Lucas et al., 2000). NPRA is considered as the dominant subtype of the NP receptors found in peripheral organs and mediates most of the known functions of ANP and BNP hormones. Nevertheless, NPRB is localized mainly in the central nervous system and vascular tissues, which is thought to mediate the actions of CNP in the brain and also in the vascular bed. There are increasing numbers of other GC receptors; however, the specific ligands for these receptors are still being identified (**Table 2**). The third member of the NP receptor family, NPRC, constitutes a large extracellular domain of 496-amino acids, a single transmembrane domain, and a very short 37-amino acid cytoplasmic tail that contains no sequence homology with any other known membrane receptor proteins and has been given the name by default as clearance receptor (Fuller et al., 1988). The extracellular region of NPRC is approximately 30% identical to both GC-A/NPRA and GC-B/NPRB. Studies using the ligand receptor binding as a criterion, have shown that NPRC has much less stringent specificity and affinity for structural variants of ANP than does NPRA (Bovy, 1990). The extracellular domain of NPRC possesses two pairs of cysteine residues along with one isolated cysteine near the transmembrane domain of the receptor. However, three potential signals for N-glycosylation and several serine and threonine for O-linked glycosylation sites are known to be present in the extracellular domain of NPRC (Fuller et al., 1988). Previously, it has been suggested that NPRC may function as a clearance receptor to remove and clear NPs from the circulation, however, a number of studies have provided the evidence that NPRC plays roles in the biological actions of NPs (Anand-Srivastava and Trachte, 1993; Matsukawa et al., 1999; Zhou and Murthy, 2003). Thus, it is evident that the clearance name carries only by a default nomenclature to NPRC.

#### **INTERNALIZATION AND DOWN-REGULATION OF GC-A/NPRA**

The ligand-dependent internalization plays important role in the receptor down-regulation and signaling process. Down-regulation of GC-A/NPRA has been reported in a number of cells, including PC-12 cells containing endogenous receptors and transfected


**Table 2 |The distribution of natriuretic peptide receptors (NPRA, NPRB, and NPRC) and their gene-knockout phenotype.**

NPRA, natriuretic peptide receptor-A; Npr1, coding for guanylyl cyclase/natriuretic peptide receptor-A; NPRB, natriuretic peptide receptor-B; Npr2, coding for guanylyl cyclase/natriuretic peptide receptor-B; NPRC, natriuretic peptide receptor-C; Npr3, coding for natriuretic peptide clearance receptor.

COS-7 and HEK-293 cells harboring recombinant receptors (Rathinavelu and Isom, 1991; Pandey, 1993; Pandey et al., 2000a, 2002). The carboxyl-terminal deletion mutation has shown that the specific sites in the GC catalytic domain and KHD of NPRA, play critical roles in the endocytosis and sequestration of the receptor (Pandey et al., 2000a). Previous studies have also indicated that after prolonged treatment of cultured cells with ANP, both the receptor density and GC activity were decreased with simultaneous reduction in mRNA of levels NPRA (Fujio et al., 1994; Cao et al., 1995, 1998; Hum et al., 2004). In addition, transforming growth factor-β1 (TGF-β1), angiotensin II (ANG II), and endothelin (ET-1) have also been shown to reduce mRNA levels of NPRA in various types of cultured cells (Fujio et al., 1994; Chen and Gardner, 2003; Garg and Pandey, 2003; Arise and Pandey, 2006). Those previous studies demonstrated that a decrease in mRNA levels of GC-A/NPRA correlated with the repressed transcriptional activity of the receptor. On the other hand, mRNA levels of NPRA are greatly increased by retinoic acid and histonedeacetylase inhibitor treatments (Kumar et al., 2010, 2014a,b). It has been suggested that NPRA exists in the phosphorylated state and the addition of ANP causes a decrease in the phosphate contents as well as reduction or desensitization of the ANP-dependent GC catalytic activity of NPRA Potter and Garbers, 1992). The apparent mechanism of desensitization of NPRA is in contrast to many other cell-surface hormone receptors, which appear to be desensitized by phosphorylation (Sibley et al., 1987; Hugnir and Greengard, 1990; Lefkowitz et al., 1998; Sorkin and von Zastrow, 2002). The initial findings have also indicated that ANP stimulates phosphorylation of NPRA (Ballerman et al., 1988; Pandey, 1989; Duda and Sharma, 1990; Larose et al., 1992). Later, it was suggested that cGMPdependent PKG, a serine/threonine kinase is also phosphorylates NPRA (Airhart et al., 2003).

The down-regulation of NPRC also seems to be associated with increased internalization of the ligand-receptor complexes involving receptor-mediated endocytosis and trafficking mechanisms of this receptor protein (Pandey, 1992). The phenomenon of down-regulation of NPRC has been largely documented in cultured VSMCs, which predominantly contain a high density of NPRC (Neuser and Bellermann, 1986; Hirata et al., 1987; Hughes et al., 1987; Pandey, 1992; Anand-Srivastava, 2000). The metabolic processing of ANP involving NPRC has been reported by several investigators utilizing VSMCs (Hirata et al., 1985; Napier et al., 1986; Murthy et al., 1989; Nussenzveig et al., 1990; Pandey, 1992, 2005; Cohen et al., 1996; Anand-Srivastava, 2000). It has been suggested that a population of the internalized NPRC also recycles back to the plasma membrane (Pandey, 1992).

## **ACTIVATION OF GC-A/NPRA GENERATES INTRACELLULAR SECOND MESSENGER cGMP**

It is believed that cGMP is generated as a result of ANP binding to the extracellular domain of GC-A/NPRA, which probably allosterically regulates an increased activity of the receptor protein (Pandey, 1993, 2005, 2011; Drewett and Garbers, 1994; Pandey et al., 2002; Sharma, 2002; Sharma and Duda, 2010). The initial findings showed that ANP markedly increases cGMP in target tissues in a dose-related manner (Hamet et al., 1984; Waldman et al., 1984; Pandey et al., 1985). Previous studies have also indicated

that the binding of ANP to GC-A/NPRA by itself is probably not sufficient to stimulate GC catalytic activity and the production of cGMP, however, it requires ATP (Kurose et al., 1987; Chinkers et al., 1991; Goraczniak et al., 1992). Because the non-hydrolyzable analogs of ATP mimicked ANP effect, it was suggested that ATP acts directly by allosteric regulation of GC catalytic activity of NPRA. Both the ligand binding and the interaction of ATP with the KHD of the receptor increase the cGMP production without affecting the affinity for the substrate (Kurose et al., 1987; Chang et al., 1990; Duda et al., 1991). Molecular cloning and overexpression of NPRA demonstrated that GC catalytic domain cannot be activated by ANP alone without ATP-binding to KHD region of the receptor (Chinkers et al., 1991; Larose et al., 1991; Wong et al., 1995). Further studies provided the evidence that ATP binding to KHD of NPRA is important for the effectors coupling of GC family of receptors (Goraczniak et al., 1992; Sharma, 2002).

Deletion of the KHD of GC-A/NPRA and GC-B/NPRB has been suggested that KHD represses the GC catalytic activity of these receptors (Chinkers et al., 1989). At the same time, another model was proposed indicating that KHD was not a repressor; however, ATP was required to activate the catalytic domain of NPRA (Goraczniak et al., 1992; Sharma, 2002). Both NPRA and NPRB contain a glycine-rich ATP binding motif within the KHD, which is known as glycine-rich consensus sequence (Duda et al., 1991, 1993; Goraczniak et al., 1992). The juxtamembrane hinge structure of NPRA undergoes a significant conformational change in response to ligand binding, and it may play an important role in transmembrane signaling process (Huo et al., 1999). The amino acid sequence near the transmembrane region is well conserved in GC-A/NPRA that contains several closely located proline residues and a pair of cysteine residues. The mutation of one of the proline in this region renders the receptor to bind the ligand but blocks GC catalytic activity (Huo et al., 1999). Similarly, in the juxtamembrane hinge region, the elimination of disulfide bond of cysteine residues resulted in constitutive activation of NPRA. Those previous findings suggested that juxtamembrane hinge region of NPRA may play a critical role in receptor activation and signal transduction mechanisms of GC-coupled receptors.

The glycosylation of the receptor seems to be essential for ligand binding activity of GC-A/NPRA (Lowe and Fendly, 1992; Fenrick et al., 1997). However, it has also been suggested that glycosylation may not be required for ligand binding of NPRA (Miyagi et al., 2000). The mutational analyses of N-linked glycosylation consensus sites in guanylyl cyclase-C (GC-C) have indicated that certain amino acid residues might be important for receptor stability (Hesegawa et al., 1999). The glycosylation sites onto the GC-A/NPRA binding domain have been found to be scattered on the surface of the receptor with the exception of the hormone binding site and dimer interface (van den Akker, 2001). The glycosylation sites have been implicated to function in proper folding and stability of NPRA (Lowe and Fendly, 1992; Koller et al., 1993; Heim et al., 1996). Nevertheless, the glycosylation of the extracellular domain of NPRA can be considered of significant importance for receptor orientation and packaging on the cell surface similar to that of other plasma membrane receptor proteins (Wormald

and Dwek, 1999). Nevertheless, it should be noted that there is no appreciable conservation of the precise position of the glycosylation sites within the members of GC-receptor family. Clearly, more studies are needed to confirm the functional roles of glycosylation in the transmembrane signaling processes of both GC-A/NPRA and GC-B/NPRB protein molecules.

## **ANP/NPRA SIGNALING INHIBITS PHOSPHOINOSITIDE HYDROLYSIS, Ca2<sup>+</sup> RELEASE, AND PKC ACTIVITY**

Previous studies have demonstrated that ANP significantly decreased the hydrolysis of phosphoinositide in murine Leydig tumor (MA-10) cells in a dose-dependent manner and the H-8, a specific inhibitor of PKG, reversed the inhibitory effect of ANP on the generation of inositol phosphates, supporting the involvement of PKG in this process (Khurana and Pandey, 1995). ANP has also been shown to inhibit both autophosphorylation and enzymatic activity of PKC in different cell systems (Pandey, 1989, 1994a,b; Kumar et al., 1997). It is not yet clear if the ANPdependent inhibitory effects on the phosphoinositide metabolism and PKC autophosphorylation and/or enzyme activity are exerted in a composite manner to negatively regulate the phosphoinositide, Ca2+, and PKC involving ANP/NPRA/cGMP/PKG cascade. It is also possible that the effect of ANP is transmitted to block the IP3 and Ca2<sup>+</sup> signaling pathways independently in response to particular agonist stimulation. It has been suggested that potassium channels can be stimulated by ANP through the activation of PKGs, which require ATP and G-proteins (White et al., 1993). However, the possible involvements of potassium channels in the ANP-dependent inhibitory responses on the generation of inositol phosphates are not yet clearly understood. ANP has also been shown to stimulate the formation of inositol phosphates in culturedVSMCs, however, in the inner medullary collecting duct cells and smooth muscle tissues, ANP stimulated the production of inositol phosphates at lower dosages, and inhibited the formation of these metabolites at higher dosages, which increase intracellular generation of cGMP (Resink et al., 1988; Hirata et al., 1989; Teitelbaum et al., 1990; Berl et al., 1991). Thus the heterogeneity of NP receptors and their diverse cellular distribution suggest that different mechanisms might be involved in the cellular action of ANP/NPRA/cGMP (Anand-Srivastava and Trachte, 1993; Pandey, 2001, 2002, 2011). It has also been shown that ANP inhibits the thrombin-induced synthesis and release of endothelin in cultured rat aortic endothelial cells by blocking the phosphoinositide breakdown (Emori et al., 1993).

In addition to the stimulatory effect of ANP on GC activity, it has also been shown to reduce adenylyl cyclase and phospholipase C activities, sodium influx, and Ca2<sup>+</sup> concentrations (Brenner et al., 1990; Anand-Srivastava and Trachte, 1993; Pandey, 2005). The increased production of cGMP in response to ANP correlates with the effects of dibutyryl-cGMP. The most compelling evidence supporting a role for cGMP effects was obtained with selective NPRA antagonists, A71915 and HS-121-1 in the kidneys (von Geldern et al., 1990; Sano et al., 1992). Those previous studies established that ANP effect is largely mediated by cGMP through the activation of GC-A/NPRA. In general, evidence suggests that biological activity of ANP/NPRA enhances the generation of the intracellular second messenger cGMP and decreases

the levels of cAMP, Ca2+, and IP3 along with the antagonistic effects on PKC and mitogen-activated protein kinases (MAPKs) in target cells (**Figure 1**). ANP has been reported to induce cGMP-dependent acrosomal reaction in both capacitated and non-capacitated spermatozoa (Anderson et al., 1994). Furthermore, the acrosome reaction was essentially equal in magnitude when induced with ANP or Ca2<sup>+</sup> ionophore A23187. However, higher concentrations of ANP were required to induce acrosomal reaction in capacitated as compared with non-capacitated spermatozoa. Those previous findings indicated that ANP-induced human acrosomal reaction does not require physiological concentrations of extracellular Ca2+. Acrosomal reaction is known to involve various extracellular signals, including cAMP (Anderson et al., 1992), cGMP (Komatsu et al., 1990), prostaglandins, Ca2<sup>+</sup> and IP3 (Thomas and Meizel, 1989), and diacylglycerol (Breitbart et al., 1992).

The established biochemical and cellular effects of ANP in the adrenal glomerulosa cells showed the activation of GC activity and K<sup>+</sup> channel conductance; whereas T-type Ca2<sup>+</sup> channels conductance and adenylyl cyclase activity are suppressed (Anand-Srivastava and Trachte, 1993). The correlative evidence between ANP-induced cGMP accumulation and vasodilation has suggested the role of cGMP as the intracellular second messenger of dilator responses to ANP (Brenner et al., 1990; Anand-Srivastava and Trachte, 1993; Cao et al., 1995; Pandey, 2005). ANP as well as cGMP analogs have been found to reduce the agonist-induced increases in cytosolic Ca2<sup>+</sup> concentrations (Hassid, 1986; Lincoln et al., 1994; Pandey, 2005). It has been suggested that cGMP activates sarcolemmal Ca2+-ATPase, and this mechanism seems to be important in the ANP-induced decreases in cytosolic Ca2<sup>+</sup> in VSMCs (Rashatwar et al., 1987; Cornwell and Lincoln, 1989; Levin et al., 1998; Pandey, 2005). Nevertheless, it is anticipated that the ultimate effect of ANP in VSMCs could be due to production of cGMP and the activation of PKG (Lincoln et al., 1994; Kumar et al., 1997). However, more studies are needed to define the biochemical and molecular basis of NP actions in vasculature, including VSMCs and endothelial cells.

Initial studies from our laboratory and data published from others have also shown that both ANP and cGMP inhibited the autophosphorylation and enzymatic activity of PKC in the plasma membrane preparations of various target cells (Rogers et al., 1988; Sauro and Fitzpatrick, 1990; Pandey, 1994a,b; Kumar et al., 1997). The activation of PKC triggers the agonist-dependent phosphorylation and activity of numerous cellular proteins causing alteration in many physiological and pathophysiological conditions, including hypertension, cardiac hypertrophy, ischemia, atherosclerosis, stroke, and neurological disorders (Louis et al., 1988; Turla et al., 1990; Komuro et al., 1991; Kumar et al., 1997). PKC is believed to be a multigene family, consisting of at least 12 isoenzymes that can be classified into classical, novel, and atypical forms (Hug and Sarre, 1993; Dekker and Parker, 1994). These PKC isoenzymes are multifunctional serine/threonine kinases that are largely activated by Ca2+/phospholipids and phorbol esters. However, some of these isoforms (e, δ, η´, and ϕ) do not require Ca2+, while other isoforms (ζ and ε) do not require Ca2<sup>+</sup> or phospholipid for PKC enzymatic activity. Previous studies have indicated that vasoconstrictive agents, including ANG II and

indicated.

ET-1, were able to activate several-fold PKC activity in cultured VSMCs, however, ANP potently antagonized the ANG II- and ET-1-stimulated PKC activity in the ANP/NPRA-dependent manner (Kumar et al., 1997; Pandey, 2005). The inhibitory effect of ANP was greatly amplified if cell were transfected with both PKC-α and NPRA cDNAs. The pretreatment of cells with NPRA antagonist A-71915, significantly blocked the production of cGMP as well as the inhibitory effect of ANP on PKC activity (Kumar et al., 1997). The results of those previous studies provided strong evidence that ANP antagonizes the PKC activation involving ANP/NPRA/cGMP signaling cascade. Agonists that activate PKC also produce two distinct second messengers, IP3, which activates cytosolic free Ca2<sup>+</sup> and diacylglycesol, which stimulates PKC activity (Berridge and Irvine, 1989; Exton, 1990; Rasmussen et al., 1995; Kumar et al., 1997). Our previous studies have suggested that ANP inhibits the formation of IP3 in a cGMP-dependent manner in the intact cells, suggesting that the inhibitory effect of ANP on PKC activity might be linked with its antagonistic action on IP3 formation, however, more studies are needed to support these observations in various ANP-responsive cell and tissues systems.

a number of pathways including; intracellular formation of cAMP, Ca2+, IP3;

## **EFFECT OF NPRA ON THE INHIBITION OF MAPKs ACTIVITY AND CELL PROLIFERATION**

It has been shown that cGMP analogs mimicked the antiproliferative action of ANP, indicating that it exerts the antimitogenic effects largely through the intracellular second messenger cGMP (Lincoln et al., 1994; Hutchinson et al., 1997; Pandey et al., 2000b; Sharma et al., 2002). ANP has been shown to inhibit collagen synthesis in cardiac fibroblasts and also it inhibits hypertrophy of cardiac myocytes (Calderone et al., 1998; Masciotra et al., 1999; Silberbach et al., 1999; Horio et al., 2000; Gopi et al., 2013). Similarly, PKG has been shown to suppress extracellular matrix production in VSMCs (Dey et al., 1998). Both NPRA and NPRC, have been suggested to play a role in ANP-dependent antimitogenic responses (Prins et al., 1996; Hutchinson et al., 1997; Pandey et al., 2000b; Sharma et al., 2002; Tripathi and Pandey, 2012). ANP has been shown to act as a growth suppressor in a variety of cell types including; kidney, heart, neurons, thymus, vasculature, and fibroblasts (Levin et al., 1998; Pandey, 2005). Previous studies have demonstrated that ANP inhibits ANGII- and platelet-derived growth factor (PDGF) -dependent MAPK activity in different tissues and cell types (Sugimoto et al., 1993; Prins

et al., 1996; Pandey et al., 2000b; Sharma et al., 2002; Tripathi and Pandey, 2012). However, in astroglial cells, ANP was shown to inhibit extracellular-regulated MAPK (Erk1/2) activity through NPRC (Prins et al., 1996). In contrast, recent findings have indicated that des- (Cys105–Cys121) -ANP, a ligand selective to NPRC, did not inhibit basal or serum-stimulated MAPK, however, CNP, which acts through NPRB, potently inhibited MAPK activity in fibroblasts in a cGMP-dependent manner (Chrisman and Garbers, 1999).

It has been postulated that cGMP-dependent signaling mechanisms of GC-A/NPRA are initiated probably at the level of gene transcription; however, the exact mechanism of this activation remains to be elucidated. A previous report also indicated that cGMP/PKG signaling was able to increase the MAPK activity in contractile rat VSMCs (Komalavilas et al., 1999). However, the process by which cGMP/PKG leads to the activation of MAPKs is unclear. Similarly, cAMP- and PKG have also been shown to inhibit as well as to activate MAPKs pathways, depending on the cell types and culture conditions (Bornfildt and Krebs, 1999). However, the involvement of specific ANP receptor subtypes in the inhibitory effects of ANP on the agonist-stimulated MAPKs activity is controversial. Indeed, more studies are needed to establish the underlying mechanisms of the antiproliferative effect of ANP in target cells. ANP has also been shown to induce apoptosis in cultured VSMCs and in neonatal rat cardiac myocytes (Trindade et al., 1995; Wu et al., 1997). The apoptotic effect of ANP was mimicked by 8-bromo-cGMP, a membrane-permeable analog of cGMP, and also by nitroprusside, an activator of soluble guanylyl cyclase. Furthermore, the effect of ANP was greatly potentiated by a cGMP-specific phosphodiesterase inhibitor zaprinast. It has been indicated that norepinephrine, a myocyte growth and proliferative effector molecule, inhibited ANP-induced apoptosis via activation of β-adrenergic receptor and elevation of cAMP (Wu et al., 1997). The existence of a complementary ANP-mediated mechanism to inhibit cell growth and proliferation is not anticipated. Nevertheless, the inhibition of cell proliferation is often accompanied by an increased probability of apoptosis, whereas, growth-promoting agents and agonist hormones tend to promote cell growth and proliferation. For instance, ANG II inhibits apoptosis, in contrast, ANP and nitric oxide, both potently inhibit cell growth and proliferation and induce apoptosis (Pollman et al., 1996;Wu et al., 1997). It has been suggested that the anti-apoptotic molecule Bcl-2 homolog Mcl-1 might serve as an important target in ANP-induced apoptosis. Intriguing was the finding that the Bcl-2 homolog Mcl-1 was initially identified as a protein marker, which was up-regulated during the differentiation of the monocytoid cell line ML-1 cells (Kozopas et al., 1993; Kiefer et al., 1995; Wu et al., 1997).

#### **GENE-TARGETING OF** *Nppa* **AND** *Npr1*

Genetic-targeting strategies in mice have provided novel approaches to study the physiological responses corresponding to gene-dosage *in vivo* (Takahashi and Smithies, 1999; Kim et al., 2002). Genetically modified mice carrying *Npr1* gene-disruption or gene-duplication have provided strong support for the physiological roles of NPs and their receptors in the intact animals (John et al., 1995; Lopez et al., 1995; Kishimoto et al., 1996; Oliver

et al., 1997, 1998; Matsukawa et al., 1999; Pandey et al., 1999; Shi et al., 2001, 2003; Holtwick et al., 2002; Vellaichamy et al., 2005; Das et al., 2012; Zhao et al., 2013). Numerous studies have examined the quantitative contributions and possible mechanisms mediating the responses of *Npr1* gene copies by determining the renal plasma flow (RPF), glomerular filtration rate (GFR), urine flow, and sodium excretion following blood volume expansion in *Npr1* homozygous null mutant (*Npr1*−/−*;* 0-copy), wild-type (*Npr1*+/+*;* 2-copy), and gene-duplicated (*Npr1*++/++*;* 4-copy) mice in a *Npr1* gene-dose-dependent manner (Shi et al., 2003). Although, the blood volume expansion stimulated the release of ANP in all three *Npr1* genotypes of mice, significant functional responses (RPF, GFR, and sodium excretion) occurred only in *Npr1*−/<sup>−</sup> and *Npr1*++/++ mice but not in *Npr1*−/<sup>−</sup> mice. These findings demonstrated that the ANP/NPRA axis is primarily responsible for mediating the renal hemodynamic and sodium excretory responses to intravascular blood volume expansion. ANP responses to volume expansion led to the significantly lesser excretion of Na+ and water in 0-copy null mutant mice and significantly greater excretory responses along with reduced tubular reabsorption in 4-copy mice as compared with 2-copy wild-type mice. Similarly, during the volume expansion, urinary cGMP concentration was significantly lower in null mutant mice and greater in gene-duplicated mice. Our previous findings have established that NPRA is a hallmark receptor, which plays a critical role in mediating the natriuresis, diuresis, and renal hemodynamic responses to acute blood volume expansion (Shi et al., 2003).

Genetic mouse models with disruption of both *Nppa* and *Npr1* genes have provided strong support for the role of this hormone-receptor system in the regulation of blood pressure, cardiac hypertrophy, and other physiological functions (John et al., 1995; Lopez et al., 1995; Oliver et al., 1997, 1998; Melo et al., 1999; Pandey et al., 1999; Shi et al., 2001, 2003; Holtwick et al., 2002; Vellaichamy et al., 2005; Kishimoto et al., 2011; Pandey, 2011). Therefore, the genetic defects that reduce the activity of ANP and its receptor system can be considered as candidate contributors to essential hypertension and CHF (John et al., 1995; Pandey et al., 1999; Zhao et al., 1999; Knowles et al., 2001; Holtwick et al., 2002; Shi et al., 2003; Vellaichamy et al., 2005). Interestingly, complete absence of NPRA causes hypertension in mice and leads to altered renin and ANG II levels, cardiac hypertrophy, and lethal vascular events similar to those seen in untreated human hypertensive patients (Oliver et al., 1997; Shi et al., 2001, 2003; Zhao et al., 2007). In contrast, increased expression of *Npr1* reduces the blood pressures and inflammatory responses, protects heart, and increases the intracellular second messenger cGMP concentrations corresponding to the increasing number of *Npr1* gene copies (Oliver et al., 1998; Pandey et al., 1999; Shi et al., 2003; Vellaichamy et al., 2007, 2014; Zhao et al., 2013). Recent evidence also indicates that CNP and its receptor NPRB can play important role in regulating the cardiac hypertrophy and remodeling as a potential drug target for the treatment of cardiovascular diseases (Del Ry, 2013).

#### **CONCLUSION**

The field of NPs has been advanced to examine the function and signaling mechanisms of their receptors and the role of second messenger cGMP in physiology and pathophysiology of hypertension, renal hemodynamics, cardiovascular functions, and neural plasticity. The development of gene-knockout and gene-duplication mouse models along with transgenic mice have provided a framework for understanding both the physiological and pathophysiological functions of NPs and their receptors in the intact animals *in vivo*. Although, a considerable progress has been made, the transmembrane signal transduction mechanisms of NPs and their receptors remain unresolved. Future investigations should include; the identification and characterization of cellular targets of intracellular second messenger cGMP produced by NPs, including cytosolic and nuclear proteins, role in gene transcription, cell growth and proliferation, apoptosis, and differentiation. A more vigorous studies of the crosstalk with other signaling mechanisms namely, PKC,MAPKs, cAMP,Ca2+, and IP3 needs to be pursued systematically. NPs are considered as circulating markers of CHF, however, their therapeutic potential for the treatment of cardiovascular diseases such as hypertension, renal insufficiency, cardiac hypertrophy, CHF, and stroke is still lacking. The ultimate goal of the investigations is this field is to fully appreciate the mechanisms of cGMP generation after ligand binding to GC-coupled receptors and the pathways leading to elicit cellular and physiological functions in relation to other signaling molecules with special emphasis to Ca2+, IP3, and cAMP levels. Identification of the discrete switch points in signal transmission of NPs and their cognate receptors that specify unique directional responses need to be vigorously pursued.

#### **ACKNOWLEDGMENTS**

I thank my wife Kamala Pandey for her generous help in the preparation of this manuscript. The research in the author's laboratory is supported by the grants from the National Institutes of Health (HL 57531 and HL 62147).

#### **REFERENCES**


receptor-A gene: interactive roles of modified histones, HATS, p300, and Sp1. *J. Biol. Chem.* 289, 6991–7002. doi: 10.1074/jbc.M113.511444


of heart failure: an analysis from the breathing not properly multinational study. *Am. J. Kidney Dis.* 41, 571–579. doi: 10.1053/ajkd.2003.50118


*Biochem. Biophys. Res. Commun.* 271, 374–379. doi: 10.1006/bbrc.2000. 2627


clearance receptor Potential role in the inhibition of astrocyte proliferation. *J. Biol. Chem.* 271, 14156–14162. doi: 10.1074/jbc.271.24.14156


peptide receptor-A exhibit the cardiac expression of proinflammatory mediators in a gene-dose-dependent manner. *Endocrinology* 155, 1045–1056. doi: 10.1210/en.2013-1416

Vellaichamy, E., Khurana, M. L., Fink, J., and Pandey, K. N. (2005). Involvement of the NF-kappa B/matrix metalloproteinase pathway in cardiac fibrosis of mice lacking guanylyl cyclase/natriuretic peptide receptor A. *J. Biol. Chem.* 280, 19230– 19242. doi: 10.1074/jbc.M411373200

Vellaichamy, E., Zhao, D., Somanna, N., and Pandey, K. N. (2007). Genetic disruption of guanylyl cyclase/natriuretic peptide receptor-A upregulates ACE and AT1 receptor gene expression and signaling: role in cardiac hypertrophy. *Physiol. Genomics* 31, 193–202. doi: 10.1152/physiolgenomics.00079.2007


Wilson, E. M., and Chinkers, M. (1995). Identification of sequences mediating guanylyl cyclase dimerization. *Biochemistry* 34, 4696–4701. doi: 10.1021/bi00014a025

Wong, S. K., Ma, C. P., Foster, D. C., Chen, A. Y., and Garbers, D. L. (1995). The guanylyl cyclase-A receptor transduces an atrial natriuretic peptide/ATP activation signal in the absence of other proteins. *J. Biol. Chem.* 270, 30818–30822. doi: 10.1074/jbc.270.51.30818

Wormald, M. R., and Dwek, R. A. (1999). Glycoproteins glycan presentation and protein-fold stability. *Structure* 7, R155–R160. doi: 10.1016/S0969- 2126(99)80095-1

Wu, C. F., Bishopric, N. H., and Pratt, R. E. (1997). Atrial natriuretic peptide induces apoptosis in neonatal rat cardiac myocytes. *J. Biol. Chem.* 272, 14860–14866. doi: 10.1074/jbc.272.23.14860

Yoshihara, F., Tokudome, T., Kishimoto, I., Otani, K., Kuwabara, A., Horio, T., et al. (2014). Aggravated renal tubular damage and interstitial fibrosis in mice lacking guanylyl cyclase-A (GC-A), a receptor for atrial and B-type natriuretic peptides. *Clin. Exp. Nephrol.* doi: 10.1007/s10157-014-0982-1 [Epub ahead of print].

Yoshimura, M., Yasue, H., Morita, E., Sakaino, N., Jougasaki, M., Kurose, M., et al. (1991). Hemodynamic, renal, and hormonal responses to brain natriuretic peptide infusion in patients with congestive heart failure. *Circulation* 84, 1581– 1588. doi: 10.1161/01.CIR.84.4.1581


**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.

*Received: 10 April 2014; accepted: 05 August 2014; published online: 22 August 2014. Citation: Pandey KN (2014) Guanylyl cyclase/natriuretic peptide receptor-A signaling antagonizes phosphoinositide hydrolysis, Ca2+ release, and activation of protein kinase C. Front. Mol. Neurosci. 7:75. doi: 10.3389/fnmol.2014.00075*

*This article was submitted to the journal Frontiers in Molecular Neuroscience. Copyright © 2014 Pandey. 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.*

## Regulation of hippocampal synaptic plasticity thresholds and changes in exploratory and learning behavior in dominant negative NPR-B mutant rats

*Gleb Barmashenko1,2 , Jens Buttgereit 3,4 , Neil Herring4,5 , Michael Bader <sup>4</sup> , Cemil Özcelik3,4 †, Denise Manahan-Vaughan2 and Karl H. Braunewell 1\**

<sup>1</sup> Guest Group, In vitro-Electrophysiology Laboratory, Department of Neurophysiology, Medical Faculty, Ruhr University Bochum, Bochum, Germany

<sup>2</sup> Department of Neurophysiology, Medical Faculty, Ruhr University Bochum, Bochum, Germany

<sup>3</sup> Experimental and Clinical Research Center, Max Delbrück Center for Molecular Medicine, Charité Medical Faculty, Berlin, Germany

<sup>4</sup> Max Delbrück Center for Molecular Medicine, Berlin, Germany

<sup>5</sup> Department of Physiology, Anatomy and Genetics, Burdon Sanderson Cardiac Science Centre – BHF Centre of Research Excellence, University of Oxford, Oxford, UK

#### *Edited by:*

Clint Lawrence Makino, Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA

#### *Reviewed by:*

Yi-Ping Hsueh, Institute of Molecular Biology – Academia Sinica, Taiwan Teresa Duda, Salus University, USA

#### *\*Correspondence:*

Karl H. Braunewell, Guest Group, In vitro-Electrophysiology Laboratory, Department of Neurophysiology, Medical Faculty, Ruhr University Bochum, Universitätsstraße 150, Building MA 4/150, 44780 Bochum, Germany

e-mail: kbraunewell@gmail.com

#### *†Present address:*

Cemil Özcelik, KLINIKUM VEST GmbH, Dorstener Straße 151, 45657 Recklinghausen, Germany The second messenger cyclic GMP affects synaptic transmission and modulates synaptic plasticity and certain types of learning and memory processes.The impact of the natriuretic peptide receptor B (NPR-B) and its ligand C-type natriuretic peptide (CNP), one of several cGMP producing signaling systems, on hippocampal synaptic plasticity and learning is, however, less well understood. We have previously shown that the NPR-B ligand CNP increases the magnitude of long-term depression (LTD) in hippocampal area CA1, while reducing the induction of long-term potentiation (LTP). We have extended this line of research to show that bidirectional plasticity is affected in the opposite way in rats expressing a dominant-negative mutant of NPR-B (NSE-NPR-B-KC) lacking the intracellular guanylyl cyclase domain under control of a promoter for neuron-specific enolase. The brain cells of these transgenic rats express functional dimers of the NPR-B receptor containing the dominant-negative NPR-B-KC mutant, and therefore show decreased CNP-stimulated cGMP-production in brain membranes. The NPR-B transgenic rats display enhanced LTP but reduced LTD in hippocampal slices. When the frequency-dependence of synaptic modification to afferent stimulation in the range of 1–100 Hz was assessed in transgenic rats, the threshold for both, LTP and LTD induction, was shifted to lower frequencies. In parallel, NPR-B-KC rats exhibited an enhancement in exploratory and learning behavior. These results indicate that bidirectional plasticity and learning and memory mechanism are affected in transgenic rats expressing a dominant-negative mutant of NPR-B. Our data substantiate the hypothesis that NPR-B-dependent cGMP signaling has a modulatory role for synaptic information storage and learning.

**Keywords: cGMP, exploratory, hippocampus, LTP, LTD, metaplasticity, NPR-B, memory**

#### **INTRODUCTION**

The natriuretic peptides, ANP, BNP, and CNP (A-, B-, and C- type natriuretic peptide) and their receptors, the natriuretic peptide receptors (NPRs) are widely distributed in the central nervous system (CNS). They constitute a peptide hormone-receptor signaling system with a variety of potential roles in modulating physiological brain functions (Potter et al., 2009; Potter, 2011). Whereas ANP and BNP activate the type I transmembrane guanylyl cyclase receptor, natriuretic peptide receptor-A (NPR-A), CNP activates a related cyclase, the natriuretic peptide receptor-B (NPR-B), leading to the production of the second messenger cGMP (Potter et al., 2009). NPR-A and -B display several functional domains, including an extracellular ligand-binding domain, a transmembrane domain, and an intracellular domain, which consists of a kinase homology domain (KHD), a hinge region, and a guanylyl cyclase domain catalyzing the conversion of Mg-GTP into cGMP. A ligand-dependent receptor homodimerization is

essential for the enzymatic activity of the guanylyl cyclase domain (Potter, 2011).

Natriuretic peptide receptor-B leads to various physiological effects ranging from bone growth (Bartels et al., 2004; Potter et al., 2009; Potter, 2011) to CNS effects, such as axonal sprouting in dorsal root ganglion neurons (Schmidt et al., 2007; Potter et al., 2009), effects on synaptic plasticity (Decker et al., 2008, 2009, 2010), and anxiogenic effects in rats and humans (Jahn et al., 2001; Kellner et al., 2003). CNP and its receptor NPR-B are widely expressed in the rat brain including expression in the limbic system, particularly in the hippocampal regions CA1-3 (Langub et al., 1995; Herman et al., 1996). In the hippocampus NPR-B is expressed in pyramidal cells and GAD (65/67)-immunopositive interneurons. Sub-cellular expression of NPR-B can be demonstrated in neuronal process in primary hippocampal cell cultures (Brackmann et al., 2005). The second messenger cGMP and cGMP signaling-cascades are implicated

in the modulation of long term potentiation (LTP), long-term depression (LTD) as well as other forms of synaptic plasticity, such as short-term plasticity (Schuman and Madison, 1991; Arancio et al., 1996; Son et al., 1998). The effect of cGMP on synaptic plasticity has been attributed to the activity of soluble guanylyl cyclases (sGCs; Arancio et al., 1996). Moreover, downstream cGMP targets such as cyclic nucleotide-gated channels and cGMP-dependent kinases (cGK or protein kinase G, PKG) contribute to different forms of synaptic plasticity (Klyachko et al., 2001; Kleppisch et al., 2003; Liu et al., 2003). More recently, we investigated whether activation of another cGMP-signaling system, the membrane guanylyl cyclase NPR-B, via activation of by its ligand CNP has an effect on hippocampal synaptic plasticity (Decker et al., 2008, 2009, 2010). When LTD and LTP stimulation was applied in area CA1 at 1 and 5 Hz and 30– 100 Hz, respectively, CNP increased the magnitude of LTD while LTP induction was reduced. Thus, in the presence of CNP the threshold for LTP induction was shifted to higher stimulus frequencies (Decker et al., 2010). C-type natriuretic peptide (CNP) also decreased hippocampal network oscillations in adult rats *in vitro*, which are believed to be involved in storage of information and memory consolidation *in vivo* (Decker et al., 2009). In line with these results earlier work showed that direct application of CNP into the lateral brain ventricles affects the performance of rats in a passive avoidance learning paradigm (Telegdy et al., 1999, 2000).

Knock out of the gene for NPR-B in mice causes severe dwarfism, and the NPR-B knock out mice display seizures, female sterility, and priapism and are therefore not suitable for detailed cardiovascular or neurophysiological phenotyping (Tamura et al., 2004). To overcome this limitation, transgenic rats with ubiquitous overexpression of a dominant-negative NPR-B mutant, lacking the cytoplasmic domain (CMV-NPR-B-KC) were generated (Langenickel et al., 2006). Receptor homodimerization of NPR-B, leading to a tight contact between the guanylyl cyclase domains, is essential for enzymatic activity. Interestingly, Tamura and Garbers (2003) have described that different NPR-B isoforms resulting from alternative splicing of the primary transcript are present in mouse tissues. In addition to full-length GC-B1, GC-B2 contains a 25 aa deletion in the KHD, and GC-B3 only retains a part of the extracellular ligand-binding domain. When GC-B2 or GC-B3 is expressed coincident with GC-B1, they act as dominant negative isoforms by virtue of blocking formation of active GC-B1 homodimers (Tamura and Garbers, 2003). It appears that these splice variants serve as dominant negative regulators of full-length GC-B. Similarly, overexpression of NPR-B-KC leads to functional down-regulation of NPR-B signaling as indicated by blunted CNP-induced cGMP production (Langenickel et al., 2006).

In order to test whether cGMP signaling via neuronal NPR-B affects bidirectional synaptic plasticity in the rat hippocampus, and simultaneously learning and memory, we have generated dominant-negative NPR-B mutant under the control of a neuronspecific promoter (NSE-NPR-B-KC) to specifically inhibit NPR-B signaling in the brain. After confirming the expression of the NPR-B-KC mutant and reduction of CNP-dependent cGMP generation in brain membranes we have analyzed bidirectional plasticity *in vitro* over the range of 1–100 Hz stimulation in the

hippocampal CA1 region. To further evaluate these animals in behavioral experiments we have performed open field, novel object recognition and spatial object recognition (SOR) tests. Our results support the hypothesis that the cGMP-signaling cascade linked to NPR-B and its ligand CNP play an important role in the brain, and in the hippocampus in particular, to modulate bidirectional synaptic plasticity and learning behavior.

#### **MATERIALS AND METHODS**

#### **GENERATION OF TRANSGENIC RATS WITH SPECIFIC INHIBITION OF NPR-B SIGNALING IN NEURONS**

Transgenic rats overexpressing a dominant-negative NPR-B mutant under the control of a neuron-specific promoter (NSE-NPR-B-KC; **Figure 1A**) were generated and identified by Southern blotting (**Figure 1B**) essentially as previously described (Langenickel et al., 2006). After breeding rats to homozygosity, tissue-specific expression of NSE-NPR-B-KC in the brain was verified by reverse transcription (RT)-PCR and Western blotting. RT-PCR, Western blotting and measuring of cGMP levels in brain membrane preparation was performed as previously described (Langenickel et al., 2006; Buttgereit et al., 2010). Primary anti-flag antibodies (Life Technologies, Carlsbad, CA, USA) in combination with secondary anti-rabbit-HRP antibodies were used for detection of NSE-NPR-B-KC-flag in Western blots. For RT-PCR the following primer pairs were used: NSE-NPR-B-KC forward: GACAGAGAGACTGATTTCGTCC, reverse: TCACTTGTCGTCATCGTCTTTG: GAPDH: forward: CCATGGAGAAGGCTGGGG, reverse: CAAAGTTGTCATGGAT-GACC. Transgenic rats and control Sprague Dawley rats, aged 6–8 weeks were kept under standard conditions with 12-h light– dark cycle and free access to water and food. The present study was carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) for care of laboratory animals and after approval of the local government ethics committee. All efforts were made to minimize the number of animals used.

#### **SLICE PREPARATION**

Animals were decapitated under deep ether anesthesia. The brain was rapidly taken out of the skull and immersed in ice-cold (1– 4◦C) sucrose-based solution (87 mM NaCl, 25 mM NaHCO3, 10 mM D-glucose, 75 mM sucrose, 2.5 mM KCl, 1.25 mM NaH2PO4, 0.5 mM CaCl2, 7 mM MgCl2, bubbled with 95% O2 and 5% CO2, pH 7.4). Horizontal hippocampal slices (400 μm) were cut on a vibratome (Leica VT 1200S). Slices were kept at room temperature. For recordings, the slices were transferred to a submerged recording chamber. During the experiments, the chamber was perfused at a flow rate of 3 ml/min with oxygenated artificial cerebrospinal fluid (ACSF; 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 25 mM NaHCO3, 25 mM D-glucose, 2 mM CaCl2, 1 mM MgCl2, pH 7.4). All recordings were performed at 32◦C.

#### **RECORDING OF EXTRACELLULAR FIELD POTENTIALS**

Recording electrodes were pulled from borosilicate glass and filled with standard Artificial Cerebrospinal Fluid (ACSF, resistance 1– 3 M). Extracellular field potentials (FPs) were recorded from

stratum radiatum (SR) of area CA1 every 60 s for at least 10 min (baseline recording) and then for at least 60 min following highor low frequency stimulation (HFS or LFS). A concentric bipolar stimulation electrode was placed in SR of area CA1 to stimulate Schaffer collaterals and commissural fibers. FPs were recorded using a MultiClamp 700B amplifier, filtered at 3 kHz and sampled to a computer disk at 10 kHz using the Clampex 10.2 software (Molecular Devices, Sunnyvale, CA, USA). The stimulus intensity was set to evoke 30% of the maximum FP response in LTP, and 50% in LTD experiments (50–200 μA with stimulus duration of 100 μs). The HFS paradigm applied to SR consisted of a 1 s long stimulus train applied at frequencies of 100, 50, 30, or 10 Hz (pulse duration: 100 μs, each). For LFS trains 900 pulses were applied at 1 or 5 Hz (pulse duration: 100 μs, each; 4–6). FP recordings obtained from SR were analyzed by determining the slope of the ?eld excitatory postsynaptic potential (EPSP) between 10 and 90% of the peak amplitude. Summary graphs were prepared by normalizing all responses to the baseline obtained during 10 min

prior to HFS or LFS and then averaged across experiments. Only slices with a stable baseline FP response were used for further analysis. All changes in long-term synaptic plasticity were evaluated by averaging responses at 50–60 min post-HFS or LFS and averaging across experiments. For metaplasticity experiments, the frequencies used were HFS 100 Hz for 1 s and LFS 1 Hz for 900 times for 15 min. After the first HFS or LFS for 15 min, LTP was measured for 60 min, then a second HFS or LFS for 15 min was applied, and LTP was measured for another 60 min. CNP (Calbiochem, Merck KGaA, Darmstadt, Germany) was prepared as stock solution, frozen, finally thawed on the day of the experiment and added to ACSF to reach the desired final concentration (100 nM).

#### **BEHAVIORAL TESTS**

The temporal order to test the 11 wild type and the 12 transgenic rats were chosen randomly and changed for each experiment. All trials were recorded by a video camera and exploration time was measured by video analysis. The rats' behavior was determined as exploration time when the animal approached closer than 2 cm to an object with its snout or forepaws (or both). After each trial, the arena and objects were cleaned carefully to avoid the presence of olfactory cues for the next rat. The available objects differed slightly in their shape, size and color. Before the first test – the novel object recognition task – rats were allowed to explore the arena during a habituation trial for 5 min, so that they could adapt to the new environment.

The object recognition task (ORT) was conducted in a squareshaped arena (100 cm × 100 cm), and used a similar approach as described in a previous reportGoh andManahan-Vaughan (2013). In the training trial (1st trial), two novel objects (i.e., A and B) were presented. Fifty minutes later (second trial) one familiar and one novel object (i.e., A and C) were presented. A further 24 h later (third trial) object A was presented with a new object (D). The objects were presented in the same location as the objects in the first trial. The objects and the recording chambers were cleaned thoroughly between task trials to ensure the absence of olfactory cues. The objects were distinctly different from one another and heavy so that they could not be moved by the rats.

In the SOR task the objects always remained the same (A and B), and the task was performed as previously described Goh and Manahan-Vaughan (2013). The position of object A remained constant, but in the second and third trial object B was placed in a distinctly new position in the chamber. As in the ORT experiments, the animals were allowed to explore the objects for 5 min. The second and third trials began 50 min and 24 h after the first trials, respectively.

In the open field locomotion test we examined motor function by means of measuring spontaneous activity in an open field (100 cm × 100 cm). Here, we differentiated between the entire running time of the animals and the time spent in the center of the area (without touching the walls).

#### **DATA ANALYSIS**

Data were expressed as mean ± SEM. Statistical analysis was done with the aid of SigmaStat software (SPSS Inc., Chicago, IL, USA) performing Student's *t*-test (unpaired), Mann–Whitney *U*-test or one-way analysis of variance (ANOVA) with repeated measures. The sample sizes (*n* values) for electrophysiological experiments indicate the number of slices from at least three different animals. Differences were considered significant with ∗*p* < 0.05 or ∗∗*p* < 0.01.

## **RESULTS**

#### **NPR-B-DEPENDENT cGMP SIGNALING IN NSE-NPR-B**-**KC RATS**

Transgenic rats overexpressing a dominant-negative NPR-B mutant under the control of a neuron-specific promoter (NSE-NPR-B-KC; **Figure 1A**) were generated and identified by Southern blotting (**Figure 1B**). After breeding rats to homozygosity, tissue-specific expression of NSE-NPR-B-KC in the brain was verified by RT-PCR. Other tissues, such as heart, lung, aorta and kidney do not express the transgene (**Figure 1C**). Western blot analysis showed that the NPR-B deletion construct was properly translated into a protein with the expected Mw of about 45 kDa in the transgenic animals but not in wild type rats (**Figure 1D**). To test

whether expression of the dominant negative NSE-NPR-B-KC mutant leads to the functional down-regulation of the NPR-B signaling, guanylyl cyclase assays in brain membrane preparations were performed (**Figures 1E,F**). The cGMP production of brain membranes derived from NSE-NPR-B-KC rats stimulated with 10, 100, and 1000 nM ANP was comparable with those derived from wild-type rats, indicating that signaling via NPR-A was not affected in the transgene line (**Figure 1E**). In contrast, after stimulation with the same doses of CNP cGMP production significantly dampened in brain membrane preparations from NSE-NPR-B-KC rats compared to wild-type rats. This result confirms the receptor-specific dominant-negative effect of the transgene expression on NPR-B signaling in the brain (**Figure 1F**).

#### **EFFECT OF NPR-B-DEPENDENT cGMP SIGNALING ON SYNAPTIC PLASTICITY IN HIPPOCAMPAL SLICES** *IN VITRO*

Here, we investigated the role of NPR-B and cGMP signaling in hippocampal synaptic plasticity. Field extracellular postsynaptic potentials (EPSP) were recorded from the SR of the CA1 by stimulating the Schaffer collaterals using high-frequency stimulation (HFS) at 100 Hz, which typically produces potentiation of the EPSP in the form of LTP. HFS resulted in LTP of 157 ± 12% 1 h after HFS was applied (*n* = 13) compared to baseline in wild-type (wt) rats (**Figure 2A**). In NPR-B-KC rats, showing reduced CNPdependent cGMP levels in brain membranes, HFS resulted in LTP that in its induction phase was similar to that seen in wt slices but after 60 min reached 285 ± 17% of baseline levels (*n* = 16, *p* < 0.01). On the opposite, application of 100 nM CNP to activate hippocampal NPR-B significantly inhibited LTP in wt rats (115 ± 10%, *n* = 7). Since the EPSP recordings in NPR-B-KC rats did not reach saturation levels 60 min after HFS (**Figure 2A**), a longer time period of 200 min was recorded (**Figure 2B**). EPSP recordings showed a steep rise over the first 30 min, followed by a slower rise over the next 100 min and finally reached saturation levels at ∼250% of baseline values. This potentiation subsequently decreased somewhat to ∼200% of baseline values 180 min after HFS (**Figure 2B**).

#### **THE FREQUENCY-DEPENDENCE OF LTP AND LTD IS MODULATED BY NPR-B ACTIVATION THROUGH CNP AND FUNCTIONAL DOWN-REGULATION OF NPR-B SIGNALING**

Next, we compared the frequency-dependence of synaptic plasticity in wt, homo- and heterozygous NPR-B-KC animals. The extent of change of synaptic strength was assessed 60 min after application of afferent stimulation of Schaffer collateral-CA1 synapses (**Figure 3**). In wt animals LTD was elicited at frequencies of 1, 5, and 10 Hz stimulation. Stimulation at 30, 50, and 100 Hz led to potentiation of EPSP over baseline. In hetero- and homozygous NPR-B-KC animals the frequency-dependence of synaptic plasticity was distinctly different to wt controls at frequencies of 5, 10, 30, 50, and 100 Hz (*p* < 0.01, *p* = 0.36 for 1 Hz stimulation, **Figure 3**). Thus, depending on gene dose the NPR-B-KC mutation shifts LTP induction to lower stimulation frequencies compared to wt rats.

We then tested the effect of CNP application (100 nM) on synaptic plasticity in wt slices (**Figure 3**). Here, we observed

**FIGURE 2 | (A)** Time course of the relative changes in the excitatory postsynaptic potential (EPSP) slope expressed in % of baseline stimulation induced by 100 Hz HFS for 1 s in hippocampal slices from NPR-B-KC, from wild type rats (wt) and from wild type slices incubated with the NPR-B agonist CNP (100 nM). Sample traces of the electrically evoked EPSPs for the NPR-B-KC, wild type slices, and wild type slices incubated with 100 nM CNP are shown on the right, and were recorded 10 min before and 60 min following high frequency stimulation (HFS, 100 Hz for 1 s). Note that LTP was more strongly expressed in slices from NPR-B-KC rats (black circles) compared to wt (white circles). Bath application of 100 nM CNP (black triangles) caused attenuation of LTP. **(B)** The increase of the EPSP slope in NPR-B-KC rats reaches saturation levels at 285% of baseline stimulation after 120 min. EPSPs decrease slowly to ∼200% of baseline stimulation after 180 min.

that although LTD which was elicited using 1 Hz stimulation was equivalent to that seen in untreated wt controls, LTD induced by 5 Hz stimulation was significantly enhanced, and potent LTD was elicited by 10 Hz stimulation. In contrast, CNP-treatment resulted in a failure to induce LTP using 30 Hz stimulation. Stimulation at 50 and 100 Hz resulted in only a very small potentiation that was significantly less than that seen in untreated wt controls (*p* < 0.01). Taken together, these data suggest that under conditions of low cGMP levels the threshold for induction of LTP is lowered, whereas when cGMP is strongly elevated via the CNP/NPR-B/cGMP pathway the threshold for LTP induction is raised and LTD induction is facilitated.

## **POSSIBLE MECHANISMS OF THE EFFECT OF FUNCTIONAL DOWN-REGULATION OF NPR-B SIGNALING ON LTP AND LTD IN HIPPOCAMPAL SLICES** *IN VITRO*

and 100 Hz.

**agonist CNP (100 nM).** Stimulation frequencies used were 1, 5, 10, 30, 50,

In search of possible molecular and cellular mechanisms responsible for the observed steady increase in EPSPs over a 60 min period following HFS, we repeated the experiments in the presence of anisomycin (20 μM), an inhibitor of protein synthesis (**Figure 4**). Anisomycin did not influence the induction of LTP but slightly reduced the potentiation of the EPSP signals in slices from wt rats, and reduced potentiation back to control levels in slices of NPR-B-KC rats. These results indicate that protein synthesis is involved in the continuous increase of EPSP over a 60 min period in slices of NPR-B-KC rats. However, the induction of LTP was unaffected by anisomycin in control and NPR-B-KC rats.

We then tested whether NPR-B-KC affects the intrinsic excitability in hippocampal neurons. Here we measured the EPSP in relation to the stimulation current (**Figure 5A**). We observed significantly higher responses in slices of NPR-B-KC rats compared to wt using stimulation currents of 0.5–2 mA (*p* < 0.05). This increase in the input–output relationship indicates positive plastic changes in intrinsic excitability caused by reduced cGMP signaling that may underlie the observed lowering of LTP induction thresholds in NPR-B-KC rats compared to wt rats.

Moreover, to test whether hippocampal inhibitory activity is affected in NPR-B-KC rats we measured paired-pulse facilitation. Using different interpulse intervals (**Figure 5B**) we show larger facilitation in slices of NPR-B-KC rats, particularly at the interpulse interval of 20 and 25 ms (*p* < 0.01), which is typically indicative of a loss of inhibitory GABAergic activity.

#### **THE EFFECT OF FUNCTIONAL DOWN-REGULATION OF NPR-B SIGNALING ON METAPLASTICITY IN HIPPOCAMPAL SLICES** *IN VITRO*

Long-term potentiation and LTD depend on the current state of synapses as modulated by intrinsic but also extrinsic influences (Goh and Manahan-Vaughan, 2013). In other words, synaptic plasticity can be modified by activity including previous synaptic inhibition, the activity of modulatory afferents, and hormonal activity, for instance, of natriuretic peptide hormones. This plasticity of synaptic plasticity, also called metaplasticity, may play a role in some mechanisms of memory and learning (Dudek and Bear, 1993). Thus, we assessed effects of NPR-B-KC on metaplasticity using repetitive HFS (100 Hz) or LFS (1 Hz) 60 min after an initial HFS (100 Hz) or LFS (1 Hz; **Figure 6**). Pretreatment with HFS did not increase LTP further following the second HFS in wild type rats, but increased LTP in NPR-B-KC rats. Pretreatment with HFS abolished LTD in response to LFS in wt rats, but further increased LTP in NPR-B-KC rats. Similarly, LFS pretreatment further increased LTP following the second HFS in NPR-B-KC rats, but had no effect in the LFS/LFS combination. Thus, in all combinations of LFS and HFS a significant bigger potentiation was observed after the second stimulation (% of potentiation 60 min after first stimulation) in NPR-B-KC rats compared to wt rats (*p* < 0.05). These results indicate that the effect of the NPR-B/cGMP signaling system on LTP and LTD also affects metaplasticity.

#### **MEASURING ANXIETY/CURIOSITY AND LEARNING AND MEMORY IN TRANSGENIC RATS EXPRESSING DOMINANT-NEGATIVE NPR-B***-***KC**

To evaluate changes in learning behavior of NPR-B-KC animals in behavioral experiments, and relate these results with effects of NPR-B-KC on synaptic plasticity, we performed open field, novel object recognition and SOR tests. In the open field test, anxiety and curiosity were compared between NPR-B-KC and

**FIGURE 5 | (A)** EPSP slope as a function of stimulus current in slices of NPR-B-KC (functional down-regulation of NPR-B signaling) and wild type (wt) rats. EPSP slopes recorded as mV/ms were determined between 0.5 and 2 mV. Note the greater excitability in hippocampal slices of NPR-B-KC rats with significant differences at 0.5–2 mV stimulus current. **(B)** Pairedpulse facilitation as a function of interpulse interval current in slices of NPR-B-KC (NPR-B) and wild type (wt) rats. Paired pulses as determined by the ratio of EPSP slope 2/EPSP slope 1 were recorded at 20, 25, 40, 50, 100, and 300 ms time intervals. Note the larger facilitation in slices of NPR-B-KC rats at interpulse intervals of 20 and 25 ms. Asterisks indicate significant changes (\*p < 0.05 and \*\*p < 0.01).

control rats. We observed significant differences in locomotion of NPR-B-KC rats in the first trial in the arena (habituation; **Figure 7A**, *p* < 0.01). The greater willingness of NPR-B-KC rats to run can be understood as a sign of greater curiosity and reduced anxiety. However, in the second and third trial after 5 min and 1 h, respectively, the rats from both groups did not show significant differences in the total distance moved. On the other hand, the NPR-B-KC rats moved significantly longer distances in the central area of the arena compared to the total distance traveled (**Figure 7B**, *p* < 0.05). NPR-B-KC rats also made significantly more visits to the central area (**Figure 7B**, *p* < 0.01). These results indicate that NPR-B-KC rats

show reduced levels of anxiety leading to enhanced exploratory behavior.

In the novel object recognition test (ORT), upon exposure to the familiar object (A) and the unfamiliar object (C at 50 min, D at 24 h), one would expect that the animals show a markedly greater interest in the unfamiliar objects compared to object A (22). This was the case both for wt and transgenic animals (**Figure 8A**). The analysis also revealed a significantly higher exploratory activity for NPR-B-KC rats in the initial 5 min when objects (A and B) were introduced into the arena (**Figure 8A**, *p* = 0.065). Object discrimination ratios were calculated to verify that the animals showed a preference of the novel vs. the familiar object (**Figure 8B**). No significantly different discrimination ratios were recorded in the first trial suggesting that the animals found both novel objects equally interesting. Upon exposure to object A and novel object C in the second trial, a significant preference for object C over object A was seen in both animal groups. A preference for the novel object was still evident 24 h after first object exposure, whereby the transgenic animals showed a significantly greater preference for object D compared to wt controls (*p* < 0.05). This indicates that long term memory is facilitated in NPR-B-KC rats. In the SOR test, the two groups exhibited the expected learning performance, but no significant differences between the two groups were observed (**Figure 9A**). In the first trial AB1, the second trial after 50 min. and the third trial after 24 h NPR-B-KC rats did not exhibited a significantly changed discrimination ratio (**Figure 9B**). Although NPR-B-KC rats showed a tendency for an increased discrimination ratio AB3 in the third trial (**Figure 9B**), SOR was not significantly different between NPR-B-KC rats and wt animals.

#### **DISCUSSION**

The second messenger cGMP modulates synaptic transmission and learning processes. This has been attributed mainly to calcium- and nitric oxide (NO)-dependent activation of sGCs (Schuman and Madison, 1991; Arancio et al., 1996; Son et al., 1998). However, other cGMP-producing enzymes including the transmembrane guanylyl cyclase NPR-B are expressed in the brain, including the hippocampal formation. NPR-B is activated by the CNP and produces cGMP in hippocampal neurons (Brackmann et al., 2005). CNP and its receptor NPR-B are strongly expressed in hippocampal regions CA1-3 (Langub et al., 1995; Herman et al., 1996), which are well-described model systems for studying

animals. Asterisks indicate significant changes (\*p < 0.05 and \*\*p < 0.01).

synaptic plasticity and metaplasticity and their relation to learning and memory (Dudek and Bear, 1993; Manahan-Vaughan and Braunewell, 1999; Braunewell and Manahan-Vaughan, 2001). Evidence that the CNP/NPR-B system is able to modulate synaptic transmission came from our recent study showing that application of CNP, the ligand for NPR-B, negatively affects LTP, but positively affects LTD in hippocampal slices (Decker et al., 2008, 2009, 2010). Our present results indicate that transgenic rats expressing a dominant-negative mutant (NPR-B-KC) of NPR-B in the brain, display changes of bidirectional plasticity in the opposite way. Rats expressing the NPR-B-KC mutant, which specifically inhibits cGMP production of NPR-B without affecting

NPR-A, showed enhanced LTP and reduced LTD induction. When frequency-dependence of synaptic modification in the range of 1–100 Hz was assessed transgenic rats expressed LTP at lower stimulation frequencies than wild-type controls and at the same time exhibited enhancements in exploratory and learning behavior. These results indicate that modulation of bidirectional plasticity via NPR-B-dependent cGMP signaling is related to learning and memory behavior in rats, and are in line with the notion that NPR-B and the associated cGMP signaling pathway constitutes a negative feed forward or feedback loop regulating synaptic plasticity.

calculation reveals no significant differences in the interest for the object in

a novel place after 50 min (B2) and 24 h (B3) between NPR-B-

It is worth mentioning that exploratory activity and spatial learning are associated with LTP and LTD. In rats, exploratory behavior of a new environment containing unfamiliar objects

type rats (wt).

KC and wild

and/or familiar objects in new locations facilitated LTD and impaired LTP, whereas exploration of the new environment itself, in the absence of objects, impaired LTD and facilitated LTP (Manahan-Vaughan and Braunewell, 1999; Braunewell and Manahan-Vaughan, 2001; Kemp and Manahan-Vaughan, 2004, 2007, 2008). More recently, it was shown in mice that object recognition facilitates hippocampal LTD and impairs LTP, indicating that both forms of plasticity work in concert in the creation of spatial memory. However, LTP and LTD might contribute to different components of a spatial representation and learning (Goh and Manahan-Vaughan, 2013). In NPR-B-KC mutant rats, showing enhanced LTP and reduced LTD, we found improved object recognition. Although our experiments did not investigate the exact contributions of LTP and LTD to SOR, these results further substantiate the link between LTP and LTD as cellular mechanism required for spatial memory tasks (Goh and Manahan-Vaughan, 2013), and link cGMP signaling via NPR-B with synaptic plasticity and learning.

In regard to the possible molecular and cellular mechanisms of how NPR-B affects synaptic plasticity in the hippocampal network, our results indicate that protein synthesis is involved in the observed effect on EPSPs. The late onset but continuous increase of EPSPs over a prolonged time period in NPR-B-KC mutant rats was sensitive to the protein synthesis inhibitor anisomycin, indicating that the link between reduced cGMP production and enhanced LTP involves protein synthesis. Moreover, the increase in the input-output relationship of LTP indicates positive alterations in the intrinsic excitability of neurons probably leading to the shift in LTP induction frequencies. This is most likely a sign of loss of inhibitory GABAergic activity in the slices of NPR-B-KC mutant rats. This assumption is in agreement with our previous results showing that CNP activation of NPR-B in hippocampal slices leads to inhibition of LTP by modulating GABA(A)-mediated inhibition of the hippocampal network (Decker et al., 2008, 2009, 2010). Similarly, the effect of the inhibition of NPR-B-dependent cyclase activity on metaplasticity also indicates that there is a loss of negative feedback in the hippocampus of transgenic rats. In our experiments all combinations of LFS and HFS, with the exception of the LFS/LFS combination, led to a significant increase in LTP after the second stimulation in NPR-B-KC rats. These differences in metaplasticity are most likely explained by the fact that 1 Hz LFS did not induce LTD in the transgenic rats, and that LTP may not have been saturated at 1 h and typically reached much higher levels than in wt animals. These differences likely contribute to the observed changes in metaplasticity in transgenic animals. Thus, the NPR-B/cGMP signaling system may play a role in metaplasticity, for instance as one of the regulatory feedback mechanism, to link previous experience or the hormonal state of the organism to synaptic plasticity.

In our previous studies we have postulated that the physiological significance of the CNP/NPR-B signaling system in the hippocampus may be to restrict synaptic plasticity and possibly learning and memory behaviors in stress-linked situations and thus link anxiogenic and limbic functions. This assumption was based on preclinical studies in rats and human studies which have shown that natriuretic peptides differentially modulate

endocrine and behavioral stress responses (Kellner et al., 2003). CNP enhances the release of ACTH and cortisol (Kellner et al., 2003) and had anxiogenic effects in rats and in humans (Jahn et al., 2001). We have previously shown that exposure of rat strains to stress, in form of a novel open platform environment, induced a significant elevation in serum corticosterone levels but did not facilitate LTD expression (Manahan-Vaughan and Braunewell, 1999). However, different or longer lasting forms of stress activating the NPR-B system are likely to modulate bidirectional plasticity and thus lead to reduced learning and memory-related behavior. Interestingly, in the transgenic rats showing reduction of the NPR-B and cGMP signaling system we observe reduced anxiety levels which are observed as an enhanced exploratory behavior in the open field test as well as in object recognition tests. Enhanced exploratory and novelty seeking behavior is likely linked to learning and memory behavior (Manahan-Vaughan and Braunewell, 1999; Braunewell and Manahan-Vaughan, 2001). Although we have seen significant differences in learning behavior only in novel object recognition and not in the SOR test, enhanced exploratory behavior was noticed in open field and object recognition tests alike. More detailed behavioral analysis will be necessary to further characterize the role of NPR-B for distinct learning and memory paradigms.

Interestingly, GC-C another member of the family of transmembrane guanylyl cyclases (Potter et al., 2009; Potter, 2011), previously thought to be expressed mainly in neurons of the intestine, was identified in midbrain dopaminergic neurons, particularly in midbrain ventral tegmental area and substantia nigra compacta (VTA/SNc) neurons that regulate many important behavioral processes (Gong et al., 2011). Dysfunction of these neurons is associated with attention deficit hyperactivity disorder (ADHD) and schizophrenia. Mice in which GC-C has been knocked out exhibit hyperactivity and attention deficits which resemble behavioral symptoms of ADHD (Gong et al., 2011). Similarly, the functional down regulation of NPR-B in NPR-B-KC rats showed increased activity and exploratory behavior. Thus, dysfunction of receptor guanylyl cyclases and the cGMP-pathway may be involved in a variety of additional CNS disorders, for instance deficits in cGMP signaling pathways and neuroplasticity are hypothesized to underlie the pathophysiology of major depressive disorder (MDD; Reierson et al., 2011). This is in line with the hypothesis that increased cGMP signaling in the brain restricts excitability of neurons, and thereby controls a range of behaviors including hyperactivity, anxiety, attention, learning and memory.

#### **ACKNOWLEDGMENTS**

This work was conducted within the framework of the national priority program (Schwerpunktprogram) SPP1226 Nicotine: molecular and physiological mechanisms in the central nervous system (www.nicotine-research.com) funded by the German Research Foundation (Deutsche Forschungsgemeinschaft, grant DFG 1579/8-1 and 2 to Karl H. Braunewell, MA1843/6 to Denise Manahan-Vaughan). Generation of transgenic rats was supported by DFG grants to Michael Bader and Jens Buttgereit (BA1374/14-1).

#### **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: 16 September 2014; accepted: 12 November 2014; published online: 01 December 2014.*

*Citation: Barmashenko G, Buttgereit J, Herring N, Bader M, Özcelik C, Manahan-Vaughan D and Braunewell KH (2014) Regulation of hippocampal synaptic plasticity thresholds and changes in exploratory and learning behavior in dominant negative NPR-B mutant rats. Front. Mol. Neurosci. 7:95. doi: 10.3389/fnmol.2014.00095 This article was submitted to the journal Frontiers in Molecular Neuroscience.*

*Copyright © 2014 Barmashenko, Buttgereit, Herring, Bader, Özcelik, Manahan-Vaughan and Braunewell. 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.*

## Dysfunction of outer segment guanylate cyclase caused by retinal disease related mutations

## *Patrick Zägel <sup>1</sup> and Karl-Wilhelm Koch1,2 \**

<sup>1</sup> Biochemistry Group, Department of Neurosciences, Carl von Ossietzky University Oldenburg, Oldenburg, Germany <sup>2</sup> Research Center Neurosensory Science, Carl von Ossietzky University Oldenburg, Oldenburg, Germany

#### *Edited by:*

Rameshwar K. Sharma, Salus University, USA

#### *Reviewed by:*

Rameshwar K. Sharma, Salus University, USA Teresa Duda, Salus University, USA

#### *\*Correspondence:*

Karl-Wilhelm Koch, Biochemistry Group, Department of Neurosciences, Carl von Ossietzky University Oldenburg, Oldenburg D-26111, Germany e-mail: karl.w.koch@uni-oldenburg.de

Membrane bound guanylate cyclases are expressed in rod and cone cells of the vertebrate retina and mutations in several domains of rod outer segment guanylate cyclase 1 (ROS-GC1 encoded by the gene GUCY2D) correlate with different forms of retinal degenerations. In the present work we investigated the biochemical consequences of three point mutations, one is located in position P575L in the juxtamembrane domain close to the kinase homology domain and two are located in the cyclase catalytic domain at H1019P and P1069R. These mutations correlate with various retinal diseases like autosomal dominant progressive cone degeneration, e.g., Leber Congenital Amaurosis and a juvenile form of retinitis pigmentosa. Wildtype and mutant forms of ROS-GC1 were heterologously expressed in HEK cells, their cellular distribution was investigated and activity profiles in the presence and absence of guanylate cyclase-activating proteins were measured. The mutant P575L was active under all tested conditions, but it displayed a twofold shift in the Ca2+-sensitivity, whereas the mutant P1069R remained inactive despite normal expression levels. The mutation H1019P caused the cyclase to become more labile. The different biochemical consequences of these mutations seem to reflect the different clinical symptoms. The mutation P575L induces a dysregulation of the Ca2+-sensitive cyclase activation profile causing a slow progression of the disease by the distortion of the Ca2+-cGMP homeostasis. In contrast, a strong reduction in cGMP synthesis due to an inactive or structurally unstable ROS-GC1 would trigger more severe forms of retinal diseases.

**Keywords: photoreceptor guanylate cyclase, GCAP, retinal dystrophy, neuronal calcium sensor, phototransduction**

#### **INTRODUCTION**

Inherited retinal diseases are caused by a heterogeneous group of mutations in retinal genes (Perrault et al., 2000; den Hollander et al., 2008; Athanasiou et al., 2013). Among them are a large number of genes coding for proteins involved in phototransduction processes in the outer segments of rods and cones. Genetic heterogeneity is paralleled by diverse phenotypical and clinical characteristics of retinal degenerations, which are classified in disorders like cone-rod dystrophies, retinitis pigmentosa, and leber congenital amaurosis (LCA). These disease forms lead to a loss of visual function, but they differ in the progression of specific visual impairments.

Phototransduction in rods and cones is centered on the lighttriggered hydrolysis of the second messenger cyclic GMP, its powerful re-synthesis after illumination and its role as a ligand targeting a cyclic nucleotide-gated channel in the plasma membrane (Pugh and Lamb, 2000; Kaupp and Seifert, 2002; Luo et al., 2008). Phototransduction is further under control of several negative feedback loops that are important for recovery of the photoresponse from illumination and for adjusting the cell performance to ambient background light (Fain et al., 2001). One crucial molecular factor involved in the regulation of phototransduction is Ca2+, which binds to and dissociates from Ca2+-sensor proteins under oscillating changes in cytoplasmic Ca2+-concentration [Ca2+].

A multi-protein complex operating at the interface between changing concentrations of cGMP and [Ca2+] consists of a sensory type membrane-bound guanylate cyclase that is regulated by guanylate cyclase-activating proteins (GCAPs; Stephen et al., 2008; Dizhoor et al., 2010; Koch et al., 2010; Koch and Dell'Orco, 2013). So far, mutations in patients suffering from retinal diseases are found in the gene *GUCA1A* encoding the Ca2+-sensor GCAP1 and *GUCY2D* coding for the outer segment guanylate cyclase type 1 (ROS-GC1) (Kitiratschky et al., 2008; Behnen et al., 2010; Hunt et al., 2010). Previous research was focused on elucidating the molecular causes and cellular consequences of point mutations in ROS-GC1 and GCAP1 and in a further step to test therapeutic strategies involving RNA interference techniques and transgenic mouse lines (Buch et al., 2011; Jiang et al., 2011; Boye et al., 2013).

However, disease causing mutations can also point to critical amino acid positions that determine structure-function relationships in natively folded protein structures and therefore might help to gain mechanistic insights of protein function and regulation (Duda et al., 1999a, 2000; Ramamurthy et al., 2001). For example, a biochemical study on mutations in the so-called dimerization domain of ROS-GC1 correlating with cone-rod dystrophy recently revealed that this dimerization or linker domain operates as a Ca2+-sensitive control switch module (Zägel et al., 2013). In the present work we tested the biochemical consequences of three point mutations (**Figure 1**). The mutation P575L was found in a family, where clinically affected members suffer from progressive cone degeneration (Small et al., 2008). The amino acid substitution is located in the juxtamembrane domain (JMD), which is a specific region of sensory GCs upstream of the kinase homology domain (KHD). H1019P is located in the cyclase catalytic domain (CCD) and patients with this missense mutation show the typical clinical symptoms of LCA type 1 suffering from severe visual impairment at birth (Perrault et al., 2000). Finally, P1069R is also within the CCD and a patient carrying this mutation was diagnosed with juvenile retinitis pigmentosa (Booji et al., 2005).

So far, none of these mutations have been investigated on a molecular level. Two of the main questions are for example what the consequences on guanylate cyclase activity are and which Ca2+-dependent activation profiles can be observed in the presence of GCAPs. These topics are addressed in the present contribution.

#### **MATERIALS AND METHODS**

#### **CLONING AND SITE-DIRECTED MUTAGENESIS OF HUMAN ROS-GC1 CONSTRUCTS**

The cDNA of human ROS-GC1 was cloned into the vector pIRES2-eGFP (Clonetech) exactly as described previously (Zägel et al., 2013). DNA was prepared by using the DNA purification kit of Promega according to manufacturer's instruction. The mutant P575L was produced by using the following primers for PCR amplification: 5- -GCTATCCGCCTAGCAACCAAGACG-3- (forward) and 5- -CGTCTTGGTTGCTAGGCGGATAGC-3- (reverse). Site-directed mutagenesis yielding the mutants H1019P and P1069R was achieved by performing two PCRs each using the ROS-GC1-WT gene in the pIRES2-eGFP vector as template. Primers for the PCR were for creating H1019P: 5- -CCTTACCGCATCCCCGTGAACTTGAG-3 and 5- -TCAAGAGAACTGGCCCGGCCGC-3- (forward and reverse primer, first PCR) and 5- -AAAGCTAGCCACCATGACCGCC-TGCGCCCGCC-3 and 5- -CTCAAGTTCACGGGGATGCGG-TAAGG-3- (forward and reverse primer, second PCR). P1069R was obtained with 5- -CCCATCCCCAAACGGCCTGACCTG-3 and 5- -TCAAGAGAACTGGCCCGGCCGC-3- (forward and reverse primer in the first PCR) and 5- -AAAGCTAGCCACCATGACCGC CTGCGCCCGCC-3 and 5- -CAGGTCAGGCCGTTTGGG-GATGGG-3- (forward and reverse primer in the second PCR). The resulting two fragments were used in a further PCR to obtain a complete mutated ROS-GC1-DNA construct (3320 bp) by employing the primers 5- -AAAGCTAGCCACCATGACCGCCTG-CGCCCGCC-3 and 5- -TCAAGAGAACTGGCCCGGCCGC-3- . The final construct was cut with Nhe I, phosphorylated with T4 polynucleotide kinase and ligated in the pIRES2-EGFP vector that was cut by Nhe I/Sma I. Clone products were verified by sequencing.

#### **HETEROLOGOUS EXPRESSION OF ROS-GC1 AND GCAPs**

WT and mutant forms of ROS-GC1 were expressed in HEK flip 293 cells. For this purpose cells were cultivated in minimal essential medium under standard conditions as described

(Koch and Helten, 2008). After culturing 3–5 <sup>×</sup> <sup>10</sup><sup>6</sup> cells were harvested in 100 μL of a buffer containing 5 mM KCl, 20 mM MgCl2, and 150 mM phosphate buffer pH 7.2. For transfection, 5 μg of DNA was added and cells were electroporated with the CLB system (Lonza). Further treatment of cells and harvesting for subsequent guanylate cyclase assays was done exactly as described (Zägel et al., 2013). GCAP1 and GCAP2 were expressed in *E. coli* and purified to homogeneity by anion exchange and size exclusion chromatography as previously reported (Hwang et al., 2003; Koch and Helten, 2008). Myristoylation of GCAPs during bacterial expression was accomplished by co-transforming *E. coli* cells with *N*-myristoyl-transferase from yeast and supplementation with myristic acid. To ensure a high degree of myristoylation of GCAP1, which lacks the consensus site for yeast *N*-myristoyl-transferase, we used the D6S-mutant of GCAP1 (Krylov et al., 1999; Hwang et al., 2003). The degree of myristoylation of both GCAPs was typically higher than 90% as checked by using an analytical HPLC run on a reverse-phase column.

#### **GUANYLATE CYCLASE ASSAY**

unprocessed human sequence.

Recombinant human ROS-GC1 in HEK cell membranes was measured as described before (Koch and Helten, 2008; Zägel et al., 2013). In addition to measuring the basal guanylate cyclase activity we mainly tested two effects of GCAPs on the activation profile of the cyclase. In a first set we added purified bovine GCAP1 or GCAP2 at saturating 10 μM to washed HEK cell membranes and incubated samples at different free [Ca2+] using a Ca2+-EGTA buffer system. Details of the procedure have been described in several previous publications (Hwang et al., 2003; Koch and Helten, 2008; Zägel et al., 2013). Data were evaluated to obtain the [Ca2+], at which the activation is half-maximal (IC50). In a second set of guanylate cyclase assays we incubated recombinant ROS-GC1 with increasing concentrations of GCAP1 and GCAP2 (0–20 μM) for 30 min under Ca2+-free conditions. GCAPs were liganded with Mg2<sup>+</sup> (Peshenko and Dizhoor, 2004) using the same buffer composition as described (Koch and Helten, 2008). The latter activity measurements yielded apparent affinities (EC50-values) of GCAPs for WT and mutant ROS-GC1. Activity values were obtained from at least 3–4 different data sets and were used to calculate the mean ± standard deviation (SD). A student's *t*-test was performed to check for significant differences between two sets of data.

#### **IMMUNOHISTOCHEMISTRY**

Localization of ROS-GC1 WT and mutants in HEK cells by immunofluorescence microscopy was done as described before using an Olympus fluorescence microscope (Zägel et al., 2013). The following primary antibodies were used for detection: anti-ROS-GC1 (1:100; ROS-GC1 (H-225), sc50512, rabbit polyclonal IgG; Santa Cruz Biotechnology), anti-Na+/K+-ATPase (1:200; Na+/K+-ATPase, a (H-3), sc-48345, mouse monoclonal IgG2b; Santa Cruz Biotechnology), and anti-calnexin [1:200; calnexin (E-10), sc-46669, mouse monoclonal IgG2a; Santa Cruz Biotechnology]. Secondary antibodies were donkey anti-rabbit conjugated to Fura350 (dilution 1:200) from Invitrogen and goat anti-mouse conjugated with Dylight594 from Thermo Scientific, USA used in a dilution of 1:500. Incubation and washing buffers were exactly as described (Zägel et al., 2013).

#### **RESULTS AND DISCUSSION**

#### **LOCALIZATION OF ROS-GC1 CONSTRUCTS**

Previous work has shown that ROS-GC1 can be expressed heterologously in HEK293 cells in sufficient quantity to allow subsequent biochemical studies (Hwang et al., 2003; Koch and Helten, 2008; Zägel et al., 2013). In HEK293 cells the enzyme was found to be present in cell membranes and mainly co-localized with the endoplasmic reticulum. Probing the cells with an anti-ROS-GC1 antibody (**Figure 2A**, middle panel) and using specific markers for the plasma membrane (anti-Na+/K+-ATPase antibody, top left panel in **Figure 2A**) and for the endoplasmic reticulum (anticalnexin, top right panel) we obtained a staining pattern in the overlay image (**Figure 2A**, bottom panels) in agreement with published results (Peshenko et al., 2008; Zägel et al., 2013). The localization of all investigated mutants was very similar to the cellular localization of the WT. Using western blotting we further estimated that the amount of ROS-GC1 WT and mutants was nearly the same in a suspension of cells having a similar cell density (**Figure 2B**, also Zägel et al., 2013).

#### **BASAL GUANYLATE CYCLASE ACTIVITIES**

The mutant P575L exhibited similar basal GC activities as WT (**Figure 3**). Activities in the presence of GCAP1 and 2 at saturating [Ca2+] revealed almost no difference between P575L and WT. However, a difference was observed, when GCAP1 and 2 were incubated with ROS-GC1 at low [Ca2+] (EGTA columns in **Figure 3**) resulting in a 25–30% decrease of maximal activity.

(lower panels). Secondary antibodies were indicated in the Methods section. Visualization was done in an Olympus fluorescence microscope. Cells were fixed either with paraformaldehyde (localization of plasma membrane) or with paraformaldehyde-methanol (endoplasmic reticulum). Scale bar = 10 μm. **(B)**Western blot of HEK293 cells expressing human ROS-GC1,WT and the mutants P575L, H1019P, and P1069R. Blots were probed with a primary antibody against ROS-GC1 (GC1#3, dilution 1:2000) and a secondary anti-rabbit antibody from Dianova (dilution 1:4000) as described (Zägel et al., 2013).

In sharp contrast to P575L, the mutant P1069R did not exhibit any activity, neither in the presence nor in the absence of Ca2+, EGTA, or GCAPs (data not shown). We also tested activities of cells transiently expressing the mutant or a stably transfected cell line. Since the cellular localization of this mutant appeared normal and was very similar to the active WT and P575L forms (**Figure 2A**), we conclude that the lack of GC activity was a consequence of an impaired catalytic domain.

A special case was observed with the mutant H1019P: we were unable to obtain consistent activity profiles as we did with WT and the P575L mutant. In some cell batches we observed activation by GCAP1, but not by GCAP2, whereas in other batches both GCAPs were able to activate the cyclase or the enzyme was inactive under all conditions. The large scattering of the results seems to indicate a general instability of the mutant.

#### **GCAP-MEDIATED ACTIVATION**

The key feature among ROS-GC1 properties is the Ca2+ dependent activation under control of GCAPs. We tested this activation of P575L by GCAPs and compared it with the WT activity profile. Representative data sets are shown in **Figure 4A** (+GCAP1) and **Figure 4B** (+GCAP2), a summary of the IC50 values (mean <sup>±</sup> SD) is displayed in **Table 1**. For both Ca2+-sensor proteins we observed a similar two-fold shift of the IC50 values to lower [Ca2+], in the presence of GCAP1 from 0.529 (WT) to 0.270 μM (P575L) and in the presence of GCAP2 from 0.269 μ(WT) to 0.128 μM (P575L).

A shift in the IC50 values had previously been reported for some autosomal cone dystrophy related mutations that are located in a so-called "mutation hotspot" region within or near the residue R838 in the dimerization domain (Duda et al. 1999a, 2000; Wilkie

**FIGURE 4 | Ca2+-sensitive activity profiles ofWT and ROS-GC1 mutant P575L.** Guanylate cyclase activities were measured as a function of free [Ca2+] for samples as indicated **(A)** WT or P575L, in the presence of 10 μM GCAP1. **(B)** WT or P575L, in the presence of 10 μM GCAP2. Incubation time was 30 min. Activities are normalized to allow comparison of IC50 values ([Ca2+]free at which activity is halfmaximal). Data are representative of three to four sets of incubations.

et al., 2000; Kitiratschky et al., 2008; Zägel et al., 2013). One reasonable model explaining this observed effect states that the shift in Ca2+-sensitivity originates from different apparent affinities of WT and mutant ROS-GC1 for the Ca2+-free form of GCAPs



\*Six data sets; +two data sets.

Values are the mean ± SD of mainly three to four independent measurements, if not indicated otherwise.

(Peshenko et al., 2004). We tested this hypothesis for the P575L mutant by measuring the GC activities at increasing GCAP concentrations (**Figure 5**). GCAP1 and 2 were Ca2+-free, but they had Mg2+-bound (Peshenko and Dizhoor, 2004) under the assay conditions. Apparent affinities expressed as EC50-values differed only minimally between WT and the P575L mutant and overlapped within the measured standard deviations (**Table 1**). Thus, we conclude that differences in apparent affinities for GCAPs cannot accountfor the observed shifts in Ca2+-sensitivity of the P575L mutant.

#### **STRUCTURE-FUNCTION RELATIONSHIPS**

The single point mutation P575L is located in direct C-terminal neighborhood of a small stretch of 20 amino acids, which was previously identified as an essential regulatory and/ or interaction site in bovine ROS-GC1 for GCAP1 (L559–I578 corresponding to peptide #34a in Lange et al., 1999). The region is highly conserved among different species and encompasses also the point mutation F565S (human sequence) that is linked to LCA type 1 (Perrault et al., 1996). However, the biochemical consequences of these two mutations differ significantly, since F565S leads to a more than

10-fold lower basal cyclase synthesis rate, but also to a complete loss of the GCAP1-stimulated GC activity (Duda et al., 1999b). These biochemical consequences are in contrast to the less severe effects caused by the P575L mutation.

The position H1019P is located in the short β6-sheet of the CCD in close vicinity of the β5-sheet that harbors amino acids R995, C997, L998, and F999, which are critical for binding the purine ring of the GTP substrate (Tucker et al.,1998). Replacement of H by P might cause a "kink" of the β6-sheet and therefore disrupts or destabilize the catalytic center. A mutation of A to P in α-synuclein for example was reported to have long-range and short-range effects on the three-dimensional structure including a decrease of β-sheet content (Wise-Scira et al., 2013). Furthermore, the position H1019P is also located in the GCAP2 binding site that was previously narrowed down to the amino acid stretch Y1012- N1037 in the bovine sequence (Duda et al., 2005; Pettelkau et al., 2012). In summary, these results might explain the large scattering of GC activities observed in the *in vitro* assays.

In our test system the mutation P1069R had the most dramatic consequences leading to a completely inactive enzyme. This position is sandwiched between the binding sites of GCAP2 and S100β (Duda et al., 2002) and we can only speculate that it might have the function to integrate diverse signals. Furthermore, it seems to be indispensable for the basal catalysis of ROS-GC1.

## **CONCLUSION**

Inherited retinal disorders can affect visual function in different manners. In the present study we investigated the biochemical consequences of three point mutations in the GUCY2D gene, which cause different forms of visual impairment. The mutation P575L causes a progressive form of cone degeneration in affected patients, which very often correlates with symptoms of photophobia, decrease of central vision and impairment of color vision. Since the mutation did not lead to a complete loss of guanylate cyclase function, cone cells expressing the mutated ROS-GC1 protein very likely maintain key features of photoreceptor physiology. But since the mutation induces a dysregulation of the Ca2+-sensitive cyclase activation profile, it seems to have longterm effects in the distortion of the Ca2+-cGMP homeostasis triggering apoptotic pathways. Similar consequences have been discussed for GCAP1 mutations correlating with certain forms of cone-rod dystrophies (Behnen et al., 2010; Hunt et al., 2010). The other two point mutations that were investigated in the present study correlate with LCA and a juvenile form of retinitis pigmentosa. Both mutations have more severe pathological consequences. Typical clinical features are for example an early onset of the disease in childhood and a strongly reduced or even nonrecordable electroretinogram (Perrault et al., 2000; Booji et al., 2005). Clinical manifestations of the point mutations appear to be mirrored in the biochemical properties of the expressed ROS-GC1 protein as there are: synthesis of cGMP by ROS-GC1 is either severely disrupted due to protein instability (H1019P) or completely abolished (P1069R). Thus, light-triggered hydrolysis of cGMP cannot adequately be balanced by the second photoreceptor specific guanylate cyclase (ROS-GC2, GC-F), which has a much lower expression level, at least in cone cells (Olshevskaya

et al., 2002; Baehr et al., 2007). The cGMP level in photoreceptor cells would be lower than in non-affected cell. This in turn would keep a lower number of cyclic nucleotide-gated channels open in the dark state. Since the cytoplasmic cGMP pool that is hydrolyzed after illumination is already reduced in the dark, amplitude and kinetics of the photoresponse are expected to be smaller and faster.

#### **ACKNOWLEDGMENT**

We thank Jutta Appelt for technical assistance in the cell culture laboratory and Dr. Alexander Scholten for fruitful discussions. This study was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG) to Karl-Wilhelm Koch (KO948/10-1).

#### **REFERENCES**


α-synuclein proteins with dynamics. *ACS Chem. Neurosci.* 4, 486–497. doi: 10.1021/cn300198q

Zägel, P., Dell'Orco, D., and Koch, K.-W. (2013). The dimerization domain in outer segment guanylate cyclase is a Ca2<sup>+</sup> -sensitive control switch module. *Biochemistry* 52, 5065–5074. doi: 10.1021/bi400288p

**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: 25 November 2013; accepted: 10 February 2014; published online: 26 February 2014.*

*Citation: Zägel P and Koch K-W (2014) Dysfunction of outer segment guanylate cyclase caused by retinal disease related mutations. Front. Mol. Neurosci. 7:4. doi: 10.3389/fnmol.2014.00004*

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

*Copyright © 2014 Zägel and Koch. 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.*

## Insights gained from gene therapy in animal models of retGC1 deficiency

## *Shannon E. Boye\**

Department of Ophthalmology, University of Florida, Gainesville, FL, USA

#### *Edited by:*

Wolfgang Baehr, University of Utah, USA

#### *Reviewed by:*

Wolfgang Baehr, University of Utah, USA Izhak Michaelevski, Tel Aviv University, Israel

#### *\*Correspondence:*

Shannon E. Boye, Department of Ophthalmology, University of Florida, P.O. Box 100284, Gainesville, FL 32610, USA e-mail: shannon.boye@eye.ufl.edu

Vertebrate species possess two retinal guanylate cyclases (retGC1 and retGC2) and at least two guanylate cyclase activating proteins (GCAPs), GCAP1 and GCAP2. GCAPs function as Ca2<sup>+</sup> sensors that regulate the activity of guanylate cyclases. Together, these proteins regulate cGMP and Ca2<sup>+</sup> levels within the outer segments of rod and cone photoreceptors. Mutations in GUCY2D, the gene that encodes retGC1, are a leading cause of the most severe form of early onset retinal dystrophy, Leber congenital amaurosis (LCA1).These mutations, which reduce or abolish the ability of retGC1 to replenish cGMP in photoreceptors, are thought to lead to the biochemical equivalent of chronic light exposure in these cells. In spite of this, the majority of LCA1 patients retain normal photoreceptor laminar architecture aside from foveal cone outer segment abnormalities, suggesting they may be good candidates for gene replacement therapy. Work began in the 1980s to characterize multiple animal models of retGC1 deficiency. 34 years later, all models have been used in proof of concept gene replacement studies toward the goal of developing a therapy to treat GUCY2D-LCA1. Here we use the results of these studies as well as those of recent clinical studies to address specific questions relating to clinical application of a gene therapy for treatment of LCA1.

**Keywords: LCA1, Leber congenital amaurosis,** *GUCY2D***, retGC1, GC1, guanylate cyclase, retinal gene therapy, AAV**

## **INTRODUCTION**

The ability to process light into an electrochemical signal depends on the precise regulation of levels of cGMP and Ca2<sup>+</sup> within the outer segments of rod and cone photoreceptor cells. In the absence of light stimulation, intracellular cGMP and Ca2<sup>+</sup> levels are high and a continuous flow of Na<sup>+</sup> and Ca2<sup>+</sup> ions through cGMP-gated channels and Na+/Ca2<sup>+</sup> exchangers keeps the cell in a depolarized state. Activation of rhodopsin or cone opsins in rod and cone outer segments, respectively, leads to activation of transducin that in turn activates cGMP phosphodiesterase (PDE) leading to a reduction in the concentration of cGMP in the outer segment. Reduction of cGMP leads to closure of cGMP-gated channels, reduced Na+/Ca2<sup>+</sup> influx and hyperpolarization of the cell (Pugh et al., 1997). Recovery of light stimulated photoreceptors to the dark state is hastened by the decrease in intracellular Ca2<sup>+</sup> that results from continued expulsion of Ca2<sup>+</sup> by Na+/Ca2+-K<sup>+</sup> exchangers. Calcium has several roles in photoreceptor cells, one of which is to regulate the activity of retinal guanylate cyclases, retGC1 and retGC2 (Koch and Stryer, 1988). It does so by binding to guanylate cyclase activating proteins (GCAPs), a family of calcium-binding proteins that regulate the activity of retGCs (Polans et al., 1996). Two variants of guanylate cyclases (retGC1 and retGC2) and at least two variants of GCAPs (GCAP1 and GCAP2) are present in the outer segments of vertebrate photoreceptors (Shyjan et al., 1992; Margulis et al., 1993; Dizhoor et al., 1994; Goraczniak et al., 1994, 1997; Palczewski et al., 1994; Lowe et al., 1995; Yang et al., 1995; Imanishi et al., 2004). In the dark, high levels of intracellular Ca2<sup>+</sup> promote its binding to GCAP and inhibit the ability of

this protein to activate retGCs. Reduction of free, intracellular Ca2<sup>+</sup> following light stimulation leads to a decrease in the amount of Ca2<sup>+</sup> bound to GCAP in exchange for Mg2+, which allows this protein to activate retGCs (Peshenko and Dizhoor, 2006; **Figure 1**). *In vivo*, retGC1 is the preferred target of GCAP1 (Olshevskaya et al., 2012). *In vitro*, some studies show that GCAP1 and GCAP2 can activate both retGCs, albeit with slightly different affinities and sensitivity to Ca2<sup>+</sup> (Hwang et al., 2003; Peshenko and Dizhoor, 2004; Peshenko et al., 2011) while another shows that GCAP1 cannot activate retGC2 (Haeseleer et al., 1999).

Mutations in *GUCY2D*, the gene that encodes retGC1, are associated with the severe, early onset, autosomal recessive disorder Leber congenital amaurosis-1 (LCA1; Perrault et al., 1996, 2000). LCA1 causing mutations are found throughout *GUCY2D*, can alter the enzyme's structure or stability, affect transport of other peripheral membrane associated protein, and are frequently null (Karan et al., 2010; Jacobson et al., 2013). Without fully functional retGC1 to replenish intracellular cGMP, as is the case in LCA1, cGMP-gated cation channels will remain closed as they are in the light stimulated state. Hence, mutations in *GUCY2D* are thought to induce the biochemical equivalent of chronic light exposure in photoreceptors. LCA1 patients routinely present with severely reduced visual acuities, attenuated or extinguished electroretinogram (ERG), nystagmus, digito-ocular signs, and apparently normal fundus appearance (Perrault et al., 1999a, 2000; Hanein et al., 2004; Yzer et al., 2006; den Hollander et al., 2008). While early reports indicated retinal degeneration was associated with this form of LCA (Milam et al., 2003; Porto et al., 2003),

later studies revealed preservation of retinal laminar architecture in LCA1 patients (Simonelli et al., 2007; Pasadhika et al., 2010; Jacobson et al., 2013).

Preservation of retinal structure despite profound visual disturbance suggests that LCA1 patients would be good candidates for gene replacement therapy. Before initiating such clinical trials there are major questions that need to be answered. Here we stepwise address these questions by reviewing findings of proof of concept studies utilizing different animal models of retGC1 deficiency (results summarized in **Table 1**). Additionally, we provide an update on efforts to clinically apply a gene therapy for LCA1.

### **WILL** *GUCY2D* **GENE REPLACEMENT RESTORE RETINAL FUNCTION** *IN VIVO***?**

The first animal model used to investigate gene replacement for retGC1 was the *retinal degeneration* (*rd*) chicken, also referred to as the GUCY1∗B chicken. This cone- dominant, avian model carries a null, naturally occurring deletion/insertion mutation in the gene encoding retGC1 (Semple-Rowland et al., 1998). As a consequence of retGC1 deficiency, photoreceptors of post-hatch day 1 (P1) GUCY1∗B chickens have only 10–20% cGMP relative to thatfound in age-matched, wild type controls and affected chickens are blind at hatch. They have unrecordable ERGs and lack optokinetic and volitional visual behavior (Ulshafer et al., 1984; Williams et al., 2006). These biochemical and visual disturbances occur prior to photoreceptor loss which doesnot begin until 1 week post-hatch (Cheng et al., 1980; Ulshafer et al., 1984) and proceeds in a central to peripheral fashion with cones completely lost by 3.5 months and rods lost by 8 months (Ulshafer and Allen, 1985a,b).

Due to the technical difficulty associated with performing subretinal injections in chicken, an *in ovo* treatment paradigm was developed. For its ability to stably transduce retinal progenitor cells (Miyoshi et al., 1997), HIV-1-based lentivirus (LV) was used to deliver a cDNA encoding bovine retGC1 under the control of an elongation factor 1 alpha (EF1α) promoter to the neural tube of developing GUCY1∗B embryos (Williams et al., 2006). Bovine retGC1 (bGC1) was chosen for its verified activity in the presence of chicken GCAPs *in vitro* (Williams et al., 2006). At 1 month post-hatch, ERG testing under both dark and light-adapted conditions revealed that LV-EF1α-bGC1 treatment produced modest increases in photoreceptor-mediated awave amplitudes in treated chickens (∼6% of wild type). These results were the first demonstration that retGC1 gene replacement could restore retinal function. ERG improvements were associated with a restoration of visually guided behavior, as assessed by

**Table 1 | Summary of proof of concept experiments.**


optokinetic reflex testing (Wallman andVelez, 1985) and volitional visually guided behavior tests (Williams et al., 2006). In addition, a slowing of retinal degeneration was observed (Williams et al., 2006). Taken together, these results were exciting proof of concept that gene replacement therapy could be effective for treatment of LCA1. However, there were multiple limitations associated with this study that needed to be overcome. First, the therapeutic effect was transient, with ERG and behavioral responses disappearing after the latest time points analyzed (5 weeks posthatch) and retinal degeneration continuing unabated (Williams et al., 2006; Verrier et al., 2011). Second, multiple facets of the therapeutic strategy lacked clinical translatability- (1) Therapeutic retGC1 was delivered embryonically, a currently untenable task in patients and also requiring in-utero genotyping. (2) An integrating, lentiviral vector was used. While proven effective for transducing retinal precursors, LV has demonstrated little to no ability to transduce post-mitotic photoreceptors (Miyoshi et al., 1997; Bainbridge et al., 2001; Pang et al., 2006; Lipinski et al., 2014). (3) Data was obtained in a non-mammalian model of retGC1 deficiency. Together, this highlighted the need to test therapy using more clinically relevant animal models and vector platform.

### **WILL POST-NATALLY DELIVERED retGC1 RESTORE RETINAL FUNCTION IN A MAMMALIAN MODEL OF retGC1 DEFICIENCY?**

The first mammalian model retGC1 deficiency to be described was the guanylate cyclase-1 knockout (GC1KO) mouse (Yang et al., 1999). Insertion of a neomycin resistance cassette in exon 5 of *Gucy2e* (the murine homolog of *GUCY2D*) resulted in truncation of retGC1, rendering it null (Yang et al., 1999). Cone-mediated ERGs are unrecordable by 5 weeks but cones are not completely degenerated in this mouse until ∼6 months of age (Yang et al., 1999; Coleman et al., 2004). Rods, on the other hand, do not

degenerate and maintain variable levels of function (∼30–50% of WT; Yang et al., 1999) which is owed to the presence of retGC2 in those cells (Baehr et al., 2007). GCAP1 and GCAP2 transcripts and GCAP1 expression are down-regulated (Coleman et al., 2004) and light-induced translocation of cone arrestin is disrupted as a result of retGC1 deficiency in this model (Coleman and Semple-Rowland, 2005). The temporal dissociation between loss of cone function and structure in GC1KO mice, provides a window of therapy for gene replacement.

For its safety profile and proven ability to transduce postmitotic photoreceptors following subretinal injection of murine retina (Yang et al., 2002), a serotype 5 adeno associated virus (AAV) was chosen to deliver bovine retGC1 (the same cDNA delivered to the GUCY1∗B chicken) under the control of either the ubiquitous, small CBA (smCBA) or photoreceptor-specific, murine opsin (MOPS) promoter to the GC1KO mouse (Haire et al., 2006). Postnatal day 21 (P21) treatment with AAV5-bGC1 failed to improve retinal function (ERG) in treated mice. However, AAV-mediated retGC1 expression was restricted to photoreceptor outer segments and restoration of light-induced cone arrestin translocation, a biochemical correlate of functionality in these cells, was observed at 4 weeks post-treatment (Haire et al., 2006), suggesting that the light signaling cascade in GC1KO cones had been at least partially reset following treatment.

Investigators speculated that the species non-specific nature of the delivered transgene may have dampened the therapeutic outcome. Indeed, subretinal delivery of murine *Gucy2e* via an AAV5 vector containing either the ubiquitous smCBA or photoreceptorspecific human rhodopsin kinase (hGRK1) promoter led to robust improvements (∼45% of WT) in cone-mediated ERGs (Boye et al., 2010). These improvements were stable over the course of this initial study (at least 3 months post-treatment) and provided the first evidence that an AAV-based vector could restore retinal function to a mammalian model of retGC1 deficiency (Boye et al., 2010). Later studies would go on to show that stable restoration of cone function (ERG) is achievable over the long term (Boye et al., 2011; Mihelec et al., 2011). Subretinally delivered AAV8 containing the hGRK1 promoter and human *GUCY2D* cDNA stably restored cone function (∼65% of WT) for at least 6 months (Mihelec et al., 2011). In the longest follow up reported to date, subretinally delivered AAV5 or capsid mutant AAV8(Y733F) containing either the smCBA or hGRK1 promoter and murine *Gucy2e* cDNA stably restored cone function (∼45% of WT) for at least 1 year post-treatment (Boye et al., 2011). A notable difference between these two studies was the level of cone function achieved (65 vs. 45% of WT). Mihelec et al. (2011) subretinally delivered AAV-*GUCY2D* to GC1KO mice at P10 (an age prior to natural eye opening) whereas Boye et al. (2011) delivered AAV-*Gucy2e* between P14–P25. Earlier intervention was likely more effective at combating the chronic effects of hyperpolarization that GC1KO cones endure upon light stimulation. Regardless of these differences, the robust and stable functional improvements in cone function following AAV-retGC1 treatment (including with a clinically relevant human cDNA) in GC1KO mice laid the groundwork for development of an AAV-based treatment for LCA1.

As a follow up to these studies, cone-mediated function was also evaluated in the retGC1/retGC2 double knockout (GCDKO mouse following subretinal delivery of AAV8(Y733F)-hGRK1- *Gucy2e* (Boye et al., 2013). Treatment of this model, in which cones and rods are functionally silent and progressively degenerate, also resulted in stable and robust improvements in cone function for at least 1 year post-treatment (Boye et al., 2013). As in all GC1KO studies, AAV-mediated retGC1 expression was restricted to the outer segments of rods and cones in treated GCDKO mice. The GCDKO mouse and gene replacement experiments in this model will be discussed in more detail below.

## **DOES RESTORATION OF CONE FUNCTION TRANSLATE INTO USEFUL VISION?**

Adeno associated virus-mediated expression of retGC1 in photoreceptors of GC1KO mice effectively reset the phototransduction cascade in cones, allowing them to send electrochemical signals to downstream neurons which were detected via full field ERG electrodes (Boye et al., 2010, 2011; Mihelec et al., 2011). However, the existence of functional neural circuits within the retina does not ensure that electrochemical signals are properly relayed to higher order processing centers in the brain and experienced as "vision." Even if the biochemical defect within photoreceptors is corrected via restoration of retGC1, it is unclear what role amblyopia (the inability of the retina to send visual information to the brain) will play on the treatment outcome. To gain insight into this question, multiple visually guided behavior tests were used to further characterize therapy in AAV-treated GC1KO mice.

Two tests, each of which evaluates the integrity of different visual processing centers of the brain, were used to determine whether useful cone vision could be restored to treated mice. A virtual optokinetic system was used to evaluate the integrity of subcortical retinal efferents (Douglas et al., 2005). While untreated GC1KO mice lack cone-mediated behavior, significant

improvements in cone-mediated spatial frequency thresholds and contrast sensitivities were seen in GC1KO mice treated with AAV5 hGRK1-*Gucy2e*, AAV5-smCBA-*Gucy2e*, AAV8(Y733F)-hGRK1- *Gucy2e* or AAV8-hGRK1-*GUCY2D* (Boye et al., 2010, 2011; Mihelec et al., 2011) with behavior resembling that of normal, sighted congenic mice. Improvements in cone-mediated behavior were observed out to 4 months post-treatment (Mihelec et al., 2011), the latest time point evaluated in this mouse model. Analysis of cone-mediated behavior was also evaluated in GCDKO mice via Morris Water Maze (under photopic conditions), a test used to analyze cortically driven visual behavior (Boye et al., 2013). AAV8(Y733F)-hGRK1-*Gucy2e* treated mice exhibited significantly reduced escape latencies (the amount of time required to escape water bath via a platform demarcated with a flag) relative to untreated controls. Importantly, behavior of treated mice was not significantly different from age-matched WT controls. Notably, these results were obtained in 1 year old GCDKO mice treated 11 months prior, highlighting that retGC1 supplementation confers useful cone-mediated vision to mammalian models of LCA1 over the long term. This is significant as humans rely heavily on their cone-mediated vision for activities of daily living. It is also interesting to note that while cone-mediated ERG improvements in both GC1KO and GCDKO controls ranged between 45 and 65% that of WT controls, conemediated behavior was restored to "normal" in AAV-treated mice supporting previous work showing that behavior is a more sensitive indicator of therapy (Williams and Jacobs, 2007). This was also shown to be the case in LV-retGC1 treated GUCY1∗B chickens whose modest ERG responses corresponded, albeit transiently, to robust visually guided behavior (Williams et al., 2006).

## **HOW DO ROD PHOTORECEPTORS RESPOND TO TREATMENT?**

Prior to the most comprehensive clinical characterization of LCA1 patients published to date (Jacobson et al., 2013), multiple reports had described that both rod/cone function and rod/cone structure were compromised (Perrault et al., 1999a,b; Milam et al., 2003; Porto et al., 2003; den Hollander et al., 2008; Chung and Traboulsi, 2009; Jacobson et al., 2013). It was clear that, despite the presence of retGC2 in these cells (Lowe et al., 1995), rod photoreceptors of LCA1 patients are negatively impacted by retGC1 deficiency. Thus, the specific effects of post-natal gene replacement therapy on this photoreceptor subclass needed to be analyzed. Prior studies were conducted in a model that lacks rod degeneration and retains variable levels of rod function throughout its life, the GC1KO mouse. For these reasons, it was not possible to address whether gene replacement would prevent degeneration of rods or definitively restore their function, respectively. Significant improvements in rod-mediated ERGs were only obtained when vector was delivered prior to natural eye opening (P10; Mihelec et al., 2011), but it is difficult to ascertain whether these gains are meaningful because GC1KO mice exhibit normal rod-mediated visual behavior. It wasnot until generation of the GCDKO mouse that it became possible to ask these questions (Baehr et al., 2007).

Deletion of both retGC1 and retGC2 in the GCDKO mouse (Baehr et al., 2007) renders both cones and rods functionally silent (ERGs are unrecordable) and degenerative (Baehr et al., 2007). Several cone proteins (GCAP1, GCAP2, cone opsins, cone transducin, cone PDE, GRK1) are downregulated/mislocalized) and PDE6 is absent from rods (Baehr et al., 2007) in this model. By 2 months of age, overall outer segment length is reduced to 30– 50% of normal (Baehr et al., 2007). Appreciable thinning of the outer nuclear layer (ONL) is apparent by ∼3.5 months of age and by 6 months, only 3–4 photoreceptor nuclei remain (Boye et al., 2013).

For its success in the studies described above, AAV8(Y733) hGRK1 was chosen to subretinally deliver *Gucy2e* to cohorts of GCDKO mice of various age groups (P18, P21–P25, P37–P49, and P108). ERG revealed that, like the cone results described above, rod function was also stably restored to ∼40% of normal (at least 1 year post-treatment; Boye et al., 2013). Morris Water Maze testing revealed that these functional gains translated to useful rod-mediated vision, with AAV-retGC1- treated GCDKO mice performing the task at speeds that did not differ significantly from age-matched WT controls (Boye et al., 2013). As in the treated GC1KO mice, WT-like behavior was achievable in GCDKO mice that exhibited only partial ERG recovery. Optical coherence tomography (OCT) was used to monitor the rate of photoreceptor degeneration (predominantly rods) in treated vs. untreated mice. By 7–12 months post-treatment, there was significant structural preservation in all treated GCDKOs, barring those treated at the latest time point (P108), although retinas in these mice still exhibited greater preservation than controls (Boye et al., 2013). Importantly, the rate of photoreceptor cell loss in GCDKO mice treated as late as P49 was no different from that of age-matched WT controls. GCAP1 and GCAP2 were expressed at WT-like levels at 1 year post-injection and cone opsins were present only in treated cones (Boye et al., 2013). Taken together, this study showed for the first time that, along with the rescue effects seen on cones,AAV-mediated retGC1 expression can restore rod photoreceptor function and rod-mediated visual behavior and preserve rod structure over the long term in an animal model of LCA1.

## **WHAT IS THE FUNCTIONAL EFFICIENCY OF AAV- DELIVERED retGC1 ENZYME?**

In considering clinical application of a gene therapy, an important question becomes "how efficient is the vector-derived, recombinant gene product relative to the native protein?" The ability to assess this *in vivo* provides a powerful bioassay which can be used to establish dose-response relationships and equivalencies for batches of clinical vectors. Guanylate cyclase activity assays are performed on dark-adapted retinas to determine how much cGMP is produced by retGCs in the presence of a [α-32P]GTP substrate (Olshevskaya et al., 2004; Peshenko et al., 2011). This assay doesnot discriminate between the activity of retGC1 and retGC2. Therefore, isolating the functional efficiency of exogenous retGC1 requires delivery to photoreceptors that lack both endogenous retGC1 and retGC2, such as those in the GCDKO mouse. Boye et al. (2013) compared retGC activity in retinas of GCDKO mice that were treated subretinally between P30–P60 with AAV8(Y733F)-hGRK1-*Gucy2e*, relative to untreated GCDKO or WT controls and those that received a serotype-matched control

vector expressing GFP. Maximal retGC activity in AAV-*Gucy2e* treated retinas was ∼63% of normal. The known contribution to total cyclase activity by retGC2 in this assay is between 20 and 28% (Peshenko et al., 2011). In addition, the area of retGC1 expression in treated retinas is known to be restricted to the subretinal injection bleb (Timmers et al., 2001; Cideciyan et al., 2008). "Good" subretinal injections typically result in ∼80– 90% detachment meaning not all photoreceptors expressed the enzyme. Taking these two factors into account, the level of retGC1 activity suggested its near-complete restoration in the area exposed to vector. When the maximal activity of AAV-retGC1 was normalized to that seen in WT retinas, it was determined that the calcium sensitivity of the exogenous and endogenous enzymes were identical (Boye et al., 2013). Moving forward, this assay may be used to evaluate the relative potency of clinical vectors.

## **WHAT ARE THE CONSEQUENCES OF retGC1 DEFICIENCY ON HUMAN ROD AND CONE PHOTORECEPTOR STRUCTURE AND FUNCTION?**

While there is general consensus that LCA1 is associated with severely attenuated or ablated ERG (Perrault et al., 1999b, 2000; Hanein et al., 2004; Yzer et al., 2006; den Hollander et al., 2008), reports on the extent of photoreceptor degeneration associated with this form of LCA have been conflicting (Milam et al., 2003; Porto et al., 2003; Simonelli et al., 2007; Pasadhika et al., 2010). The most thorough clinical characterization of LCA1 performed to date focused on a cohort of patients ranging in age from 6 months to 37 years with different mutations in *GUCY2D* and has provided new insight on the effects of retGC1 deficiency on retinal structure and function (Jacobson et al., 2013).

Similar to previous reports Simonelli et al. (2007), Pasadhika et al. (2010), Jacobson et al. (2013) found that LCA1 patients retained normal retinal laminar architecture aside from foveal cone outer segment abnormalities and, in some cases, foveal cone loss. Within rod-dominant retina, ONL thickness and outer segment lengths were normal in all patients (Jacobson et al., 2013). Unlike previous reports, Jacobson et al. (2013) found that LCA1 patients can retain substantial rod function. Full field sensitivity testing (FST) revealed all patients effectively used their rods to detect blue stimuli, albeit at reduced sensitivities. These psychophysical findings were supported by ERG, pupillometry and mobility testing, all of which revealed variable levels of retained rod function that did not correlate with the patient's age (Jacobson et al., 2013). In contrast, the majority of patients lacked cone sensitivity (FST) which correlated to severely reduced visual acuity and a lack of color perception. FST, microperimetry and mobility tests revealed a small subset had detectable, but reduced cone function with central fixation and some color perception. None of the patients had recordable cone-mediated ERGs (Jacobson et al., 2013).

To better understand these differential cone phenotypes, Jacobson et al. (2013)investigated the activity of patient-specific retGC1 mutants *in vitro*. Their activity was assayed by transfecting cDNA for each form, along with GCAPs in HEK293 cells under appropriate ionic conditions and quantifying the amount of cGMP produced relative to that achieved with wild type retGC1. Of the patient-specific mutants tested, some exhibited reduced catalytic activity (4–5 fold lower than WT) while others demonstrated no loss of function (Jacobson et al., 2013). Not surprisingly, those mutants that retained biochemical activity were expressed by LCA1 patients with measurable cone function. It is expected that those with normal activity *in vitro* may exhibit defective folding, expression or impaired trafficking to photoreceptor outer segments*in situ*. Investigations are currently underway to understand the fate of these mutants *in vivo* via AAV-mediated expression in GCDKO mice. retGC1 mutants that exhibited no activity *in vitro* corresponded to patients with profound loss of cone function (Jacobson et al., 2013).

## **WHAT FORM MIGHT A CLINICAL TRIAL FOR** *GUCY2D***-LCA1 TAKE?**

A few important choices must be made prior to clinical application of a gene replacement therapy for LCA1. These include the injection route and location, AAV serotype and promoter, and therapeutic endpoints. Route and location of injection should be dictated by the calculated risk(s)/benefit. The structural integrity of the patient's retina and locus of existing vision must be considered. Evidence suggests *GUCY2D*-LCA1 is unlike any form of LCA studied in detail to date (Jacobson et al., 1998, 2003, 2005, 2007a,b, 2009b, 2011; Cideciyan et al., 2007). Their hallmark retinal preservation suggests that LCA1 patients may be good candidates for subretinal injection of AAV-*GUCY2D* (Simonelli et al., 2007; Pasadhika et al., 2010; Jacobson et al., 2013). Because retGC1 deficiency leads to profound visual impairment, stemming primarily from cones, the central retina should be the treatment target. In the small subset of patients that exhibit foveal cone losses, para- or peri-foveal areas may be targeted in the hopes that, like some *RPE65*-LCA2 patients, they will develop an eccentric locus of fixation (Cideciyan et al., 2009).

AAV5 andAAV8-based vectors were used in the aforementioned studies for developing a treatment for LCA1. These vectors have proven utility for photoreceptor-targeted therapy following subretinal injection in multiple mouse and dog models of inherited retinal disease (Min et al., 2005; Alexander et al., 2007; Boye et al., 2010, 2013; Gorbatyuk et al., 2010; Komaromy et al., 2010; Mao et al., 2011; Pang et al., 2011, 2012; Yao et al., 2011; Beltran et al., 2012; Petit et al., 2012; Lheriteau et al., 2014). Because AAV transduction profiles and the activity of promoters can vary across species, the final decision should be dictated by the serotype and promoter combination's behavior in a species most closely related to man, non-human primate (NHP). In LCA1, photoreceptors of the central retina are the treatment target. Therefore, serotype and promoter combinations that effectively transduce cone photoreceptors in NHP are required. Transduction of both AAV5 and AAV8 vectors have recently been described in subretinally injected NHP (Boye et al., 2012; Vandenberghe et al., 2013). Only partial cone transduction was achieved following subretinal delivery of elevated doses of AAV8 containing the ubiquitous CMV promoter driving GFP (1011 vg delivered; Vandenberghe et al., 2011, 2013). At a lower dose (1010 vg delivered), AAV8-CMV-GFP failed to transduce foveal, parafoveal or perifoveal cones (Vandenberghe et al., 2011). In contrast, foveal, parafoveal, perifoveal and peripheral cones of NHP (as well as rods) were all

effectively transduced following subretinal delivery of an AAV5 vector containing the photoreceptor-specific hGRK1 promoter (1010 vg delivered; Boye et al., 2012). This mirrors earlier findings in rodent that hGRK1 has exclusive activity in cones and rods (Khani et al., 2007; Tan et al., 2009; Boye et al., 2010, 2011; Pawlyk et al., 2010; Sun et al., 2010). Taken together, these results support the use of subretinally delivered AAV5-hGRK1*-GUCY2D* in a clinical setting.

Because LCA1 patients exhibit profound cone-mediated visual impairment and variable levels of rod-mediated vision, typical measures of vision must be supplemented with assays that are more suited to this severe disease. Most patients have nystagmus, making full field ERG recordings difficult. Additionally, because many LCA1 patients have unrecordable ERGs to begin with, it would be impossible to measure any negative treatment-associated changes with this assay. Tests which control for nystagmus such as microperimetry can be used to assess cone-mediated central vision (Cideciyan et al., 2012). Tests such as chromatic FST which was used in the most recent clinical characterization (Jacobson et al., 2013) and in the clinical trial for RPE65-LCA2 (Jacobson et al., 2012) could be used to measure localized improvements in rod function. The limits of variability for this assay have already been defined in patients with various retinal degenerations, including LCA (Roman et al., 2005; Roman et al., 2007; Jacobson et al., 2009a). OCT should be performed to monitor the safety of injections performed under the fovea or para/peri- fovea, with emphasis placed on the former given the slow return of cone outer segment structure observed in LCA2 patients treated subfoveally (Jacobson et al., 2012). The potential effects of amblyopia must also be considered. It is possible that AAV-mediated expression of wild type retGC1 in cones of LCA1 patients will reset their signaling cascade allowing for electrochemical signals to propagate through the retina, but that these signals will not be effectively relayed to the brain. Magnetic resonance imaging studies could be used to confirm the integrity of higher order visual processing pathways prior to treatment. Fully understanding how the brain of an LCA1 patient will accommodate *GUCY2D* gene replacement and whether this will differ depending the patient's mutation and/or level of cone preservation will only be accomplished through a clinical trial.

## **CONCLUSION**

Information gleaned from proof of concept experiments in multiple animal models of retGC1 deficiency have laid the groundwork for development of an AAV-based treatment for *GUCY2D*-LCA1 (Haire et al., 2006; Williams et al., 2006; Boye et al., 2010, 2011; Mihelec et al., 2011). Thorough clinical characterization of patients has provided new insights into the pathophysiology of this severe early onset inherited retinal disease, pointing to the central retina as the target for treatment and highlighting the importance of selecting appropriate outcome measures by which to score therapeutic efficacy in a clinical trial (Jacobson et al., 2013). Studies in NHP have revealed an optimal AAV serotype (AAV5) and promoter (hGRK1) combination for delivery of therapeutic transgene to all cone subclasses and rods (Boye et al., 2012). The majority of genes (13/19) known to account for some form of

LCA, including *GUCY2D*, encode photoreceptor specific proteins (den Hollander et al., 2008; Wang et al., 2009; Estrada-Cuzcano et al., 2011; Sergouniotis et al., 2011; Abu-Safieh et al., 2013). In fact, inherited retinal diseases as a whole are caused primarily by defects in proteins expressed by photoreceptors (Wright et al., 2010), a cell type yet to be targeted by AAV in a clinical setting. Thus, developing a treatment for *GUCY2D*-LCA1 will serve as a framework for other photoreceptor-targeted gene replacement strategies.

#### **ACKNOWLEDGMENT**

I would like to thank Sanford L. Boye for his thoughtful critique of this review.

#### **REFERENCES**


caused by RPE65 mutations. *Invest. Ophthalmol. Vis. Sci*. 50, 2368–2375. doi: 10.1167/iovs.08-2696


in a mouse model of autosomal dominant retinitis pigmentosa. *Hum. Gene Ther.* 22, 567–575. doi: 10.1089/hum.2010.140


**Conflict of Interest Statement:** Shannon E. Boye is a co-inventor on US patent # 61/327,521 which covers some aspects of the material discussed within.

*Received: 30 January 2014; accepted: 26 April 2014; published online: 14 May 2014. Citation: Boye SE (2014) Insights gained from gene therapy in animal models of retGC1 deficiency. Front. Mol. Neurosci. 7:43. doi: 10.3389/fnmol.2014.00043*

*This article was submitted to the journal Frontiers in Molecular Neuroscience. Copyright © 2014 Boye. 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.*

## RNA interference gene therapy in dominant retinitis pigmentosa and cone-rod dystrophy mouse models caused by GCAP1 mutations

## *Li Jiang1\*, Jeanne M. Frederick1 andWolfgang Baehr 1,2,3 \**

<sup>1</sup> Department of Ophthalmology and Visual Sciences, John A. Moran Eye Center, University of Utah Health Science Center, Salt Lake City, UT, USA

<sup>2</sup> Department of Biology, University of Utah, Salt Lake City, UT, USA

<sup>3</sup> Department of Neurobiology and Anatomy, University of Utah Health Science Center, Salt Lake City, UT, USA

#### *Edited by:*

Rameshwar K. Sharma, Salus University, USA

#### *Reviewed by:*

Karl-Wilhelm Koch, Carl von Ossietzky University Oldenburg, Germany James Ames, University of California, Davis, USA Alexander Dizhoor, Salus University,

USA

#### *\*Correspondence:*

Li Jiang and Wolfgang Baehr, Department of Ophthalmology and Visual Sciences, John A. Moran Eye Center, University of Utah Health Science Center, 65 Mario Capecchi Drive, Salt Lake City, UT 84132, USA e-mail: li.jiang@hsc.utah.edu; wbaehr@hsc.utah.edu

RNA interference (RNAi) knockdown is an efficacious therapeutic strategy for silencing genes causative for dominant retinal dystrophies.To test this, we used self-complementary (sc) AAV2/8 vector to develop an RNAi-based therapy in two dominant retinal degeneration mouse models. The allele-specific model expresses transgenic bovine GCAP1(Y99C) establishing a rapid RP-like phenotype, whereas the nonallele-specific model expresses mouse GCAP1(L151F) producing a slowly progressing cone-rod dystrophy (CORD). The late onset GCAP1(L151F)-CORD mimics the dystrophy observed in human GCAP1-CORD patients. Subretinal injection of scAAV2/8 carrying shRNA expression cassettes specific for bovine or mouse guanylate cyclase-activating protein 1 (GCAP1) showed strong expression at 1 week post-injection. In both allele-specific [GCAP1(Y99C)-RP] and nonallelespecific [GCAP1(L151F)-CORD] models of dominant retinal dystrophy, RNAi-mediated gene silencing enhanced photoreceptor survival, delayed onset of degeneration and improved visual function. Such results provide a "proof of concept" toward effective RNAi-based gene therapy mediated by scAAV2/8 for dominant retinal disease based on GCAP1 mutation. Further, nonallele-specific RNAi knockdown of GCAP1 may prove generally applicable toward the rescue of any human GCAP1-based dominant cone-rod dystrophy.

**Keywords: photoreceptor guanylate cyclase, guanylate cyclase-activating protein 1, short-hairpin RNA, RNA interference, self-complementary adeno-associated virus, cone-rod dystrophy,** *retinitis pigmentosa*

## **INTRODUCTION**

Cone-rod dystrophies constitute a rare (1/40,000 prevalence) and heterogeneous class of hereditary retinal disease (Hamel, 2007). Symptoms of cone-rod dystrophy (CORD) may include photoaversion, attenuation of central visual acuity, achromatopsia, and eventually, extinction of peripheral vision. To date, ten genes are associated with dominant CORD:*PROM1* (Prominin-1)*, PRPH2* (Peripherin/rds)*, GUCA1A* (GCAP1)*, RIMS1* (Regulating Synaptic Membrane Exocytosis)*, GUCY2D* (Guanylate Cyclase 1)*, AIPL1* (Arylhydrocarbon-Interacting receptor Protein-Like 1)*, PITPNM3* (Phosphatidyl Inositol Transfer Membraneassociated family member 3)*, UNC119* (Uncoordinated 119 or HRG4)*, CRX* (Cone-Rod otX-like photoreceptor homeobox transcription factor), *and SEMA4A* (Semaphorin 4A; RETNET at https://sph.uth.edu/retnet/disease.htm). *GUCA1A,* encoding guanylate cyclase-activating protein 1 (GCAP1), is one of the most fully-characterized dominant CORD genes (Baehr and Palczewski, 2009) and involves about one dozen families with >100 affected members harboring various *GUCA1A* mutations (Jiang et al., 2011).

Guanylate cyclase-activating protein 1 plays a key role in inhibiting photoreceptor guanylate cyclase activity at high free Ca2+, and accelerating guanylate cyclase activity in low free Ca2+. Hydrolysis of cyclic guanosine monophosphate (cGMP) by the phototransduction cascade closes cGMP-gated channels, reducing

influx of Ca2<sup>+</sup> ions (**Figure 1**). Continuous extrusion of Ca2<sup>+</sup> by the light-insensitive NCKX exchanger lowers cytoplasmic Ca2+, thereby activating GCAP1, and GC and re-establishing dark cGMP levels. In rods, cGMP levels are regulated by two guanylate cyclases (GC1 and GC2; Goraczniak et al., 1994, 1997; Lowe et al., 1995; Duda et al., 1996) and two GCAPs (GCAP1 and GCAP2; Dizhoor et al., 1994, 1995; Palczewski et al., 1994; Gorczyca et al., 1995), while cone phototransduction relies on GC1 and GCAP1 exclusively. The two GCAPs overlap partially in regulating the GCs of rods (Makino et al., 2008; Dizhoor et al., 2010; Peshenko et al., 2011), but with differential signaling modes (Duda et al., 2005, 2012). Both GCAPs contribute to rod recovery after photolysis (Mendez et al., 2001; Howes et al., 2002; Makino et al., 2008, 2012). Germline deletion of both GCAPs renders GC activity in rods and cones Ca2+-insensitive (Mendez et al., 2001). Transgenic GCAP1 could restore normal rod and cone response recovery (Howes et al., 2002; Pennesi et al., 2003).

GCAPs feature four EF hand motifs, of which three (EF2-4) are high-affinity Ca2<sup>+</sup> binding sites and one (EF1) is inactive (Palczewski et al., 2004; Baehr and Palczewski, 2009). Six missense mutations in GCAP1 associated with autosomal dominant CORD3 are found in EF3 (E89K, Y99C, D100E, D100G, N104K, I107T; Dizhoor et al., 1998; Payne et al., 1998; Sokal et al., 1998; Jiang et al., 2008; Kitiratschky et al., 2009; Kamenarova et al., 2013; Nong et al., 2014). In EF4, only four missense

**cGMP and Ca2+.** Dark levels of cGMP and Ca2<sup>+</sup> are high in rod and cone photoreceptors. Light activation of rhodopsin initiates rod phototransduction, activated PDE6 rapidly hydrolyzes cytoplasmic cGMP, and cGMP-gated cation channels close. Continued extrusion of Ca2<sup>+</sup> by the light-insensitive Na+-K+/Ca2<sup>+</sup> exchanger (NCKX) lowers free Ca2<sup>+</sup> which activates guanylate cyclase-activating proteins (GCAPs) and guanylate cyclases (GCs). Restoration of cGMP dark levels re-opens cation channels and Ca2<sup>+</sup> levels equilibrate to dark levels.

mutations have been identified (I143NT, L151F, E155G, E155A, G159V; Wilkie et al., 2001; Nishiguchi et al., 2004; Jiang et al., 2005; Sokal et al., 2005; Huang et al., 2013; **Figure 2**). These dominant GCAP1 mutations alter Ca2+-association, decrease Ca2<sup>+</sup> sensitivity, and produce constitutive activity of photoreceptor GC1 at normal "dark" Ca2<sup>+</sup> levels. Persistent stimulation of GC1 in the dark increases cGMP to toxic levels. (Dizhoor et al., 1998; Sokal et al., 1998; Woodruff et al., 2007). Elevated

**FIGURE 2 | Structure of myristoylated GCAP1 (PDB 2R2I; Baehr and Palczewski, 2009).** N, N-terminal. C, C-terminal. EF hand helix-loop-helix structures are shown: EF1 (inactive and gray), EF2 (red), EF3 (dark blue), and EF4 (turquoise). Ca2+-binding loops (green) and approximate positions of relevant missense mutations associated with adCORD (red arrows) are indicated. The N-terminal myristoyl group (orange–brown) is buried and nearly invisible.

levels of cGMP open more CNG channels, elevate free Ca2<sup>+</sup> in outer segments, and trigger cell death by unknown mechanisms. Elevated Ca2<sup>+</sup> levels have been suspected to trigger cell death, however a massive accumulation of cGMP in *Cnga3*−/−;*Nrl*−/<sup>−</sup> mice lacking cation channels and elevated Ca2<sup>+</sup> also correlated with photoreceptor apoptotic death. This result excludes Ca2<sup>+</sup> as a death trigger and supports a role of cGMP accumulation as the major contributor to cone death and a role cGMPdependent protein kinase G (PKG) regulation in cell death (Xu et al., 2013).

In a four-generation British family, CORD was mapped to chromosome 6p21.1 (Payne et al., 1998). The disease-causing mutation was identified as Y99C, a change that was absent in over 200 unrelated controls. The same mutation was later identified independently in two ancestrally related families (Downes et al., 2001). The Y99C mutation in GCAP1 has also been reported to cause isolated macular dysfunction (Michaelides et al., 2005). Thus far, families presenting with cone and CORD have been independently linked to a GCAP1(L151F) mutation. In the first family (Sokal et al., 2005), hemeralopia, dyschromatopsia and reduced visual acuity became evident by the second-to-third decades of life with non-recordable photopic ERG responses. In the second pedigree spanning five-generations (Jiang et al., 2005), 11 of 24 individuals displayed photoaversion, color vision defects, and central acuity loss with onset of legal blindness during the second-to-third decades of life. The GCAP1(L151F) gene product, known to disrupt Ca2<sup>+</sup> coordination at EF-hand four, and alter Ca2<sup>+</sup> sensitivity (Sokal et al., 2005), represents a conservative substitution.

#### **GCAP1 MUTANT MOUSE MODELS**

To study GCAP1-associated retinal degeneration disease, several GCAP1 mouse models have been generated, expressing different GCAP1 mutants. The first generated GCAP1 transgenic mice expressing bovine GCAP1(Y99C) mutant gene displayed severe *retinitis pigmentosa*-like phenotypes as the mutant transgene was specifically expressed in rod photoreceptors under control of a rhodopsin promoter (Olshevskaya et al., 2004). A GCAP1(E155G) knock-in mouse could mimick human patients and presented lateonset and slowly progressive cone-rod photoreceptor degeneration (Buch et al., 2011). Expressing GCAP1(E155G) in mouse rods also caused severe early-onset rod-cone degeneration (Woodruff et al., 2007). Our laboratory generated three transgenic mouse lines expressing wild-type GCAP1-EGFP, as well as mutant GCAP1(L151F), and GCAP1(L151F)-EGFP (Jiang et al., 2013; **Figure 3**). The three transgenes were modified from mouse genomic *Guca1a* gene fragment, containing its native regulatory elements. They are expressed at a level comparable with endogenous GCAP1 in the transgenic mice, and the mutant transgenic mice develop retina pathology slowly and recapitulate features of human CORD (Jiang et al., 2013).

Our aim was to develop first a self-complementary AAV2/8 (scAAV2/8) knockdown virus that expresses an allele-specific short hairpin RNA (shRNA) targeting the mutant *Guca1a* gene in the GCAP1(Y99C) transgenic mice. Further, we intended to target both mutant and native *Guca1a* in GCAP1(L151F)- EGFP transgenic mice (nonallele-specific shRNA targeting). In

the first set, we used two transgenic mouse lines expressing bGCAP1(Y99C) cDNA under the control of the rhodopsin promoter (**Figure 3**). One line expressed GCAP1 at near normal levels and the second line overexpressed GCAP1 (3–4X; Olshevskaya et al., 2004). These lines are models for dominant *retinitis pigmentosa* representing moderate (line L52H) or severe (line L53) retinal degeneration phenotypes (Olshevskaya et al., 2004). An allele-specific knockdown bGCAP1(Y99C) is expected to significantly slow down the dystrophy and even delay onset of degeneration.

The second set of transgenic mice contained the L151F mutation introduced in exon 4 of the *Guca1a* gene. We employed the entire *Guca1a* gene including promoter, all exons and introns, and the 3- -UTR containing the polyadenylation signals. We established two mutant GCAP1 mouse lines, GCAP1(L151F), and GCAP1(L151F)-GFP (**Figure 3**). A third line expressing a GCAP1- EGFP fusion protein (no L151 mutation) served as a WT control (**Figure 3**). The EGFP tag allowed detection of transgene expression by live fluorescence microscopy, and distinction of transgenic GCAP1-EGFP (50 kDa) from native GCAP1 (23 kDa). We

of the CMV promoter. Hp consists of a sense-loop-antisense construct

et al., 2013).

**Figure 2** (Jiang et al., 2011), and **Figure 7** (Jiang et al., 2013).

intended to knockdown both WT and mutant GCAP1s as removal of photoreceptor GCAP1 does not affect retina development or morphology, and GCAP1/2 double knockout mice do not exhibit retina degeneration.

#### *IN VITRO* **SCREENING FOR POTENT shRNA SEQUENCES**

For successful knockdown via shRNA, it is essential to screen candidate knockdown sequences for specificity and efficacy. We developed an *in vitro* screening strategy consisting of a shRNA expression vector with reporter gene, to be co-expressed with a second vector expressing GCAP1-GFP as a prey (**Figures 4A,B**). This *in vitro* screening system permits the testing of a number of candidates and eventual identification of the most powerful shRNA. The shRNA vector, hH1pro, contained an shRNA expression cassette driven by a human H1 pol-III promoter, and a CMV-driven mCherry reporter gene arranged in a tail-to-tail array (**Figure 4B**). For allelespecific targeting we generated four shRNA constructs, bovine (b) G1hp1-4, expressing four candidate anti-bovine GCAP1 shRNAs specifically targeting b*Guca1a* mRNA (**Figure 4C**). Similarly, we generated four anti-mouse GCAP1 shRNA expression constructs, mG1hp1-4, for nonallele-specific targeting of both transgenic and endogenous mouse GCAP1 mRNAs. (**Figure 4D**).

In HEK293 cell culture, we tested the knockdown efficiencies of four putative bGCAP1 shRNA by cotransfection of each expression construct (bG1hp1-4) with the target bGCAP1-EGFP (**Figure 5A**). The expression levels of bGCAP1-EGFP in the transfected cells were detected by live cell fluorescence microscopy and immunoblotting assay 48–50 h after cotransfection. Under a similar expression level indicating by the mCherry reporter (**Figure 5A**, lower panel), four bGCAP1 shRNA candidates suppressed GCAP1-GFP expression at different efficiencies (**Figure 5A**, upper panel). Among them, bG1hp4 was shown the highest knockdown efficiency, presented by the lowest GFP fluorescence in the cotransfected cell cultures, and 79% reduction of GCAP1-GFP compared to 68, 47, and 58% for other three shRNAs in semi-quantitative immunochemistry assay (**Figure 5B**). By using the same *in vitro*

**FIGURE 6 |Transduction efficiencies of four gene delivery methods for mouse photoreceptors tested by expression of fluorescent protein reporter. (A)**, Fluorescence microscopy of retina transversesections of mice injected subretinally with pCMV-Zeo-GFP, peptide/polymer nanoparticles. Below, retina section probed with anti-GFP antibody. **(B)**, Fluorescence microscopy of retina transfected

with pCAG-GFP using recombinant adenovirus 5, Ad5-RGD. Below, section at higher magnification. **(C)**, pCMV-GFP-transfected retina using recombinant AAV vector, AAV2. Underneath, section at higher magnification. **(D)**, pCMV-mCherry-transfected retina using electroporation. Micrographs were recorded at various days post-injection (PI).

screening experiment, we demonstrated that among four antimouse GCAP1 shRNAs (mG1hp1-4), mG1hp4 is the most efficient, as it deceased mGCAP1-GFP expression by more than 70% in cell culture. Knockdown specificities of bG1hp4 and mG1hp4 were tested with their 2-nucleotide mismatch shRNA controls, bG1hp4m2 and mG1hp4m2. Compared to bG1hp4 and mG1hp4, the respective mismatch controls were unable to suppress expression of co-transfected bGCAP1-GFP and mGCAP1-GFP (**Figure 4**).

## **GENERATION OF SELF-COMPLEMENTARY ADENO-ASSOCIATED VIRUS**

To efficiently express the shRNAs in mouse photoreceptors, we tested four currently used gene delivery vectors: a peptide/polymer nanoparticle (CK30PEG; Farjo et al., 2006), recombinant adenovirus (Ad5-RGD; Cashman et al., 2007; Sweigard et al., 2010), recombinant AAV2 (Wu et al., 2006), and electroporation with naked DNA. No GFP reporter gene expression was detected in the CK30PEG-transduced retina (**Figure 6A**), and the Ad5-RGDtransduced retina had much more robust GFP expression in RPE cells than in photoreceptors (**Figure 6B**). The AAV2-transduced retina showed the most robust GFP expression in photoreceptors (**Figure 6C**), suggesting AAV vector is a highly efficient gene vector for shRNA delivery to mouse photoreceptors. Although neonatal electroporation of subretinally injected plasmid DNA showed specific and efficient photoreceptor transfection (**Figure 6D**), this method is only useful for undifferentiated and mitotic photoreceptors and is therefore, not applicable for human gene therapy.

Among currently available AAV vectors, the scAAV vector, scAAV2/8, shows the most effective and stable transgene expression in mouse photoreceptors (Natkunarajah et al., 2008). We packaged the shRNA expression cassette with an mCherry reporter into the recombinant viral vector to generate two scAAV2/8 viral constructs expressing shRNAs targeting bovine and mouse GCAP1s, respectively, scAAVbG1hp4 and scAAVmG1hp4. By examining the mCherry reporter expressed in the scAAVbG1hp4 transduced mouse retinas, we demonstrated that scAAV2/8 vector could generate a long-term transgene expression lasting to 1 year (**Figure 7**).

## **THERAPEUTIC EFFECTS OF KNOCKDOWN VIRUS**

To test allele-specific knockdown of GCAP1 *in vivo*, we injected the scAAVbG1hp4 virus into the mouse subretinal space using two different bGCAP1(Y99C) transgenic mouse lines, L53, and L52H. Expression of the mutant bGCAP1(Y99C), but not that of endogenous mouse GCAP1, was significantly suppressed by scAAV2/8 mediated bG1hp4 expression in the retinas of transgenic mice. In the severe and rapid retinal degeneration mouse line L53, bG1hp4 significantly delayed photoreceptor cell death, which was observed at 30 and 45 days post-injection. With the moderate retinal degeneration line L52H, we demonstrated a long-term therapeutic effect of scAAVbG1hp4 virus from 1 month up to 11 months postinjection, assayed by retinal morphology (**Figure 8A**) and function (**Figures 8B,C**).

As deletion of GCAPs in mouse apparently has no detrimental morphological nor disease-causing defect, except for increased sensitivity and delayed recovery from the dark state (Mendez

**FIGURE 8 | Long-term therapeutic efficiency of allele-specific shRNA vector, scAAV2/8-bG1hp4. (A)**, Direct fluorescence microscopy of retinal cross-sections examines retinal morphology of the bG1hp4 treated and untreated L52H transgenic mice. Viral vectors were injected subretinally at mouse ages P21–P30. Both treated and untreated eyes were harvested at four representative times from 1–11 months post-injection. Red, mCherry expression demonstrating scAAV2/8 virus transduction. Blue, DAPI staining of nuclei. Note significant preservation of ONL thickness at 11 months post-treatment (∼12 months of age) compared to non-treated controls. **(B,C)**. Scotopic and photopic ERG amplitudes recorded from bG1hp4-treated (red) and untreated (blue) L52H transgenic mice (Jiang et al., 2011). Subretinal injection of scAAV2/8-bG1hp4 in the transgenic mouse models delayed progression of both rod and cone dysfunction significantly.

**vector, scAAV2/8-mG1hp4. (A)**, Fluorescence fundus images of mG1(L151F)-GFP transgenic mice with subretinal injection of the virus vectors at P21–P30 (left) and non-injected control (right). GFP signal (top) represents the mG1(L151F)-GFP expression level in photoreceptors, and mCherry (bottom) signal indicates scAAV2/8 virus transduction.

**(B)**, Immunoblot of the transgenic mG1(L151F)-GFP and endogenous GCAP1 protein levels in the injected and the non-injected retinas at 30 days PI. Transgenic mG1(L151F)-GFP (∼50 kD) and endogenous GCAP1 (∼25 kD) proteins were detected by UW101 antibody directed against GCAP1. -actin served as endogenous loading control. See also **Figure 9** (Jiang et al., 2013).

et al., 2001), we assumed that nonallele-specific shRNA knockdown of both wild-type and mutant GCAP1s may be a general therapeutic strategy to treat retinal degeneration in patients carrying GCAP1 mutations. We tested *in vivo* knockdown efficiency of scAAV2/8-mediated mG1hp4 in the mGCAP1(L151F)-GFP transgenic mice. Fluorescence fundus imaging indicated significant suppression of the mGCAP1(L151F)-GFP mutant transgene, but not in uninjected eyes (**Figure 9A**). Broad and robust mCherry reporter gene expression in scAAVmG1hp4-treated mouse retinas signaled even expression levels of the virus. Semi-quantitative analysis of GCAP1 levels in treated retinas revealed that mG1hp4 simultaneously knocked down *in vivo* expression of the mutant mGCAP1(L151F)-GFP transgene by 70% and endogenous wildtype GCAP1 by 90% (**Figure 9B**). Thus our experiment provides "proof of principle" that nonallele-specific RNA interference (RNAi) knockdown may be a strategy applicable to all GCAP1 mutations.

#### **CONCLUSION**

The experimental goal was to test whether allele-specific or nonallele-specific knockdown of a dominant GCAP1 mutant is able to ameliorate photoreceptor dystrophy. We demonstrated the feasibility of shRNA knockdown first with an allele-specific approach in a *retinitis pigmentosa* mouse model expressing GCAP1(Y99C). A scAAV robustly expressed shRNAs in photoreceptors at 1 week post-injection and gene silencing activity persisted as long as 1 year without any apparent off-target interference. Delayed disease onset, significantly improved rod photoreceptor survival and increased visual function support that the methodology can be useful for human gene therapy.

For nonallele-specific knockdown, we generated a sophisticated set of transgenic mouse models expressing GCAP1-EGFP fusion proteins with and without L151F mutation. The L151F mutation, discovered in our lab, was shown to cause dystrophy in two unrelated families. Nonallele-specific shRNA knockdown of both wild-type and mutant GCAP1s may serve as a therapeutic strategy to rescue the dominant degeneration caused by any of the eleven known EF-hand GCAP1 mutations. An advantage of dominant GCAP1 mutations is that a nonalle-specific approach promises to be successful, while mutations in other CORD genes require gene replacement to rescue the disease. Successful knockdown by RNAi suppression of both wild-type and mutant GCAP1s may be a potent therapeutic strategy, applicable to all affected family members with CORD based on GCAP1 mutations in EF3 and EF4, as long as the shRNA guide strand is located external to the disease-causing mutations.

#### **REFERENCES**


pigmentosa mouse model. *Proc. Natl. Acad. Sci. U.S.A.* 108, 18476–18481. doi: 10.1073/pnas.1112758108


**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: 21 January 2014; accepted: 19 March 2014; published online: 07 April 2014. Citation: Jiang L, Frederick JM and Baehr W (2014) RNA interference gene therapy in dominant retinitis pigmentosa and cone-rod dystrophy mouse models caused by GCAP1 mutations. Front. Mol. Neurosci. 7:25. doi: 10.3389/fnmol.2014.00025 This article was submitted to the journal Frontiers in Molecular Neuroscience.*

*Copyright © 2014 Jiang, Frederick and Baehr. 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.*

## Regulation of intraocular pressure by soluble and membrane guanylate cyclases and their role in glaucoma

#### *Emmanuel S. Buys <sup>1</sup> \*, Lincoln R. Potter 2, Louis R. Pasquale3 and Bruce R. Ksander <sup>4</sup>*

*<sup>1</sup> Department of Anesthesia, Critical Care, and Pain Medicine, Anesthesia Center for Critical Care Research, Harvard Medical School, Massachusetts General Hospital, Boston, MA, USA*

*<sup>2</sup> Department of Pharmacology, University of Minnesota Medical School, Minneapolis, MN, USA*

*<sup>3</sup> Department of Ophthalmology, Glaucoma Service Mass Eye and Ear Infirmary and Channing Division of Network Medicine, Harvard Medical School, Brigham and Women's Hospital, Boston, MA, USA*

*<sup>4</sup> Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, Schepens Eye Research Institute, Harvard Medical School, Boston, MA, USA*

#### *Edited by:*

*Clint Lawrence Makino, Harvard Medical School, USA*

#### *Reviewed by:*

*Bin-Nan Wu, Kaohsiung Medical University, Taiwan Alan Marmorstein, Mayo Clinic, USA*

#### *\*Correspondence:*

*Emmanuel S. Buys, Department of Anesthesia, Critical Care, and Pain Medicine, Anesthesia Center for Critical Care Research, Harvard Medical School, Massachusetts General Hospital, 55 Fruit Street, Thier 511B, Boston, MA 02114, USA e-mail: ebuys@mgh.harvard.edu*

Glaucoma is a progressive optic neuropathy characterized by visual field defects that ultimately lead to irreversible blindness (Alward, 2000; Anderson et al., 2006). By the year 2020, an estimated 80 million people will have glaucoma, 11 million of which will be bilaterally blind. Primary open-angle glaucoma (POAG) is the most common type of glaucoma. Elevated intraocular pressure (IOP) is currently the only risk factor amenable to treatment. How IOP is regulated and can be modulated remains a topic of active investigation. Available therapies, mostly geared toward lowering IOP, offer incomplete protection, and POAG often goes undetected until irreparable damage has been done, highlighting the need for novel therapeutic approaches, drug targets, and biomarkers (Heijl et al., 2002; Quigley, 2011). In this review, the role of soluble (nitric oxide (NO)-activated) and membrane-bound, natriuretic peptide (NP)-activated guanylate cyclases that generate the secondary signaling molecule cyclic guanosine monophosphate (cGMP) in the regulation of IOP and in the pathophysiology of POAG will be discussed.

**Keywords: guanylate cyclase, nitric oxide, natriuretic peptides, glaucoma, open-angle, intraocular pressure**

### **PRIMARY OPEN ANGLE GLAUCOMA**

Primary open-angle glaucoma (POAG) is a leading cause of blindness, affecting over 2.2 million patients in the US alone, and is associated with an estimated health care cost upwards of \$1.5 billion/year. Vision loss, occurring due to loss of retinal ganglion cells (RGCs) and degeneration of the optic nerve, has far-reaching effects on the patient's ability to function independently, dramatically affecting quality of life from a physical, mental and social well-being perspective. Currently, there is no definitive cure for POAG and although multiple risk factors for POAG have been identified [including intra-ocular pressure (IOP), race, age, and genetic factors], the molecular signaling involved in POAG pathogenesis remains largely unknown.

Although multiple POAG risk factors have been identified, the etiology of POAG remains to be elucidated, likely because the disease can be stratified into various subtypes defined by discrete but yet unknown biochemical pathways. Two major pathophysiologic mechanisms for POAG have been proposed. In the "mechanical theory" optic neuropathy is caused by increased IOP (see "Intraocular pressure: a risk factor for POAG" below). Alternatively, a vascular component has been hypothesized to contribute to POAG pathophysiology (see "Glaucoma and vascular dysfunction" below). However, the extent to which vascular dysfunction contributes to glaucomatous optic neuropathy remains to be elucidated and is controversial (Vajaranant and Pasquale, 2012).

## **INTRAOCULAR PRESSURE: A RISK FACTOR FOR POAG**

POAG is often associated with elevated IOP. Measurement of IOP typically requires an anesthetized ocular surface and a cooperative patient, as corneal applanation is needed for accurate measurement of this ophthalmic vital sign. While POAG is a strongly age related disease (Sommer et al., 1991; Mukesh et al., 2002) and elevated IOP is a major risk factor for this condition, IOP does not necessarily increase with age. One longitudinal analysis found that among people aged 50–59, IOP increased after 9 years of follow-up but among subjects 60 and older, a slight decrease in IOP was noted in the same time period (Wu et al., 2006). Furthermore while African heritage is an undisputed risk factor for POAG, people of African heritage do not necessarily have higher IOP than Caucasian subjects drawn from the same population (Sommer et al., 1991). While the heritability of IOP is relative highly (56–64%) (Carbonaro et al., 2009) very few common loci for this trait have been discovered thus far (Van Koolwijk et al., 2012; Ozel et al., 2014). Other factors such as body mass index (Klein et al., 1992; Wu and Leske, 1997; Oh et al., 2005), blood pressure (Klein et al., 2005), and diabetes (Wu et al., 2006) have only modest positive associations with IOP. Data regarding how IOP might change while blood pressure varies in an individual patient are lacking. When a patient (regardless of glaucoma status) transitions between a seated to supine position there is a predictable increase in IOP of ∼3 mm Hg (Lee et al., 2012, 2013). Only a few medical conditions impact the level of IOP (Arevalo et al., 1996; Garcia Filho et al., 2011) but steroid exposure is notorious for leading to elevated IOP via changes in trabecular meshwork permeability (Clark et al., 1995). It is thought that IOP fluctuates more in POAG patients than controls (Sacca et al., 1998) but the literature is conflicting regarding the role of IOP fluctuation on open-angle glaucoma disease progression (Nouri-Mahdavi et al., 2004; Bengtsson et al., 2007).

IOP is determined by the balance between the production/secretion of aqueous humor (AqH) by the ciliary processes and by the drainage of AqH via the iridocorneal angle. AqH can exit the eye through various routes: the conventional, the uveoscleral, and the uveolymphatic pathways. The conventional pathway consists of the trabecular meshwork (TM), Schlemm's canal, collecting channels, and the episcleral venous system. It is generally accepted that IOP reflects the pressure necessary to overcome the intrinsic resistance to aqueous outflow that occurs at the junction where the juxtacanalicular part of the TM meets the inner wall of Schlemm's canal (Ethier et al., 1986). In uveoscleral outflow, AqH drains through the ciliary muscle (CM) and exits through the supraciliary space and across the anterior or posterior sclera, into choroidal vessels (Fautsch and Johnson, 2006). Uveoscleral outflow, accounting for anywhere between 3% and 80% of outflow depending on the species studied (4–60% in humans) (Fautsch and Johnson, 2006), is particularly impacted by age, an important observation considering the age-dependency of POAG in both patients (Kass et al., 2002, 2010; Mukesh et al., 2002; De Voogd et al., 2005; Rudnicka et al., 2006; Leske et al., 2008) and animal models (Buys et al., 2013). Incidentally, also in animal models of secondary angle-closure glaucoma, IOP and prevalence of optic neuropathy and retinal lesions increased with age (John et al., 1998; Saleh et al., 2007). More recently, the existence of a third outflow route was postulated: lymphatic channels in the stroma of the ciliary body and interstitial spaces between CM muscle bundles may function as a backup outflow system (Yucel et al., 2009). The relevance of this uveolymphatic pathway and whether NO-cGMP signaling (see "NO-cGMP signaling in the eye" below) modulates contractility of ocular lymphatic vessels remains to be determined. Importantly, a central role for NO in lymphatic function was recently identified (Liao et al., 2011). Whether the effect of NO on lymphatic contractions is mediated by cGMP remains unknown.

In contrast to what is observed in angle closure glaucoma, POAG is not associated with apparent blockage of the anterior chamber angle. In POAG, there is variable elevation of IOP associated with impaired AqH outflow that occurs despite apparently normal anterior segment anatomy and an open iridocorneal angle (Weinreb and Khaw, 2004). IOP rises gradually over time, likely as a consequence of decreased drainage of AqH. IOP-lowering treatment significantly cuts the risk of developing glaucoma in ocular hypertensives (Kass et al., 2002, 2010).

### **GLAUCOMA AND VASCULAR DYSFUNCTION**

While IOP reduction continues to be a successful treatment to reduce the progression of POAG (Leske et al., 2003; Kass et al., 2010), the pathogenesis of POAG seemingly also depends on factors other than increased IOP. Compounds that do not lower IOP dramatically may have properties that address the underlying glaucomatous disease process and therefore could be suitable therapeutic agents (Weinreb and Kaufman, 2009; Chen et al., 2011). For example, brimonidine was superior to timolol in stabilizing visual field deterioration, despite producing a similar IOP-lowering effect (Krupin et al., 2011). Also, ocular hypertension does not necessarily lead to POAG and ocular normotension does not preclude the development of POAG (Leske et al., 2001), suggesting that other pathologies, including neurologic (similar to other chronic central nervous system diseases such as Alzheimer's diseases or Multiple Sclerosis Quigley, 2011) or vascular dysfunction [e.g., as characterized by impaired retinovascular autoregulation (Feke and Pasquale, 2008) or peripheral vascular endothelial dysfunction (Henry et al., 1999; Su et al., 2008)], may contribute to the etiology of POAG. Vascular dysfunction contributes to the development of systemic hypertension (Mendelsohn, 2005; Michael et al., 2008) and several studies have provided evidence that blood pressure affects POAG risk (Kaiser and Flammer, 1991; Tielsch et al., 1995; Hulsman et al., 2007; Leske et al., 2008; Memarzadeh et al., 2010; Cherecheanu et al., 2013). In addition, the idea that vascular dysfunction contributes to the pathogenesis of POAG is based on the hypothesis that decreased perfusion of the optic nerve leads to neurodegeneration (Harris et al., 2005; Vajaranant and Pasquale, 2012).

Both impaired blood flow and impaired vascular autoregulation have been described in glaucoma patients (Flammer et al., 2002; Feke and Pasquale, 2008; Moore et al., 2008; Feke et al., 2011). Large epidemiologic studies have found adverse associations between POAG and low ocular perfusion pressure (OPP) (Hulsman et al., 2007; Leske, 2009) or low systemic blood pressure (Bonomi et al., 2000). In addition, the Early Manifest Glaucoma Trial observed that predictors of POAG disease progression included lower systolic blood and perfusion pressure (Leske et al., 2007). Especially in normal tension glaucoma patients, systemic vascular abnormalities have been described, including impaired flow- or acetylcholine-mediated vasodilation (Gasser and Flammer, 1991; Henry et al., 1999; Su et al., 2008). Also POAG patients with early paracentral visual field loss are more likely to have systemic vascular risk factors such as hypotension and migraines (Park et al., 2011, 2012). It is important to note that blood flow regulation in the optic nerve head seems to be strongly dependent on IOP and OPP (Schmidl et al., 2011). The interaction between systemic blood pressure and IOP (defining OPP) (Tielsch et al., 1995; He et al., 2011) is particularly relevant for this review, given the role of NO-cGMP signaling (see "NO-cGMP signaling in the eye" below) in both the regulation of IOP and systemic blood pressure.

### **NO-cGMP SIGNALING IN THE EYE**

NO is an exceptionally well-characterized signaling molecule, with important roles in a wide variety of physiological and pathophysiological processes, including cardiovascular homeostasis, neuronal function, and inflammation. NO is synthesized from L-arginine by a family of three enzymes referred to as NO synthases (NOSs) (**Figure 1**) (Moncada and Higgs, 1993). NOS2, or inducible NOS, was first identified in macrophages but has since been detected in a wide variety of cells exposed to endotoxin and cytokines. Following bacterial infection, especially with gramnegative organisms, high levels of NO are produced by NOS2. NOS1 and NOS3 were initially described to be constitutively expressed in neuronal cells and endothelial cells, respectively

(Moncada and Higgs, 1993). Under physiological conditions, the low levels of NO produced by the two constitutive Ca2+ dependent enzymes NOS1 and NOS3 have diverse functions ranging from neurotransmission and vasodilatation to inhibition of platelet adhesion and aggregation. NO, is an important modulator of smooth muscle function and all three NOS isoforms are expressed in the eye (**Table 1**). Multiple studies have reported on the ability of NO to modulate resistance in the AqH outflow pathway and IOP (see "NO-cGMP-mediated regulation of IOP: implications for POAG" below). Therefore, NO is an attractive candidate as a factor that could modify both mechanical and vascular events in POAG pathogenesis.

NO has numerous targets, reacting with a variety of intracellular and extracellular molecules typically via thiol groups or transition metal centers (Chiamvimonvat et al., 1995; Torres et al., 1995; Davis et al., 2001; Jaffrey et al., 2001; Stamler et al., 2001). A major target of NO is the obligate heterodimer soluble guanylate cyclase (sGC) (Mergia et al., 2003, 2006; Nimmegeers et al., 2007; Vermeersch et al., 2007; Buys et al., 2008), a hemecontaining heterodimeric enzyme, consisting of one α and one β subunit (**Figure 1**). cGMP interacts with a variety of effector proteins including cGMP-dependent protein kinases (PKGs), cGMP-regulated phosphodiesterases (PDE's), and ion channels. cGMP is also synthesized by receptor guanylate cyclases that are activated by NPs (NPRs, **Figure 1**). However, cGMP produced by NPRs and sGC may have differential effects, possibly due to differential spatiotemporal distributions of cGMP produced by the two guanylate cyclase families (Su et al., 2005; Castro et al., 2006; Piggott et al., 2006).

Two functional isoforms of sGC exist: sGCα1β1, the predominant isoform in most tissues (Mergia et al., 2003), and sGCα2β<sup>1</sup> (Hobbs, 1997; Russwurm et al., 1998; Bamberger et al., 2001). sGCα and β subunits were shown to be expressed in several anatomical sites relevant for glaucoma (**Table 1** and **Figure 2**). sGC activity was detected in rabbit (Haberecht et al., 1998), rat (Kajimura et al., 2003), turtle (Blute et al., 1998), and mouse retina (Blom et al., 2012; Buys et al., 2013). sGC is expressed both in retinal ganglion cells and photoreceptors and in the vascular smooth muscle layer of retinal arterioles. Finally, and most relevant to the potential ability of NO-sGC to modulate AqH outflow, sGC is abundantly expressed in isolated human TM cells (Ellis et al., 2009) and in both human and mouse CM (Buys et al., 2013). Also NP-activated membrane bound guanylate cyclase are expressed in various ocular tissues (see "NP receptors and their effects on IOP" below and **Table 1**).

## **NO-cGMP-MEDIATED REGULATION OF IOP: IMPLICATIONS FOR POAG**

Multiple studies have yielded evidence suggesting that NO-cGMP signaling regulates AqH outflow and IOP. For example, both an NO donor and a cGMP-analog decreased IOP (Kotikoski et al., 2002) and increased outflow facility in rabbits (Kotikoski


#### **Table 1 | Ocular localization of nitric oxide synthase (NOS), soluble guanylate cyclase (sGC), and NP receptors (NPR).**

*ONH, optic nerve head; RGC, retinal ganglion cell; IPL, inner plexiform layer; CM, ciliary muscle; TM, trabecular meshwork; SC, Schlemm's canal; ONL, outer nuclear layer; INL, inner nuclear layer; OPL, outer plexiform layer.*

et al., 2003). The NO-induced increase in outflow facility was impaired by treating perfused porcine eyes with the sGC inhibitor 1H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-1 (ODQ), suggesting that the effects of NO on outflow are cGMP-dependent (Ellis et al., 2009). The non-isoform specific NOS-inhibitor L-NG-Nitroarginine methyl ester (L-NAME) decreased and an NO-donor compound increased AqH flow rate in perfused human donor eyes. This increase in flow was associated with an increase in cGMP levels measured in the perfusate, again suggesting a central role for sGC and cGMP in the ability of NO to modulate outflow (Schneemann et al., 2002). Accordingly, NOdonor compounds lower IOP (Nathanson, 1988, 1992; Schuman et al., 1994; Behar-Cohen et al., 1996; Krauss et al., 2011) and enhance tissue oxygenation of the optic nerve head in preclinical animal models (Khoobehi et al., 2011). Overexpressing NOS3 in mice lowers IOP by increasing pressure-dependent drainage (Stamer et al., 2011), confirming previous reports suggesting that the ability of NO to lower IOP is mediated by a decrease in the AqH resistance rather than a change in the rate of AqH secretion (Nathanson and Mckee, 1995b). Other possible mechanisms by which NO and cGMP can modulate outflow include regulation of Schlemm's canal cell volume (Dismuke et al., 2008; Ellis et al., 2010) and TM cell volume (Dismuke et al., 2009).

Impaired NO-cGMP signaling has been implicated in POAG. For example, NO metabolites and cGMP levels were lower in both plasma and AqH samples from POAG patients than in those from individuals without POAG (Galassi et al., 2000, 2004; Doganay et al., 2002). Also, NADPH-diaphorase (NADPH-d) reactivity, a marker for NO production, was decreased in TM, SC, and anterior longitudinal CM fibers isolated from POAG eyes (Nathanson and Mckee, 1995a) and serum levels of L-arginine analogs (endogenous NOS inhibitors) were found to be elevated in patients with advanced glaucoma (Javadiyan et al., 2012). In addition, NOS3 gene variants were associated with POAG in women (see "Genetics of POAG" below Kang et al., 2010; Magalhaes Da Silva et al., 2012). Together, these studies suggest that impaired NO-cGMP signaling can contribute to the etiology of POAG, identifying the NO-cGMP signaling pathway as a potential therapeutic target for POAG. However, the molecular mechanisms mediating cGMP's effects in the eye, and exactly how NO-cGMP signaling regulates IOP or impacts optic neuropathy remain unclear.

It is important to recognize that the role of NO-cGMP signaling in the eye is likely complex with physiological and pathophysiological effects. For example, contrary to the observations described above that NO lowers IOP, topical application of the

NOS-inhibitor L-NAME reduced IOP in a rabbit model of ocular hypertension (Giuffrida et al., 2003). This reduction of IOP, observed only in hypertensive eyes and not normotensive controls, was suggested to be associated with a decreased formation of AqH production (Giuffrida et al., 2003). Opposing effects of cGMP itself on AqH dynamics have been reported: AqH flow rate decreased after intravitreal administration of the cell permeable cGMP analog 8-Br-cGMP but increased after intra-cameral administration of 8-Br-cGMP, underscoring that the route of administration of any drug aimed at lowering IOP or preventing POAG progression may significantly determines its therapeutic efficacy (Kee et al., 1994).

## **NO MODULATES RGC VIABILITY**

NO can also exert a direct neurotoxic effect on RGC's (Morgan et al., 1999; Takahata et al., 2003). For example, in acute angleclosure glaucoma patients, higher levels of NO in AqH were postulated to contribute to RGC and optic nerve damage (Chang et al., 2000; Chiou et al., 2001). Similarly, in a rat model of chronic glaucoma, elevated retinal NOS1 expression was suggested to contribute to cytotoxicity and selective RGC loss (Park et al., 2007). Furthermore, NO generated by NOS2 can contribute to RGC death seen in response to increased IOP: treatment of rats with chronically elevated IOP with an inhibitor of NOS2 prevented RGC loss (Neufeld et al., 1999, 2002). More clinically relevant, expression of NOS2 was increased in the TM of POAG patients and activity of NOS2 in the TM of patients with POAG was reported to be proportional to the observed visual field defects (Fernandez-Durango et al., 2008). Together, these data could be interpreted to suggest that inhibiting NOS activity may be useful for the treatment of glaucoma by protecting RGCs from stress (e.g., as associated with elevated IOP). However, as described in the studies referred to above, there is ample evidence that impaired NO-cGMP signaling contributes to the pathogenesis of POAG. Together with the fact that NO has multiple downstream targets, of which sGC is arguably a predominant one, these observations support focusing on sGC and cGMP as therapeutic targets in POAG (see "The NO-cGMP pathway as a therapeutic target for POAG" below).

## **MODULATION OF TM AND CM CONTRACTILITY BY NO-cGMP AND ITS RELEVANCE IN IOP REGULATION**

It is generally accepted that the TM and CM are active structures rather than passive filters. The CM is a smooth-muscle like structure and the TM contains contractile smooth-muscle-specific alpha-actin filaments. Both are involved in fine-tuning the regulation of AqH outflow (Erickson-Lamy et al., 1991; Wiederholt, 1998; Wiederholt et al., 2000). Several studies (reviewed in Wiederholt et al., 2000) have suggested the existence of a functional antagonism between the TM and the CM, and TM resistance may be regulated by contractile state of CM (Nathanson and Mckee, 1995b): contraction of the CM would decrease TM resistance, increase AqH outflow and decrease IOP.

Alterations in smooth muscle contractility have previously been implicated in glaucoma. For example, RhoA-Rho kinase-mediated signaling, which regulates the phosphorylation status of myosin light chain, thereby directly influencing TM contraction, was reported to influence AqH drainage (Russ et al., 2010). Inhibiting Rho kinase, an enzyme that plays a critical role in regulating the contractile tone of smooth muscle tissues, increased AqH outflow in various animal models (Honjo et al., 2001; Rao and Epstein, 2007; Lu et al., 2008). The ability of Rhokinase inhibitors to lower IOP is currently being tested in clinical trials (Chen et al., 2011; Tanihara et al., 2013).

The presence of NOS3 and sGC in the CM and TM (**Table 1**) suggest that the NO-sGC pathway may modulate outflow resistance by regulating CM and TM contractility. In fact, both NO-donor compounds and a cGMP analog were previously demonstrated to regulate the contractile state of bovine CM and TM in organ bath experiments (Wiederholt et al., 1994; Masuda et al., 1997; Kamikawatoko et al., 1998). The ability of NO-cGMP to regulate smooth muscle cell relaxation is well known. For example, vascular reactivity was attenuated in NOS3 mutant mice (Atochin et al., 2007) and in sGCα−*/*<sup>−</sup> <sup>1</sup> mice (Mergia et al., 2006; Nimmegeers et al., 2007). NO-cGMP signaling (facilitating smooth muscle relaxation Surks et al., 1999) and RhoA-Rho kinase signaling (causing smooth muscle contraction Bennett et al., 1988) have opposing effects on smooth muscle function and NO-cGMP signaling controls activity of the RhoA-Rho kinase pathway (Sauzeau et al., 2003). Together, these findings raise the possibility that impaired NO-cGMP signaling, previously suggested to participate in the regulation of AqH outflow and IOP (see "NO-cGMP-mediated regulation of IOP: implications for POAG" above) has a similar impact on AqH drainage as does increased RhoA-signaling: both could result in augmented contractility of the smooth muscle-like CM, thereby potentially decreasing AqH drainage and increasing IOP (Honjo et al., 2001).

There is, however, an apparent paradox in the dynamic regulation of IOP by the CM and the role of NO-cGMP therein: impaired NO-cGMP signaling would be expected to result in a decreased ability of the SMC-like CM to relax, just as it does in other SMC-like structures (Mergia et al., 2006; Friebe et al., 2007; Nimmegeers et al., 2007; Vanneste et al., 2007; Decaluwe et al., 2010), and hence decrease TM resistance. Importantly, there are two types of CM in humans: circular, involved in accommodation, and longitudinal (Flugel et al., 1990; Sharif et al., 2004). The effect of CM contraction and relaxation on IOP is not uniform: the circular and longitudinal components of the CM appear to have opposite roles in controlling IOP. Contraction of the circular muscle via muscarinic stimulation (e.g., in response to pilocarpine) causes tightening of the trabecular ring leading to an opening of the trabecular juxtacanalicular tissue, decreased resistance, and increased outflow facility, ultimately lowering IOP (Kaufman, 2008). On the other hand, there is evidence that contraction of the longitudinal muscle can impair outflow and increase IOP. The latter was demonstrated in patients that suffered significant blunt trauma where the circular CM is disabled without affecting the longitudinal muscle. When these patients were treated with pilocarpine, the contraction of the longitudinal muscle paradoxically increased IOP (Bleiman and Schwartz, 1979). Therefore, relaxation of the posterior longitudinal muscle in the CM (e.g., with Rho-kinase inhibitors Honjo et al., 2001) likely has an outflow enhancing effect, lowering IOP. On the other hand, decreased relaxation of longitudinal CM fibers associated with impaired NO-cGMP signaling (e.g., in sGCα−*/*<sup>−</sup> <sup>1</sup> mice, see "sGC-deficient mice: a model of elevated IOP and POAG" below), would result in increased IOP. In addition, repeated contraction and NO-mediated relaxation of any CM component may decreases outflow resistance and lower IOP, likely due to a direct mechanical effect on the TM. Lack of NO-sGC-cGMP signaling in the CM may prevent flushing of the TM, possibly resulting in altered (increased) outflow resistance.

## **THE NO-cGMP PATHWAY AS A THERAPEUTIC TARGET FOR POAG**

The NO-cGMP pathway can be perturbed by a variety of mechanisms. Decreases in NOS3 expression or uncoupling of NOS3 result in lower NO levels (Takimoto et al., 2005). Alternatively, NO bioavailability can be reduced by interaction of NO with reactive oxygen species (ROS) (Munzel et al., 2010). sGC itself can be converted to an NO-insensitive state as a consequence of increased oxidative stress: it is conceivable that the mechanism by which increased oxidative stress results in POAG (Majsterek et al., 2011a) may involve direct oxidation and inactivation of sGC (Stasch et al., 2006). Also, sGC expression can be repressed under pro-inflammatory conditions (Marro et al., 2008). The latter is relevant based on the hypothesis that inflammation contributes to the development of glaucoma (Vohra et al., 2013). Lastly, NO-cGMP signaling can be modulated by genetic variation (see "Genetics of POAG" below) (Kang et al., 2010, 2011; Ehret et al., 2011; Buys et al., 2013). Whether any of the mechanisms that alter activity of the NO-cGMP pathway contribute to the etiology of POAG remains a matter of active investigation. Nonetheless, the studies described here and in the "Modulation of TM and CM contractility by NO-cGMP and its relevance in IOP regulation" section certainly imply the potential usefulness of focusing on NO-cGMP signaling as a therapeutic approach to the treatment of POAG.

Various therapeutic agents have been and are being developed to enhance NO-sGC-cGMP signaling. NO-independent sGC stimulators are already being clinically tested and have been approved for treating cardiovascular disease (Ghofrani et al., 2013). These compounds could conceivably be developed to treat POAG. NO-independent and heme-independent sGC activators preferentially activate sGC when the heme is oxidized or missing (Stasch et al., 2006). The selectivity of sGC activators for oxidized/heme-free sGC could be exploited therapeutically to preferentially target diseased tissue (Boerrigter et al., 2007), especially since oxidative stress may have a pathogenic role in POAG (Majsterek et al., 2011a). Another advantage of specifically activating sGC is that it bypasses any toxicity associated with activating NOS (see "NO-cGMP-mediated regulation of IOP: implications for POAG" above). In addition, targeting sGC rather than NOS circumvents concerns related with reduced NO bioavailability (e.g., in a setting of increased oxidative stress) and allows for a more specific approach since the biological actions of NO are not only mediated by cGMP but also by cGMP-independent mechanisms.

Other potential therapeutic approaches specifically aimed at cGMP signaling that may be developed for treatment of POAG include targeting other enzymes that control GMP levels and which are abundantly present in the eye, including cGMPcatabolizing PDE's (Francis et al., 2011) and NP activated membrane bound guanylate cyclades (see "NP receptors and their effects on IOP" below, Mckie et al., 2010). Pilot studies testing the ability of the PDE5 inhibitor sildenafil to lower IOP were not successful. A single oral dose of 50–100 mg sildenafil failed to impact IOP in POAG patients (Grunwald et al., 2001) and healthy volunteers (Vobig et al., 1999; Sponsel et al., 2000), but did increase blood flow in both healthy subjects (Sponsel et al., 2000; Foresta et al., 2008) and patients with systemic vascular dysfunction (Koksal et al., 2005). Also chronic treatment with sildenafil (twice per week for 3 months) failed to impact IOP (Dundar et al., 2006). It is noteworthy that the cohorts in which the effect of sildenafil on IOP was studied were of limited size (5–15 subjects). It is conceivable that these studies were underpowered to detect small decreases in IOP associated with sildenafil treatment. This is a relevant concern since modest changes in IOP have been reported to impact POAG risk in humans. For example, a 2 mmHg difference in IOP distinguished between progression and non-progression in POAG patients (Leske et al., 2007; Konstas et al., 2012). Also, most of the cohorts studied consisted of men only. In light of a possible gender-specificity of POAG (Vajaranant et al., 2010; Pasquale and Kang, 2011; Tsai et al., 2012) and the role of NO-cGMP signaling therein (Kang et al., 2010), it would be of interest to see whether preventing cGMP catabolism by PDE5 would impact IOP in women. Finally, the half-life of sildenafil is rather limited (∼4 h) (Smith et al., 2013). It may be worthwhile testing the effect on IOP of other PDE5 inhibitors with longer half-lives (e.g., tadalafil with a half-life of ∼17 h Smith et al., 2013) and/or performing a carefully controlled dose response experiment. Finally, various methods to deliver the drug should be tested. Whether pharmacological modulation of the activity of enzymes that ultimately modulate cGMP levels may prevent

## **sGC-DEFICIENT MICE: A MODEL OF ELEVATED IOP AND POAG**

POAG progression remains to be determined.

Mice lacking the α<sup>1</sup> subunit of the NO receptor soluble guanylate cyclase (sGCα−*/*<sup>−</sup> <sup>1</sup> mice) were recently reported to represent a novel and translatable animal model of POAG, characterized by thinning of the retinal nerve fiber layer (RNFL) and loss of optic nerve axons (Buys et al., 2013). The optic neuropathy associated with sGCα1—deficiency was accompanied by modestly increased IOP, on average 2 mmHg higher in 39 ± 14 weeks old female sGCα−*/*<sup>−</sup> <sup>1</sup> mice than in age-matched female wild-type (WT) mice. No IOP difference was detected between age-matched <sup>15</sup> <sup>±</sup> 6 weeks old female sGCα−*/*<sup>−</sup> <sup>1</sup> and age-matched WT mice: age was a predictor of elevated IOP in female sGCα−*/*<sup>−</sup> <sup>1</sup> but not in WT mice. The increase in IOP was accompanied by a decrease in AqH turnover in the context of an open iridocorneal angle: sGCα−*/*<sup>−</sup> <sup>1</sup> mice presented with a normal ciliary body, a well-defined TM, and a patent Schlemm's canal. In addition, spectral domain optical coherence tomography analysis of the iridocorneal angle did not reveal any evidence for angle-closure in sGCα−*/*<sup>−</sup> <sup>1</sup> mice. Similarly, biomicroscopy did not reveal any evidence of exfoliation syndrome, pigment dispersion syndrome, or other conditions that could produce elevated IOP.

Highlighting the multi-faceted role of NO-cGMP signaling in ocular (patho)physiology, retinal vascular dysfunction was observed in sGCα−*/*<sup>−</sup> <sup>1</sup> mice. Other studies had postulated that vascular dysfunction contributes to the etiology of POAG (Henry et al., 1999; Feke and Pasquale, 2008; Su et al., 2008) and systemic vascular dysfunction had been previously reported in sGCα−*/*<sup>−</sup> 1 mice (Nimmegeers et al., 2007; Atochin et al., 2010; Buys et al., 2012). Whether retinal vascular dysfunction underlies the retinal degeneration and optic neuropathy in sGCα−*/*<sup>−</sup> <sup>1</sup> mice remains to be determined. Vascular dysfunction may also contribute to the observed increases in IOP in sGCα−*/*<sup>−</sup> <sup>1</sup> mice, possibly by impairing uveolymphatic outflow (see "Intraocular pressure: a risk factor for POAG" above). Finally, it cannot be excluded that loss of RGCs in sGCα−*/*<sup>−</sup> <sup>1</sup> mice is caused by a direct effect of sGCdeficiency on RGCs, possibly modulating the susceptibility of RGCs to stress imposed by increased IOP or vascular dysfunction. Taken together, targeting sGC may represent a multi-pronged approach, aimed at lowering IOP, ameliorating vascular function, and protecting RGCs from stress-induced dysfunction and death. To test this hypothesis, additional studies, including in animal models of POAG, such as the sGCα−*/*<sup>−</sup> <sup>1</sup> mouse model, are required.

## **GENETICS OF POAG**

First-degree relatives of glaucoma patients have a 22% risk of developing glaucoma versus a 2% lifetime risk of relatives of controls (Wolfs et al., 1998). In recent years, familial aggregation, genome wide and candidate gene association studies have uncovered an important genetic component to POAG (**Table 2**). Several dozen genetic loci have been linked to POAG (Fan et al., 2006). The role of many of the identified genes in the etiology of POAG, some of which were identified in single studies that need to be replicated, remains controversial. Among the genes identified to date, three are involved in the nitric oxide (NO)-cGMP system, highlighting a central role of this signaling pathway in the development of POAG. First, variants were identified in or near the genes encoding caveolin 1 and 2 (*CAV1/CAV2*) (Thorleifsson et al., 2010; Wiggs et al., 2011). Caveolins modulate the ability of NOS3 to generate NO (Mineo and Shaul, 2012). At least one variant in the *CAV1/CAV2* locus was subsequently found to be associated with increased IOP (Van Koolwijk et al., 2012; Ozel et al., 2014). Second, A candidate gene association study in 527 incident cases and 1543 controls revealed interactions between *NOS3* gene variants, potentially affecting expression and/or activity of NOS3, and high tension POAG in females (Kang et al., 2010). In addition, a functional NOS3 polymorphism (T-786C) was associated with POAG and appears to interact with gender and age in modulating the risk of POAG (Magalhaes Da Silva et al., 2012). The same variant also affects the interaction of systemic hypertension and cigarette smoking with POAG risk, highlighting the complex gene-environment interactions that impact the etiology of POAG (Kang et al., 2011). And

#### **Table 2 | POAG-associated genes.**


*AGTR2, Angiotensin II receptor type 2; APOE, Apolipoprotein E; ASB10, ankyrin repeat and SOCS box-containing 1; ATOH7, atonal homolog 7; C7, complement component 7; CAV1, caveolin 1; CAV2, caveolin 2; CDKN1A, cyclin-dependent kinase inhibitor 1A; CDKN2B-AS1, CDKN2B antisense RNA 1; CGAS, candidate gene association study; CYP1B1, cytochrome P450 subfamily I polypeptide 1; EDNRA, endothelin receptor type A; ELOVL1, elongation of very-long-chain-fatty acids 1; GAS7, growth arrest-specific protein 7; GLC1B, glaucoma 1, open angle, B; GSTM1, glutathione S-transferase mu-1; GWAS, genome-wide association study; HK2, hexokinase 2; HTG, high-tension glaucoma; IGF2, insulin-like growth factor II; IL1B, interleukin 1-beta; IOP, intra-ocular pressure; JOAG, juvenile open angle glaucoma; LOXL1, lysyl oxidase-like 1; MMP1, matrix metalloproteinase 1; MTHFR, 5,10-methylenetetrahydrofolate reductase; MYOC, myocilin; NCK2, NCK adaptor protein 2; NOS3, nitric oxide synthase 3; NPPA, natriuretic peptide precursor; NTG, normal-tension glaucoma; OCLM, oculomedin; OPA1, optic atrophy 1; OPTN, optineurin; POAG, primary open angle glaucoma; RFTN1, raftlin lipid raft linker 1; Six1/Six6, six homeobox1/6; SRBD1, S1 RNA-Binding Domain-Containing Protein 1; TAP1, transporter ATP-binding cassette major histocompatibility complex 1; TMCO1, transmembrane and coiled-coil domains 1; TNF, tumor necrosis factor; TP53, tumor protein p53; WDR36, WD repeat domain 36; XFG, exfoliation glaucoma.*

thirdly, in a recent candidate gene association study in POAG patients from the GLAUGEN cohort (Wiggs et al., 2011), a variant (rs11722059) was identified in the *GUCY1A3/GUCY1B3* locus (containing the genes encoding the α<sup>1</sup> and β<sup>1</sup> subunits of sGC, arranged in tandem) (Wiggs et al., 2011). Intriguingly, the association was only present in POAG characterized by early paracentral visual field loss (Buys et al., 2013). Early paracentral visual field loss is a subtype of POAG previously postulated to be associated with ocular vascular dysregulation (Park et al., 2011). This was a particularly interesting finding in light of the identification of sGCα−*/*<sup>−</sup> <sup>1</sup> mice with systemic and retinal vascular dysfunction, as a model of POAG with moderately elevated IOP: as described above, POAG patients with early paracentral visual field loss tend to have more frequent systemic vascular risk factors (Park et al., 2011, 2012), and low OPP is a risk factor for POAG (Leske, 2009; Cherecheanu et al., 2013). In addition, rs11722059 is in linkage disequilibrium with a *GUCY1A3/GUCY1B3* variant associated with blood pressure in a large genome wide association study (GWAS) (Ehret et al., 2011). However, the extent to which vascular dysfunction contributes to glaucomatous optic neuropathy remains to be elucidated (Flammer et al., 2002; Vajaranant and Pasquale, 2012).

For neither the *NOS3* or *GUCY1A3/GUCY1B3* variants, an association with IOP was detected. Whether this truly means that there is no association with the identified variants remains unsure: it is conceivable that the GWAS and candidate gene association studies discussed were underpowered to detect small effect on IOP. For example, the GWAS in which the sGC blood pressure variant was identified included 200,000 subjects (Ehret et al., 2011). In contrast, the studies aimed at identifying associations with POAG and IOP (like systemic blood pressure a continuous variable) were much smaller in scope (hundreds to several thousand subjects).

## **NP RECEPTORS AND THEIR EFFECTS ON IOP**

There are three NPs in mammals: atrial NP, B-type NP and Ctype NP (Potter et al., 2009). A separate gene encodes each NP, and mouse "knock out" experiments have shown that each NP has unique functions, although all members stimulate vasorelaxation. ANP and BNP are cardiac endocrine hormones that decrease blood pressure and volume. CNP is a paracrine-signaling molecule that stimulates long bone growth, causes bifurcation of neurons in the spinal cord and inhibits meiosis in the oocyte.

There are also three known NP peptide receptors (Potter, 2011a,b,c). The natriuretic clearance receptor (NPR-C) controls the local concentration of all three natriuretic peptides through an undefined receptor-mediated internalization and degradation process, but it has also been reported to signal through a Gprotein-dependent pathway. The primary signaling receptor for ANP and BNP is guanylate cyclase-A (GC-A), which is also known as natriuretic peptide receptor-A or NPR1. It consists of an extracellular ligand-binding domain, a single membranespanning region and intracellular guanylate cyclase domain. GC-B, also known at NPR-B or NPR2, is the primary signaling receptor for CNP and it is topologically and structurally similar to GC-A. The vast majority of NPs functions are mediated through elevation of intracellular cGMP concentrations synthesized by the guanylate cyclase domains of GC-A or GC-B.

Early studies reported that binding sites for 125I-ANP were found in rat and rabbit ciliary processes (Bianchi et al., 1986). Shortly thereafter, topical, intracameral and intravitreal application of ANP was shown to decrease IOP in rabbits (Sugrue and Viader, 1986). The pressure reductions were longer lasting with the intravitreal treatments but even the topical application reduced IOP. Additional studies found that ANP stimulated guanylate cyclase activity in ciliary processes of the rabbit eye and that intravitreous injection of ANP decreased IOP in the rabbit eye for up to 24 h (Mittag et al., 1987). The reduction of IOP by intravitreal ANP was later correlated to a decrease in AqH flow in the rabbit eye (Korenfeld and Becker, 1989).

Relatively low concentrations of CNP (∼2 nmol/L) were also shown to increase cGMP concentrations in the AqH of the rabbit eye and reduce IOP in rabbit eyes by a process associated with an increase in total outflow facility (Takashima et al., 1998). In the same study, CNP-like immunoreactivity was detected in rabbit and porcine AqH at about two-fold higher concentrations than that found in plasma. In a more recent study, CNP was shown to be a more potent reducer of IOP than ANP in rabbit eyes, and a ring-deleted analog of ANP that blocked binding to NPR-C and increased the concentration of all endogenous natriuretic peptides, maintained IOP reductions longer that other natriuretic peptides (Fernandez-Durango et al., 1999). Thus, whether the reductions in IOP observed by early investigators were due to cross-activation of GC-B or whether activation of GC-B and GC-A leads to decreased IOP is not known. Regardless, CNP is the most potent natriuretic peptide for reducing IOP in mammals, which suggests that the degradation resistant CNP analog developed for the treatment of achondroplasia may be an ideal peptidebased molecule to reduce IOP in the clinic (Lorget et al., 2012).

### **CONCLUSION**

In conclusion, POAG remains a major cause of blindness in the USA and worldwide. The identification of new therapeutic targets for the treatment of POAG has been hampered by lack of understanding of the etiology of POAG and the limited number of animal models available that likely represent only a small subset of human POAG cases. There continues to be an urgent need for biomarkers that allow to detect/diagnose/track POAG progression and treatment efficacy. Also, the lack of a definitive cure underscores the need to develop novel therapeutic approaches for POAG.

Evidence obtained from animal models indicates that cGMP signaling plays an important role in POAG pathogenesis. These findings are supported by genetic studies of IOP and POAG risk in human subjects, highlighting the relevance of cGMP signaling in the development of POAG. Strategies aimed at modulating cGMP levels may constitute a pharmacological approach for a disease for which no definitive cure is currently available. However, additional studies are required to unequivocally determine the role of impaired cGMP signaling in POAG and to advance benchto-bedside translation of available cGMP-enhancing reagents.

#### **ACKNOWLEDGMENTS**

The authors thank Peter Mallen for generating the figures. Dr. Buys was supported by NIH/NEI 1R01EY022746-01. Dr. Pasquale was supported by a Harvard Medical School Distinguished Ophthalmology Scholar Award and by NIH/NEI RO1EY015473-10.

## **REFERENCES**


Kaiser, H. J., and Flammer, J. (1991). Systemic hypotension: a risk factor for glaucomatous damage? *Ophthalmologica* 203, 105–108.


results from the Low-Pressure Glaucoma Treatment Study. *Am. J. Ophthalmol.* 151, 671–681. doi: 10.1016/j.ajo.2010.09.026


factors in regulating expression of the nitric oxide receptor. *J. Biol. Chem.* 283, 20027–20036. doi: 10.1074/jbc.M801223200


dogs. *Vet. Ophthalmol.* 15(Suppl. 1), 60–63. doi: 10.1111/j.1463-5224.2011. 00956.x


**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: 14 March 2014; paper pending published: 19 April 2014; accepted: 21 April 2014; published online: 19 May 2014.*

*Citation: Buys ES, Potter LR, Pasquale LR and Ksander BR (2014) Regulation of intraocular pressure by soluble and membrane guanylate cyclases and their role in glaucoma. Front. Mol. Neurosci. 7:38. doi: 10.3389/fnmol.2014.00038*

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

*Copyright © 2014 Buys, Potter, Pasquale and Ksander. 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.*

## Guanylate cyclase-C/cGMP: an emerging pathway in the regulation of visceral pain

## *Gerhard Hannig, Boris Tchernychev, Caroline B. Kurtz, Alexander P. Bryant, Mark G. Currie and Inmaculada Silos-Santiago\**

Department of Discovery Pharmacology, Ironwood Pharmaceuticals, Inc., Cambridge, MA, USA

#### *Edited by:*

Clint Lawrence Makino, Massachusetts Eye and Ear Infirmary and Harvard Medical School, USA

#### *Reviewed by:*

Jakub Fichna, Medical University of Lodz, Poland Scott Waldman, Thomas Jefferson University, USA

#### *\*Correspondence:*

Inmaculada Silos-Santiago, Department of Discovery Pharmacology, Ironwood Pharmaceuticals, Inc., 301 Binney Street, Cambridge, MA 02142, USA e-mail: isilos-santiago@ ironwoodpharma.com

Activation of guanylate cyclase-C (GC-C) expressed predominantly on intestinal epithelial cells by guanylin, uroguanylin or the closely related GC-C agonist peptide, linaclotide, stimulates generation, and release of cyclic guanosine-3- ,5- -monophosphate (cGMP). Evidence that the visceral analgesic effects of linaclotide are mediated by a novel, GC-C-dependent peripheral sensory mechanism was first demonstrated in animal models of visceral pain. Subsequent studies with uroguanylin or linaclotide have confirmed the activation of a GC-C/cGMP pathway leading to increased submucosal cGMP mediated by cGMP efflux pumps, which modulates intestinal nociceptor function resulting in peripheral analgesia.These effects can be reproduced by the addition of exogenous cGMP and support a role for GC-C/cGMP signaling in the regulation of visceral sensation, a physiological function that has not previously been linked to the GC-C/cGMP pathway. Notably, targeting the GC-C/cGMP pathway for treatment of gastrointestinal pain and abdominal sensory symptoms has now been validated in the clinic. In 2012, linaclotide was approved in the United States and European Union for the treatment of adult patients with irritable bowel syndrome with constipation.

**Keywords: abdominal pain, colonic nociceptor, cyclic guanosine monophosphate, guanylate cyclase-C, irritable bowel syndrome with constipation, linaclotide, uroguanylin, visceral analgesia**

#### **INTRODUCTION**

Guanylate cyclase-C (GC-C) is a type I transmembrane receptor with intrinsic guanylate cyclase activity, belonging to a larger family of enzymes comprising both soluble and receptor guanylate cyclases that respond to a diverse range of signals by catalyzing the conversion of guanosine triphosphate to cyclic guanosine-3- ,5- monophosphate (cGMP; Lucas et al., 2000; Vaandrager, 2002). This receptor is expressed predominantly on the apical surface of intestinal epithelial cells (IEC), and its important role in the regulation of intestinal electrolyte and fluid homeostasis is firmly established (Schulz et al., 1990; Carrithers et al., 1996; Castro et al., 2013; Silos-Santiago et al., 2013). The endogenous hormones guanylin and uroguanylin, cognate ligands of GC-C, are expressed in a distinct but overlapping regional pattern in enteroendocrine cells along the longitudinal axis of the gastrointestinal (GI) tract and released into the intestinal lumen (Currie et al., 1992; Kita et al., 1994; Perkins et al., 1997; Qian et al., 2000; Silos-Santiago et al., 2013). Activation of GC-C by either guanylin or uroguanylin results in cGMP accumulation in IEC. Cyclic GMP is the sole second messenger generated by GC-C and is known to regulate several canonical intracellular signaling pathways mediated primarily through direct interaction with three groups of target proteins: cGMP-dependent protein kinases, cyclic nucleotide-gated ion channels, and cGMP-regulated phosphodiesterases (Pfeifer et al., 1996; Schlossman et al., 2005). Cyclic GMP regulation of intestinal electrolyte and fluid homeostasis is dependent on protein kinase II (PKG-II) phosphorylation and activation of the cystic fibrosis transmembrane conductance regulator (CFTR) ion channel, stimulating transepithelial secretion of chloride and bicarbonate ions and concomitant inhibition of sodium absorption by the sodium/hydrogen exchanger 3 (NHE3), resulting in the net efflux of ions and water into the lumen (Forte, 1999; Vaandrager, 2002; Sindic and Schlatter, 2006). Studies in GC-C null (*Gucy2c*−/−) mice have confirmed the central role of GC-C in the control of electrolyte and fluid homeostasis (Mann et al., 1997; Schulz et al., 1997).

Our understanding of cellular processes regulated by activation of the GC-C/cGMP pathway has markedly evolved over the past two decades. Importantly, this knowledge has guided the development of novel therapeutic paradigms targeting the GC-C/cGMP pathway that have successfully translated into the clinic.

#### **LINACLOTIDE IS A POTENT AND SELECTIVE GUANYLATE CYCLASE-C AGONIST**

Linaclotide, a synthetic 14-amino acid peptide composed of naturally occurring amino acids (N-CCEYCCNPACTGCY-C; molecular weight: 1524.2 Dalton), is a potent and selective agonist of GC-C. This orally administered, minimally absorbed peptide (≤0.1% in all non-clinical species) is a member of the guanylin family of cGMP-regulating peptides that includes the natural hormones guanylin and uroguanylin (Bryant et al., 2010; Busby et al., 2010). Linaclotide is stabilized by three intramolecular disulfide bonds (Cys1-Cys6, Cys2-Cys10, Cys5-Cys13), locking this peptide into a single conformation similar in structure to guanylin and uroguanylin, with the distinct difference that the biologically active conformation (A-isoform) of these peptides contains only two

disulfide bonds (Skelton et al., 1994; Marx et al., 1998; Busby et al., 2010). Linaclotide exhibited high-affinity and pH-independent binding to GC-C on human colon carcinoma T84 cells and concomitantly stimulated a significant and concentration-dependent accumulation of intracellular cGMP, with greater potency than guanylin or uroguanylin. In standard rodent models of GI function, orally administered linaclotide elicited significant *in vivo* pharmacological effects, stimulating fluid secretion and accelerating transit (Bryant et al., 2010; Busby et al., 2010). The lack of such effects in *Gucy2c*−/−mice further confirmed GC-C as the molecular target of linaclotide and supported an underlying mechanism linking the effects of linaclotide in these models to local activation of GC-C in the intestine. The single pharmacologically active metabolite of linaclotide, MM-419447 (Des-Tyr14) mirrors the effects of linaclotide *in vitro* and *in vivo*, providing evidence that this active metabolite contributes to the pharmacology associated with oral administration of linaclotide (Busby et al., 2013).

The potent pharmacological effects of linaclotide in nonclinical models provided the rationale for development of this peptide as a novel therapeutic for the treatment of functional GI disorders associated with constipation, such as chronic idiopathic constipation (CIC), and irritable bowel syndrome with constipation (IBS-C). However, no reports had previously linked local activation of the intestinal GC-C pathway to the regulation of abdominal pain.

## **LINACLOTIDE ACTIVATION OF GUANYLATE CYCLASE-C ELICITS ANALGESIC EFFECTS IN VISCERAL PAIN MODELS**

Abdominal pain or discomfort (associated with a change in bowel function) is a defining symptom of the diagnostic criteria for IBS and is hypothesized to originate from hypersensitivity of the colon to mechanical stimuli (Drossman, 2006; Tack et al., 2006; Videlock and Chang, 2007). In clinical studies, IBS patients have shown lowered colorectal pain thresholds and increased sensory ratings, further consistent with persistent enhanced perception and responsiveness to visceral stimuli (= visceral hyperalgesia; Mertz et al., 1995; Bouin et al., 2002).

Antinociceptive effects of an orally administered GC-C agonist, linaclotide, were first demonstrated by Eutamene et al. (2010), using rodent models of visceral hypersensitivity associated with inflammation produced by trinitrobenzene sulfonic acid (TNBS) or stress models (following acute water avoidance or partial restraint) that induce this condition. In these models, visceral pain is measured using colorectal balloon distension. This well-characterized method evokes contractions of the abdominal musculature (visceromotor response) to graded distension pressures. It differs from measurements of visceral sensitivity in humans, which is primarily based on conscious perception of controlled colon distension. Orally administered linaclotide elicited significant analgesic effects during colonic distension in all models tested, without affecting basal sensitivity. Notably, the specificity of linaclotide-mediated analgesic effects was mechanistically linked to the activation of intestinal GC-C, confirmed in studies that assessed the effects of linaclotide on TNBS-induced colonic hypersensitivity in*Gucy2c* wild-type and*Gucy2c*−/<sup>−</sup> mice. In these studies, linaclotide reversed colonic hypersensitivity only in *Gucy2c*

wild-type, but not *Gucy2c*−/<sup>−</sup> mice. The analgesic effects of linaclotide were also reproduced by the endogenous GC-C ligand, uroguanylin, a finding not previously attributed to the endogenous hormones. The analgesic effects of linaclotide and uroguanylin in these models were not associated with changes in colonic pressurevolume relationships, further suggesting that the antinociceptive effects are linked specifically to GC-C activation and are not associated with altered colonic compliance (Eutamene et al., 2010; Silos-Santiago et al., 2013). These findings raised the intriguing question of whether activation of an intestinal GC-C pathway had emerged as a novel physiological pathway regulating visceral pain.

## **LINACLOTIDE INDUCES PERIPHERAL VISCERAL ANALGESIA BY ACTIVATION OF A GUANYLATE CYCLASE-C/EXTRACELLULAR cGMP PATHWAY**

Hallmarks of IBS include allodynia (persistent response to normally non-noxious stimuli) and hyperalgesia (persistent enhanced perception and responsiveness to noxious stimuli) to mechanical stimuli within the intestine, mediated by chronic changes in afferent pathways (sensitization). High-threshold colonic nociceptive sensory afferents (mesenteric, serosal afferents) that only respond at sufficiently high levels to mechanical stimulation are predominantly found in the splanchnic nerve pathway, while low-threshold stretch receptors [muscular/mucosal (M/M), muscular afferents] are predominantly found in the pelvic pathway (Hughes et al., 2009; Feng and Gebhart, 2011; Blackshaw and Brierley, 2013).

In a mouse model of chronic post-inflammatory visceral hypersensitivity (CVH) in which colonic mechanical hyperalgesia and allodynia are evident long after resolution of TNBS-induced colitis, linaclotide, and uroguanylin reversed chronic mechanical hypersensitivity in CVH colonic high-threshold splanchnic serosal nociceptors, and also inhibited mechanical hypersensitivity of healthy nociceptors. Inhibition of CVH colonic nociceptors by either peptide was greater than their inhibition of control nociceptors from healthy animals, and linaclotide-induced inhibition was more potent than that produced by uroguanylin. Importantly, *in vitro* inhibition of colonic nociceptors correlated with *in vivo* findings in which linaclotide decreased the processing of noxious colorectal distension stimuli in the thoracolumbar spinal cord indicated by a lower number of activated dorsal horn (DH) neurons within the thoracolumbar spinal cord, specifically the superficial lamina of the DH, recognized as the major termination zone for nociceptive afferents (Castro et al., 2013). Further evidence supporting a mechanism in which linaclotide inhibition of colonic nociceptors is dependent on local activation of GC-C in IEC rather than direct effects on colonic nociceptors was derived from expression studies using *in situ* hybridization in whole adult mouse, colonic segments, and spinal cord and dorsal root ganglion (DRG) sections, and studies assessing linaclotide inhibition of colonic nociceptors in *Gucy2c*−/<sup>−</sup> mice. While abundant GC-C expression was found in the intestine, no expression was detected in key sensory structures, such as DRG and spinal cord neurons. Furthermore, while mechanosensory responses of colonic nociceptors at baseline were similar in control and *Gucy2*−/<sup>−</sup> mice, linaclotide-induced inhibition was completely lost in *Gucy2c*−/<sup>−</sup> mice. This was further corroborated by findings that prior removal of the mucosa, the source of GC-C, in healthy and CVH states significantly attenuated linaclotide inhibition of colonic nociceptors (Castro et al., 2013). Finally, electric field stimulation-induced contractions of colonic tissues were not affected by linaclotide, which confirmed that antinociceptive effects of linaclotide were not linked to altered smooth muscle cell contractility, consistent with the absence of GC-C expression on colonic smooth muscle cells (Castro et al., 2013).

While these findings indicate that the analgesic effects of linaclotide are mediated entirely by peripheral expression of GC-C on the apical surface of IEC, a direct inhibitory effect of linaclotide on colonic nociceptors was further considered unlikely because of minimal absorption of this peptide, and the location of peripheral endings of colonic nociceptors in the submucosa. This provoked studies that investigated the potential role of cGMP, the common primary downstream effector of linaclotide and uroguanylin, in altered intestinal nociceptor function and peripheral analgesia. The involvement of an intracellular GC-C/cGMP pathway in the regulation of intestinal fluid and electrolyte homeostasis is firmly established; however, the effects of extracellular cGMP transported out of IEC following local GC-C activation have remained elusive. Precedence that extracellular cGMP is involved in the modulation of neuronal activity was obtained from studies that showed direct inhibitory effects of extracellular cGMP on CNS neurons, resulting in reduced excitability and inhibition of neurotransmitter release (Linden et al., 1995; Poupoloupou and Nowak, 1998; Cervetto et al., 2010). When assessed in rat models of colonic hypersensitivity, orally administered cGMP elicited significant analgesic effects, in a dose-dependent manner, reproducing the effects of linaclotide and uroguanylin in these models (Silos-Santiago et al., 2013). There is now compelling evidence supporting a model in which the potent analgesic effects of cGMP *in vivo* are mediated by a pathway linking extracellular cGMP, secreted from IEC into the submucosa following activation of the GC-C/cGMP pathway by linaclotide or uroguanylin, to altered function of colonic nociceptors resulting in peripheral analgesia. *In vitro* exposure of human intestinal Caco-2 cells to linaclotide or uroguanylin stimulated extracellular transport of cGMP into the apical and basolateral spaces, which was inhibited by the cGMP efflux pump inhibitor probenecid in a concentration-dependent manner (Castro et al., 2013; Silos-Santiago et al., 2013). This provided evidence implicating energy-dependent transport of cGMP by the cGMP efflux pumps multidrug-resistance protein (MRP) 4 and 5 (Sager, 2004). While cGMP-binding phosphodiesterases are generally recognized as the major elimination pathwayfor intracellular cGMP, MRP4/5-mediated extracellular transport of cGMP is consistent with theirfunction as overflow pumps, decreasing intracellular cGMP levels under conditions when cGMP production is strongly induced and importantly, providing extracellular cGMP for paracrine actions (Ritter et al.,2005; Zimmermann et al.,2005). Further evidence supporting a role of extracellular cGMP in the regulation of colonic afferent activity was obtained from *ex vivo* Ussing chamber assays, in which exposure of rat colonic tissue (luminal side) to uroguanylin stimulated secretion of cGMP into the submucosal space (Silos-Santiago et al., 2013). Moreover, in a rat model of TNBS-induced colonic afferent sensitization, exogenous cGMP significantly decreased pelvic afferent firing rates

in response to colonic distension, and in CVH mice colonic nociceptors were significantly inhibited by application of exogenous cGMP to the mucosal epithelium, to a greater extent than those from healthy mice (Castro et al., 2013; Silos-Santiago et al., 2013). While cGMP dose levels required for inhibition of colonic nociceptors exceeded those for linaclotide and uroguanylin, facilitating access of cGMP to colonic nociceptors by removal of the mucosa significantly increased its potency, confirming the barrier function of the epithelium for luminal cGMP to diffuse across the mucosa (Castro et al., 2013). Furthermore, direct application of cGMP to mouse colorectal receptive endings significantly decreased the response of control pelvic muscular and M/M afferents to circumferential stretch, and sensitized responses of muscular and M/M afferents to stretch were reversed (Feng et al., 2013). Similar to findings with linaclotide and uroguanylin, antinociceptive effects of extracellular cGMP were not associated with altered smooth muscle contractility (Castro et al., 2013; Silos-Santiago et al., 2013).

In conclusion, accumulating evidence now strongly supports a direct peripherally acting analgesic mechanism that, following activation of a GC-C/extracellular cGMP pathway by selective GC-C agonists, mediates inhibition of colonic nociception and decreases visceral pain (**Figure 1**). This mechanism suggests that the regulation of colonic sensation may have evolved as an effect of GC-C agonism by the endogenous hormones guanylin and uroguanylin in IEC (Castro et al., 2013; Silos-Santiago et al., 2013).

## **LINACLOTIDE DECREASES ABDOMINAL PAIN AND DISCOMFORT IN PATIENTS WITH IRRITABLE BOWEL SYNDROME WITH CONSTIPATION**

Results from two pivotal phase 3 clinical trials with linaclotide in adult patients with IBS-C have validated the approach of therapeutically targeting the GC-C/cGMP pathway for treatment of abdominal pain in this disorder (Chey et al., 2012; Rao et al., 2012). The efficacy and safety of linaclotide were assessed in a 26 week, double-blind, parallel-group, placebo-controlled, randomized trial, and a 12 week, double-blind, parallel-group, placebo-controlled, randomized trial, with a 4 week randomized withdrawal period. In both trials, four prespecified primary endpoints were evaluated, based on the FDA primary endpoint for IBS-C [responder: improvement ≥30% from baseline in average daily worst abdominal pain score and increase by ≥1 complete spontaneous bowel movement (CSBM)from baseline (same week) for at least 50% of weeks assessed], and three other primary endpoints, based on improvements in abdominal pain and CSBM for 9/12 weeks. Notably, a 30% reduction in the pain score has previously been shown of clinical importance in IBS patients and in patients reporting pain relief in general (Farrar et al., 2001; Spiegel et al., 2009).

In the 26 week trial, 804 IBS-C patients randomized (1:1) to linaclotide or placebo, received oral linaclotide (290 μg, once daily) during the 26 week treatment period (primary and secondary efficacy assessments were evaluated over the first 12 weeks of treatment; Chey et al., 2012). For the rigorous, combined IBS-C endpoint recommended in the recent *Guidance for IBS Clinical Trials* (Food and Drug Administration, 2010), 33.7% of IBS-C patients in the linaclotide group were endpoint responders,

compared with 13.9% in the placebo group (*P* < 0.0001). The FDA pain responder criterion was met by 48.9% of linaclotide-treated patients, compared to 34.5% in the placebo group (*P* < 0.0001) and similarly, 49.1% of linaclotide-treated patients met the abdominal pain endpoint for at least 13/26 weeks, compared to the placebo group (*P* < 0.0001). Furthermore, improvement in abdominal pain compared to placebo, was significant startingfrom the first week of therapy and continuing throughout the 26 week treatment period.

In the 12 week randomized trial, followed by a 4 week randomized withdrawal period, 800 patients (1:1, linaclotide or placebo) received oral linaclotide (290 μg, once daily) during the 12 week treatment period (Rao et al., 2012). Patients who had completed all 12 weeks of the double-blind treatment period were eligible to enter the double-blind 4 week randomized withdrawal period, in which patients initially randomized to linaclotide were re-randomized (1:1) to oral linaclotide (290 μg, once daily) or placebo, and patients previously randomized to placebo were assigned to receive oral linaclotide (290 μg) once a day. The combined FDA-recommended IBS-C responder endpoint was met by 33.6%, compared to 21% in the placebo group (*P* < 0.0001), and similarly a larger number of linaclotide-treated patients (50.1%) reported a reduction of ≥30% in abdominal pain, compared to the placebo group (37.5%; *P* = 0.0003). As in the 26 week clinical trial with linaclotide, improvement in abdominal pain was recorded within the first week of therapy and sustained

throughout the treatment period. During the 4 week randomized withdrawal period, patients on linaclotide had continued relief of abdominal pain, while patients re-randomized to placebo showed a gradual worsening of abdominal pain symptoms to the level experienced by patients receiving placebo during the treatment period, however, without worsening of symptoms relative to baseline.

Finally, results from several pooled *post hoc* analyses of these two phase 3 trials further confirmed the efficacy of linaclotide treatment on symptoms of abdominal pain and discomfort in patients with IBS-C (Johnston et al., 2013). Importantly, it was shown in a *post hoc* longitudinal responder analysis using the FDA responder criterion that at week 3, more than 50% of linaclotidetreated patients reported >30% reduction in abdominal pain, and that this analgesic effect increased to greater than 60% of linaclotide-treated patients at week 7 and was sustained at approximately 70% of linaclotide-treated patients for the duration of the 26 week study (Castro et al., 2013).

Based on results from these two phase 3 pivotal trials, linaclotide received FDA approval in 2012 as a first-in-class drug for the treatment of adult patients with IBS-C, and by the European Medicines Agency for the symptomatic treatment of moderate to severe IBS-C in adult patients. Linaclotide was also approved in the US for the treatment of CIC (2012) and in Canada for the treatment of both IBS-C and CIC (2013). Today, linaclotide is in worldwide development for the treatment of IBS-C.

#### **SUMMARY AND PERSPECTIVE**

This review offers a new perspective and understanding of endogenous mechanisms involved in the regulation of visceral pain. The intestinal GC-C/extracellular cGMP pathway is emerging as a novel pathway that regulates peripheral analgesia via inhibition of primary colonic afferents, which results in decreased visceral pain. Findings from both non-clinical models and clinical studies have confirmed that the analgesic effects of linaclotide are mediated by a distinct pain-regulating pathway that operates independently from improvements in bowel function. In order to maximize the benefits of these novel findings for patients work is ongoing to fully elucidate the underlying mechanisms of this pathway, such as the identification and characterization of the molecular target(s) on colonic afferents and secondly, to more broadly assess the significant therapeutic potential of targeting the GC-C/cGMP pathway in GI disorders associated with visceral pain. The efficacy of linaclotide on abdominal pain demonstrated in the phase 3 IBS-C clinical trials translates the targeting of the GC-C/cGMP pain pathway into the clinic. This pathway continues to emerge as a novel therapeutic target for GI dysfunction with the potential for broader utility in treating visceral pain.

### **AUTHOR CONTRIBUTIONS**

Alexander P. Bryant, Mark G. Currie, Gerhard Hannig, Caroline B. Kurtz, Inmaculada Silos-Santiago, and Boris Tchernychev contributed to writing and critical revisions of the manuscript.

#### **REFERENCES**


bowel syndrome with constipation. *Am. J. Gastroenterol.* 107, 1714–1724. doi: 10.1038/ajg.2012.255


**Conflict of Interest Statement:** Alexander P. Bryant, Mark G. Currie, Gerhard Hannig, Caroline B. Kurtz, Inmaculada Silos-Santiago and Boris Tchernychev are employees and stock/stock option holders in Ironwood Pharmaceuticals, Inc.

*Received: 30 January 2014; paper pending published: 19 March 2014; accepted: 01 April 2014; published online: 16 April 2014.*

*Citation: Hannig G, Tchernychev B, Kurtz CB, Bryant AP, Currie MG and Silos-Santiago I (2014) Guanylate cyclase-C/cGMP: an emerging pathway in the regulation of visceral pain. Front. Mol. Neurosci. 7:31. doi: 10.3389/fnmol.2014.00031*

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

*Copyright © 2014 Hannig, Tchernychev, Kurtz, Bryant, Currie and Silos-Santiago. 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.*