**THE REGULATED SECRETORY PATHWAY IN NEUROENDOCRINE CELLS** 

**Topic Editors Rafael Vazquez-Martinez and Stéphane Gasman**

# ENDOCRINOLOGY

#### *FRONTIERS COPYRIGHT STATEMENT*

© Copyright 2007-2014 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** 978-2-88919-246-5 **DOI** 10.3389/978-2-88919-246-5

### *ABOUT FRONTIERS*

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

### *FRONTIERS JOURNAL SERIES*

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing.

All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### *DEDICATION TO QUALITY*

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

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view.

By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### *WHAT ARE FRONTIERS RESEARCH TOPICS?*

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area!

Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **THE REGULATED SECRETORY PATHWAY IN NEUROENDOCRINE CELLS**

Topic Editors:

**Rafael Vazquez-Martinez,** University of Cordoba, Cordoba, Spain **Stéphane Gasman,** Institut des Neurosciences Cellulaires et Intégratives, CNRS UPR 3212, Strasbourg, France

The image is the property of: Haeberlé A.M., Bailly Y. and Gasman S : Institut des Neurosciences Cellulaires et Intégratives (INCI), CNRS UPR3212; Plateforme Imagerie In Vitro, Neuropôle, Strasbourg.

The regulated secretory pathway is a hallmark of neuroendocrine cells. This process comprises many sequential steps, which include ER-associated protein synthesis, post-translational modification of proteins in the Golgi complex, sorting and packing of secretory proteins into carrier granules, cytoskeleton-based granule transport towards the plasma membrane and tethering, docking and fusion of granules with specialized releasing zones. Each stage is subjected to a rigorous regulation by a plethora of factors that function in a spatially and temporarily coordinated fashion. Much effort has been devoted to characterize the precise role of the regulatory proteins participating in the different steps of this process and to identify new factors in order to obtain a unifying picture of the secretory pathway. In spite of this and given the enormous complexity of the process, certain stages are not fully understood

yet and many players remain to be identified. The aim of this Research Topic is to gather review articles and original research papers on the molecular mechanisms that govern and ensure the correct release of neuropeptides.

# Table of Contents



Paanteha K. Moghadam and Meyer B. Jackson


María Luisa Durán-Pastén and Tatiana Fiordelisio


# The regulated secretory pathway in neuroendocrine cells

### **Rafael Vazquez-Martinez 1,2\* and Stéphane Gasman<sup>3</sup>\***


\*Correspondence: bc2vamar@uco.es; gasman@inci-cnrs.unistra.fr

#### **Edited and reviewed by:**

Hubert Vaudry, University of Rouen, France

**Keywords: secretion, neuroendocrine cells, membrane trafficking, large dense core vesicles, regulated exocytosis, endocytosis, super-resolution microscopy**

The regulated secretory pathway shared by excitable cells, including neurons and neuroendocrine cells is an intricate process that comprises multiple, tightly regulated steps. After their synthesis in the endoplasmic reticulum, hormones and neuropeptides have to be sorted and packed into large dense core vesicles (also named secretory granules) in the Golgi apparatus. Granules are transported toward the plasma membrane in a cytoskeleton-dependent manner and mature into competent organelles for secretagogueinduced exocytosis. Granules are then tethered to the plasma membrane, docked, and primed, before finally releasing their contents after fusing with the plasma membrane. To ensure that neurotransmission and neuroendocrine secretion operate correctly, all these steps must be tightly regulated and coordinated both spatially and temporally. Currently, when the field of intracellular trafficking has been honored by the 2013 Nobel Prize in Physiology or Medicine (awarded to James Rothman, Randy Schekman, and Thomas Südhof for their pioneering works on vesicular transport), this issue of Frontiers in Neuroendocrine Science is aimed to providing an up-to-date overview of the cellular and molecular mechanisms governing the regulated secretory pathway in neuroendocrine cells. Reviews presented here are widely covering this topic, from the architecture of the organelle involved in secretory cargo processing and sorting, the biogenesis of secretory granules, their specific transport toward the plasma membrane to the late steps of exocytosis, and the secretory granule membrane recapture.

Early stages of the secretory pathway have been discussed by two groups. Emma Martinez-Alonso and colleagues discuss the Golgi complex architecture, as well as the regulatory proteins that govern extra- and intra-Golgi transport, and the still controversial, but not mutually exclusive, theoretical models proposed to explain cargo progression through the Golgi stack (1). The group of Richard Mains discusses the specific roles of cytosolic adaptor proteins such as AP-1A, PACS-1, and GGAs in the assembly and maturation of secretory granules (2).

Closer to the cell surface, several groups discuss the molecular mechanisms regulating the late stage of exocytosis. The group of Frédéric Meunier reviews recent insights of the role of the cortical acto-myosin network (3), whereas the role of different tethering and priming factors such as CAPS, Munc13, and Doc2 proteins is described by the groups of Ury Ashery and Tom Martin (4,5). Paanteha Moghadam and Meyer Jackson review how various synaptotagmins regulate fusion pore kinetics and control the mode of release (6). Lipids have emerged as key players of the regulated exocytosis and the group of Nicolas Vitale presents an overview on the diverse roles that lipids play in defining exocytotic sites, both by affecting membrane topology and by regulating secretory vesicle priming and fusion (7).

Finally, exocytosis cannot exist without a compensatory membrane intake process (i.e., endocytosis), which allows recycling of granule components and maintains organelle integrity. The groups of Stéphane Gasman and Alla Rynditch discuss the mechanisms that coordinate clathrin-mediated compensatory endocytosis with exocytosis, highlighting the specific role of the intersectin family of scaffold proteins in exocytosis and endocytosis (8, 9). The group of Ana-Maria Cardenas reviews the pleiotropic role of the mechano-GTPase dynamin-2, on intracellular membrane fission and fusion events, vesicle traffic, and cytoskeleton dynamics, as well as the impact of dynamin-2 mutations on the correct functioning of the secretory pathway (10).

On a more physiological point of view, Maria-Luisa Durán-Pasten and Tatiana Fiordelisio present an example of how pituitary gonadotrophs receive and transduce extracellular signals to promote luteinizing (LH) and follicle-stimulating (FSH) secretion, highlighting the tremendous plasticity of the system for adapting to different physiological demands (11). Wei-Jye Lin and Stephen Salton report that single nucleotide polymorphisms in genes encoding secreted proteins are associated with neuropsychiatric or endocrine/metabolic disorders (12). Finally, Jennifer Fitch-Tewfik and Robert Flaumenhaft demonstrate how the regulated secretory pathway is similar in mast cells compared to neuroendocrine cells from the adrenal gland (13), and Burton Dickey's group describes the regulatory mechanism of mucin secretion in a non-neuroendocrine cell model (14).

On a more technical point of view, recent improvements in detection technologies, especially in optical microscopy, continually push the limits of sensitivity and resolution. The groups of Colin Rickman and Ute Becherer discuss how advances over the last decade in fluorescence microscopy provided spatial and temporal details on the subcellular organization of the molecular machinery governing the regulated secretory pathway (15, 16), whereas the group of Rory Duncan describes how the combination of new imaging approaches with super-resolution microscopy and novel calcium indicators is appropriate for accurate study of

voltage-gated calcium channel locations, interactions, dynamics, and composition in living cells (17).

Collectively, this compilation of reviews intends to illustrate the recent progress made to understand the complex regulation of the granule secretory pathways in neuroendocrine cells. We are grateful to all the authors who have contributed to this Research Topic and to the dedicated reviewers who helped us reaching the highest quality standards.

#### **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: 10 March 2014; accepted: 24 March 2014; published online: 08 April 2014. Citation: Vazquez-Martinez R and Gasman S (2014) The regulated secretory pathway in neuroendocrine cells. Front. Endocrinol. 5:48. doi: 10.3389/fendo.2014.00048 This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2014 Vazquez-Martinez and Gasman. 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.*

# Morpho-functional architecture of the Golgi complex of neuroendocrine cells

#### **Emma Martínez-Alonso<sup>1</sup> , Mónica Tomás <sup>2</sup> and José A. Martínez-Menárguez <sup>1</sup>\***

<sup>1</sup> Department of Cell Biology and Histology, Medical School, University of Murcia, Murcia, Spain

<sup>2</sup> Department of Human Anatomy and Embryology, Medical School, Valencia University, Valencia, Spain

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Maite Montero-Hadjadje, INSERM U982, France David Cruz-Garcia, Centre for Genomic Regulation, Spain Gustavo Egea, University of Barcelona, Spain

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

José A. Martínez-Menárguez, Department of Cell Biology and Histology, Medical School, University of Murcia, 30100 Murcia, Spain. e-mail: jamartin@um.es

In neuroendocrine cells, prohormones move from the endoplasmic reticulum to the Golgi complex (GC), where they are sorted and packed into secretory granules. The GC is considered the central station of the secretory pathway of proteins and lipids en route to their final destination. In most mammalian cells, it is formed by several stacks of cisternae connected by tubules, forming a continuous ribbon. This organelle shows an extraordinary structural and functional complexity, which is exacerbated by the fact that its architecture is cell type specific and also tuned by the functional status of the cell. It is, indeed, one the most beautiful cellular organelles and, for that reason, perhaps the most extensively photographed by electron microscopists. In recent decades, an exhaustive dissection of the molecular machinery involved in membrane traffic and other Golgi functions has been carried out. Concomitantly, detailed morphological studies have been performed, including 3D analysis by electron tomography, and the precise location of key proteins has been identified by immunoelectron microscopy. Despite all this effort, some basic aspects of Golgi functioning remain unsolved. For instance, the mode of intra-Golgi transport is not known, and two opposing theories (vesicular transport and cisternal maturation models) have polarized the field for many years. Neither of these theories explains all the experimental data so that new theories and combinations thereof have recently been proposed. Moreover, the specific role of the small vesicles and tubules which surround the stacks needs to be clarified. In this review, we summarize our current knowledge of the Golgi architecture in relation with its function and the mechanisms of intra-Golgi transport. Within the same framework, the characteristics of the GC of neuroendocrine cells are analyzed.

**Keywords: golgi complex, neuroendocrine cells, morphology, transport vesicles, tubules, intra-golgi transport**

#### **INTRODUCTION**

A century ago the Italian anatomist Camillo Golgi described a new organelle that nowadays bears his name, the Golgi apparatus or Golgi complex (GC) (Golgi, 1898). Using a silver impregnation method, the "black reaction," he found a reticular structure in neurons that he called "apparato reticolare interno." Due to the difficulties and variability inherent to the technique, it was not clear for decades whether this structure was anything more than an artifact. The electron microscope clearly demonstrated that, the GC is indeed a real organelle,which is composed of flattened cisternae surrounded by tubules and vesicles (Dalton and Felix, 1956). These first ultrastructural images obtained from ultrathin sections showed the exceptional complexity of the organelle and, consequently, high voltage electron microscopy and stereology were used to obtained 3D information (Rambourg et al., 1974). Ultrastructural immunocytochemical methods provide great impetus to the morpho-functional analysis of the GC through the precise location of key molecular components (Rabouille and Klumperman, 2005). Electron tomography has increased our knowledge of the 3D architecture of the GC (Ladinsky et al., 1999). Another advance has been the use of correlative light-electron microscopy, whereby cell organelles are visualized first by light microscopy in living cells transfected with fluorescent proteins, and then the

same structures are identified under the electron microscope (Polishchuk et al., 2000; Mironov et al., 2008; van Rijnsoever et al., 2008). The combination of immunoelectron microscopy and electron tomography is a powerful approach for scrutinizing the secrets of this organelle (Zeuschner et al., 2006). In parallel to morphological approaches, biochemical and genetic analyses have described in detail the molecular machineries operating in the secretory/endocytic pathways.

The GC has two main functions. The first is the posttranslational modification of proteins and lipids arriving from the endoplasmic reticulum (ER), mainly their glycosylation. The second function is the concentration, packing, and export of these modified products to the final destination in or outside of the cell. Thus, the GC is at the same time an efficient glycan factory and a post office. Helping to carry these functions is a surprisingly complex array of membranes equipped with an accurate machinery. Despite the large volume of incoming and outgoing traffic, it is able to maintain its architecture, although it is also flexible enough to disassemble and reassemble under certain conditions, such as mitosis. In neuroendocrine cells, prohormones are frequently glycosylated and proteolytically processed before being sorted into secretory granules (reviewed in Vázquez-Martínez et al., 2012). A summary of the current knowledge on the morphology of this organelle and early steps of the secretory pathway (i.e., ER-Golgi and intra-Golgi transport) is given below. Post-Golgi events, such as secretory granule formation, and other aspects of the Golgi functions are omitted in this description and can be found in excellent reviews elsewhere.

#### **MORPHOLOGY OF THE GOLGI RIBBON**

In mammalian cells, this organelle consists of a pile of flat, disklike membranes, the cisternae (**Figures 1**–**3**). This pile of cisternae is called the Golgi stack. The number of cisternae per stacks varies between different organisms but is characteristic of each species, usually numbering between 3 and 11 (Rambourg and Clermont, 1997). The diameter of the cisternae is also cell type-dependent, and is usually 0.5–1µm (Weidman et al., 1993). The lumen of the cisternae is usually quite narrow (10–20 nm), allowing the interaction of the glycosylation enzymes present in its membranes and the cargo. Typically, cisternae are of uniform thickness in the central part but dilated near the lateral rims. In secretory cells, cisternae may show distensions filled with a material of low electron density, known as pro-secretory granules. These elements can be observed in the trans side alone (prolactin cells) or in all the cisternae, although, in the latter case, their size increases in the trans direction (Rambourg and Clermont, 1997). Cisternae may show small (fenestrae) and large holes, which are sometimes aligned to form wells (Ladinsky et al., 1999). Usually, such wells are filled

with vesicles and exposed to both the cis and lateral sides of the stacks (Ladinsky et al., 1999). Fenestrations are less abundant in the medial cisternae of the stack and increase in both cis and trans directions.

Early histochemical and immunoelectron microscopical analysis demonstrated that the Golgi stack is polarized. Thus, based on the distribution of resident proteins, the Golgi stack can be divided into three regions: cis, medial, and trans. Glycosyltransferases, sugar nucleotide transporters, and many other Golgi proteins are preferentially found in one of these sub compartments. However, the resident proteins are not restricted to a few cisternae, but exhibit a gradient of concentration through out the cisternae of the stack, suggesting a state of dynamic equilibrium (Rabouille et al., 1995).

The GC of most mammalian cells is formed of several stacks that are laterally interconnected by tubules forming the Golgi ribbon (Rambourg et al., 1979; Rambourg and Clermont, 1997; Marsh et al., 2001) (**Figures 1** and **2A**). Thus, although it is not always apparent, the stacks observed in typical electron microscopic images of the Golgi area, belong to the same ribbon. Due to their appearance, the pile of cisternae and the lateral tubular network are called the compact and non-compact zones, respectively. Usually, the tubules connect cisternae located in the same positions in the respective stacks. However, heterotypic connections, even between the cis-most and trans-most cisternae of adjacent stacks,

**FIGURE 1 | Golgi structure and transport carriers in secretory cells.** The Golgi ribbon is formed by adjacent Golgi stacks (Gs) separated by non-compact regions (Nc) containing tubules and vesicles. Golgi stacks are connected to the cis and trans Golgi networks (CGN and TGN). Newly synthesized cargo leaves the endoplasmic reticulum (ER) by COPII-coated vesicles (black). COPI-coated (red) vesicles mediate recycling from the Golgi and the ERGIC. Transport carriers at the TGN include clathrin-coated vesicles (blue), regulated secretory granules (Sg), and the poorly understood constitutive secretory carriers (orange) Clathrin and COPI coats are also associated to secretory granules. Heterotypic and homotypic tubular connections between cisternae may be involved in anterograde and/or retrograde intra-Golgi transport.

**FIGURE 2 | Structure of the Golgi ribbon.** Electron micrographs of Epon-embedded PC12 cells. **(A)** The Golgi stacks (Gs) are surrounded by tubule-vesicular elements (asterisks). Arrowheads point to the non-compact area of the ribbon which connects the stacks laterally. **(B)** The cis and trans sides of the stacks are identified by the presence of ER exit sites (arrowhead) and secretory granules (Sg), respectively. Gs, Golgi stack; L, Lysosome; N, Nucleus. Bars, 200 nm.

are abundant in some cell types (Rambourg and Clermont, 1997; Vivero-Salmerón et al., 2008). Frontal views of cisternae point to lateral networks of tubules emerging from the fenestrated rims (Weidman et al., 1993). Some of these tubular membranes fuse with the same cisterna, whereas others grow toward the cytoplasm. A single cisterna may have tubules oriented toward the cis and trans sides (Ladinsky et al., 1999). Many tubules, however, extend laterally and fuse with tubules from adjacent Golgi stacks, forming the tubular network that bridges adjacent stacks.

Usually, the ribbon is located close to the nucleus, around the microtubule organizing centers (MTOC), and the spatial configuration of the GC is closely related to the arrangement and orientation of the microtubules. The maintenance of the Golgi ribbon strongly depends on the microtubules and the action of motor proteins (Egea and Rios, 2008). Microtubule de-polymerizing agents such as nocodazole induce fragmentation of the ribbon into mini stacks (Storrie et al., 1998). In fact, the GC acts as a secondary MOTC. Golgi organization also depends on the actin cytoskeleton (Egea et al., 2006).

The structure of the ribbon is also supported by the socalled Golgi matrix, a ribosome-free area surrounding the cisternae (Xiang and Wang, 2011). This matrix can be visualized as small fibers connecting the cisternae (Mollenhauer and Morre, 1998) and also connecting the Golgi membranes and transport vesicles (Orci et al., 1998). The matrix is formed by structural proteins, some of them identified as auto-antigens and others isolated from detergent-insoluble salt-resistant Golgi fractions (Slusarewicz et al., 1994). These components include Golgi reassembly stacking proteins (GRASPs) and golgins (Xiang and Wang, 2011). These proteins are very dynamic and cycle between membrane-associated and a cytoplasmic forms.

The morphology of the GC (the number of cisternae per stack, the number of fenestrations, the complexity of associated tubulevesicular elements, etc.) is cell type specific and depends on the activity of the cell. The level of cargo reaching the GC is an important factor in Golgi appearance. In general, when the input of cargo is low, the GC decreases in size and becomes larger when the synthetic activity is stimulated (Clermont et al., 1993; Taylor et al., 1997; Aridor et al., 1999; Glick, 2000). The relationship between cell activity and Golgi organization was clearly shown in prolactin cells. When the activity of these cells is reduced by removing the litters from lactating rats, the number of cisternal fenestrations and peri-Golgi vesicles increases concomitantly with a reduction in the number of Golgi tubules and mature secretory granules (Rambourg et al., 1993).

#### **CIS AND TRANS GOLGI NETWORK**

The Golgi stack is flanked by two tubule-vesicular networks located at the cis and trans sides, which represent the entry and exit sides of the stack, respectively (**Figure 1**). At the cis-side, the cis-Golgi network (CGN) is involved in ER-Golgi transport. At the trans side, the trans Golgi network (TGN) receives and packs proteins and lipids that have traversed the stack and deliver them to their final destinations.

The CGN is formed of tubules connected to the first Golgi cisterna (Sesso et al., 1994; Rambourg and Clermont, 1997). This element is well developed in some cell types such as spermatids (Vivero-Salmerón et al., 2008) but less so in others, such as prolactin cells (Rambourg and Clermont, 1997). In early electron microscopical studies, these tubules were selectively identified by using reducing osmic for prolonged times. The functional relationship of this tubular network connected to the stack (the CGN) and ER-derived pre-Golgi elements [intermediate compartment (IC), see below] remains to be established.

Trans Golgi network is involved in the final steps of protein glycosylation and maturation and in the sorting of products to the apical and basolateral plasma membranes, early and late endosomes, and secretory granules (Griffiths and Simons, 1986; Keller and Simons, 1997; De Matteis and Luini, 2008). In neuroendocrine cells, the secretory proteins are concentrated in secretory granules that can be rapidly released after stimulation (Kelly, 1985). This

**FIGURE 3 | Coated vesicles and buds.** Electron micrographs of cryosections of PC12 cells. Using this methodology, membranes appear negatively stained, whereas coats are identified as electron dense areas around vesicles and buds. **(A)** COP- (arrows) and clathrin-coated (arrowhead) vesicles are observed in the lateral and trans Golgi sides,

respectively. Note the different thickness of these coats. **(B)** COPII-coated bud associated to the endoplasmic reticulum (arrowhead). **(C)** COPI-coated bud in the lateral rim of cisterna (arrowhead). COPII- and COPI-coated buds are identified by their locations because the thickness of these coats is identical. Bar, 200 nm.

regulated secretory pathway co-exists with the constitutive secretory pathway that is common to all cell types (Arvan and Castle, 1998). The sorting process can take place in the TGN (sorting-forentry) or in immature secretory granules (sorting-by-retention) (Borgonovo et al., 2006). Different carriers and associated molecular machineries may be used for different routes (Traub and Kornfeld, 1997). Structurally, the TGN is formed of a large tubular network in continuity with the trans-most cisterna of the Golgi stack (Griffiths et al., 1985; Clermont et al., 1995). This is not always evident and, in some cell types, TGN can be found some distance from the stack (Clermont et al., 1995). The TGN can vary significantly in both size and composition, depending on the

amount and type of cargo, and is reduced or absent in cells producing secretory granules in contrast with cells with an extensive lysosomal system (Clermont et al., 1995).

#### **VESICLES**

The Golgi stack is surrounded by a high number of 50–100 nm vesicles, the smallest ones at the cis-side and lateral rims, and the largest ones at the trans side (Marsh et al., 2001). Many of these vesicles have a coat, an electron dense proteinaceous layer on the cytoplasmic leaflets of their membranes (**Figure 3**). These coats are also found in certain areas of the secretory/endocytic compartments, which represent forming vesicles. Three types of coat complex (COPI, COPII, and clathrin) have been identified and characterized. COPI- and COPII-coated vesicles are almost identical under the electron microscope. The overall size and coat thickness are 50–60 and 10 nm, respectively. COPII-coated buds are restricted to the ER, while COPI-coated buds are found in pre-Golgi and Golgi membranes (Griffiths et al., 1985; Oprins et al., 1993; Orci et al., 1997; Martinez-Menárguez et al., 2001; Rabouille and Klumperman, 2005). The trans-most cisternae of the Golgi and the TGN contains another type of coat, the clathrin coat (Pearse and Robinson, 1990) which is also found in the plasma membrane and endosomes. Clathrin-coated vesicles are unambiguously identified by their size (100 nm) and the thickness of the coat (18 nm) (Orci et al., 1984; Heuser and Kirchhausen, 1985; Kirchhausen et al., 1986; Oprins et al., 1993; Ladinsky et al., 2002). Interestingly, clathrin and COPI coats are also observed in forming secretory granules (Martínez-Menárguez et al., 1999).

Vesicles represent the best known type of transport intermediate. A huge amount of information has been accumulated on the molecular machinery involved in regulation of intercompartmental transport in the secretory pathway. Most data refer to vesicles as transport carriers but it can be assumed that the same or similar mechanisms operate in other carriers. While the formation of vesicles and the selection of cargo depend on the coat machinery, the specific targeting and fusion of the carriers with the target membranes depend on tethering factors, Rab and SNARE (soluble *N*-ethylmaleimide-sensitive factor attachment protein receptors) proteins, and other accessory proteins (Bonifacino and Glick, 2004).

SNARE proteins are involved in docking and the specific fusion of transport intermediates with the target membranes (Bonifacino and Glick, 2004; Hong, 2005; Jahn and Scheller, 2006). The SNAREs associated with vesicles (or other transport intermediates) and target membranes have been named v- and t-SNARE, respectively. SNAREs have also been divided into R- and Q-SNAREs, according to the central residue (R/Gln or Q/Arg, respectively) of the SNARE domain, a conserved region of 60– 70 residues found in all members of this family. Commonly, v-SNARE and t-SNARE are R-SNARE and Q-SNARE, respectively. Interaction between one v-SNARE and two/three t-SNAREs induces the formation of the trans-SNARE complex, which catalyzes the fusion of the membranes. After fusion, a cis-SNARE complex is formed in the target membrane, which is later disassembled by the action of the cytosolic proteins α-SNAP (soluble *N*-ethylmaleimide-sensitive factor attachment protein) and the ATPase NSF (*N*-ethylmaleimide-sensitive factor). Now, v-SNARE can be transported back to the donor compartment to be reused. SNARE proteins are sufficient to drive membrane fusion so that they are considered the minimal membrane fusion machinery. Two SNARE complexes have been implicated in intra-Golgi transport (Malsam and Söllner, 2011). One complex is formed of v-SNARE GS15 and the t-SNAREs syntaxin5, GOS28, and Ykt6, and is involved in COPI-dependent intra-Golgi transport. The second complex is formed of v-SNARE rBet1 and the t-SNAREs syntaxin 5, membrin (GS27), ERS24/Sec22. This second complex has also been implicated in ER-to-Golgi and intra-Golgi transport (Volchuk et al., 2000). SNARE complex assembly is regulated by SM (Sec1/Munc18) proteins, a family of cytosolic proteins. Sly1 I

is the only member of this family operating in the ER-Golgi area (Laufman et al., 2009), whereas Munc18-1 is involved in the exocytosis of dense-core granules in neuroendocrine cells (Burgoyne et al., 2009).

Rab proteins are a family of small GTPases that regulate membrane transport by recruiting effectors, including sorting adaptors, tethering factors, kinases, phosphatases, and motor proteins (Jahn and Scheller, 2006; Stenmark, 2009). Rab proteins switch between an active form (GTP-bound) and a cytosolic inactive form (GDPbound). They have been implicated in vesicle budding, uncoating, mobility, and transport. Rab proteins in the GTP-bound form are reversibly associated with membranes by geranylgeranyl groups. The replacement of GDP by GTP is facilitated by guanine nucleotide exchange factors (GEFs), and their low intrinsic GTPase activity is enhanced by GTPase-activating proteins (GAP). Rab protein is a large family, including more than 60 members in humans, which are specifically associated with distinct compartments and transport events (Stenmark, 2009). Interestingly, the specific membrane recruitment of Rabs has recently been demonstrated to depend on the activity of GEFs (Blümer et al., 2013). The Golgi-associated Rab family playing a key role in Golgi maintenance and functioning includes Rab1, Rab2, Rab6, Rab33B, Rab18, and Rab43 (Liu and Storrie, 2012). Rab3 (A–D), Rab11, Rab18, Rab26, Rab27 (A,B), and Rab37 are involved in regulated secretion (Fukuda, 2008; Stenmark, 2009). Rab3A, Rab11, Rab18, and Rab27A regulate exocytosis in neuroendocrine cells by interacting directly with secretory vesicles (Vázquez-Martínez and Malagón, 2011). A few GEFs and GAPs for secretory Rabs have been identified. The specific role of secretory Rabs in the formation and maturation of the secretory vesicles, and their docking and fusion with the plasma membrane, remains controversial.

Tethering factors are a group of membrane-associated proteins or multi-subunit complexes that link transport vesicles with the target membranes to ensure correct docking and fusion. In addition to tethering, they play a role in Golgi stacking and form the Golgi matrix. They have been classified into three classes: oligomeric complexes that work as Rab effectors and bind SNARE, oligomeric complexes that function as GEFs for Rab proteins and, finally, coiled-coil tethers (Sztul and Lupashin, 2009). Golgiassociated members of the last group are called golgins. The golgin family includes p115, a protein believed to be involved in tethering COPII vesicles to pre-Golgi membranes, transport from these elements to the cis-Golgi and intra-Golgi transport. GM130 is another coiled-coil protein associated with the cis-Golgi. GM130 and p115 are also components of the Golgi matrix. Apart from their role as tethers, they are involved, together with other members of the golgin and GRASP families such as giantin and GRASP65, in maintaining the stacked morphology of the cisterna and the Golgi ribbon (De Matteis et al., 2008). NECC1 (neuroendocrine long coiled-coil protein 1) is a new component of this family, which is mainly present in neuroendocrine tissues (Cruz-Garcia et al., 2007). NECC1 is the first long coiled-coil protein described that has a role as a negative modulator of the regulated secretion in neuroendocrine cells (Cruz-García et al., 2012). Dsl1, conserved oligomeric Golgi (COG), and transport protein particle (TRAPP) are multi-subunit complexes associated with the GC (Sztul and Lupashin, 2009). Dsl1 is a three unit complex involved

in Golgi-to-ER transport,where it acts to tether COPI vesicles. The COG complex formed of eight subunits (Cog1–8) found at the cis and medial Golgi cisternae is believed to be involved in transport to the Golgi and the intra-Golgi recycling of Golgi resident proteins (Miller and Ungar, 2012). TRAPP is another multi-subunit complex that works as tethering factor (Sacher et al., 2008). At least two TRAPP complexes exist in mammals. TRAPPI tethers COPIIcoated vesicles and mediates ER-to-Golgi transport (Barrowman et al., 2010). TRAPII has been involved in intra-Golgi transport, post-Golgi traffic, endosome-to-Golgi, and autophagy (Yu and Liang, 2012). They work as GEF for Rab1 and the activation of this GTPase might recruit other tethers (Sztul and Lupashin, 2009).

Some lipidic species, such as diacylglycerol, phosphatidic acid and lysophosphatidic acid, and enzymes associated with their metabolism play a key role in carrier formation by regulating the curvature of membranes. Diacylglycerol has been implicated in the formation of post-Golgi carriers and is necessary to recruit protein kinase D, a regulator of the fission of transport carriers (Bard and Malhotra, 2006). Diacylglycerol has also seen to be involved in Golgi-to-ER retrograde transport mediated by tubules (Fernández-Ulibarri et al., 2007). Phosphatidic acid, which is generated by phospholipase D2, is involved in COPI vesicle formation (Yang et al., 2008). This phospholipid is necessary to maintain the structure of the GC and secretion in neuroendocrine cells (Siddhanta et al., 2000). Lysophosphatidic acid generated by the enzyme phospholipase A2 has been involved in the retrograde transport mediated by tubules (de Figueiredo et al., 2000; Brown et al., 2003). This enzyme is involved in the formation of tubular continuities between cisternae (San Pietro et al., 2009) and tubular transport intermediates at the TGN (Schmidt et al., 2010). Interestingly, two lipid-modifying enzymes, lysophosphatidic acid acyltransferase γ and phospholipase A2-α, which promote or inhibit COPI fission, respectively, work together, regulating the morphology of Golgi carriers (vesicles vs. tubules) (Yang et al., 2011).

#### **THE GOLGI COMPLEX OF NEUROENDOCRINE CELLS**

The hypothalamus-hypophysis system is the most important and well-known neuroendocrine system. Important clues on Golgi functioning have been obtained by studying the GC of the pituitary gland. In addition, neuroendocrine cell lines, such as corticotropic tumor AtT20, pheochromocytoma PC12, and frog melanotrope cells, provide important clues as regards secretory granule formation and regulation (Morvan and Tooze, 2008). Thus, the discovery that immature secretory granules originated from Golgi cisternae was made in mammotroph cells (Farquhar, 1961). In the same cell type it was also found that that there is a step during which secretory material is condensed in the GC (Smith and Farquhar, 1966). AtT20 cells have been used to show that there are different routes from the GC to the plasma membrane (Gumbiner and Kelly, 1982). However, detailed morphological studies of the Golgi ribbon of neuroendocrine cells, with some exceptions, are scarce. The morphology of the GC and the formation of secretory granules in prolactin cells were described in early microscopic studies. In these cells, the Golgi ribbon forms a hollow sphere in the perinuclear area. The stacks of prolactin cells have four to five (mostly flattened) cisternae and show a reduced CGN and TGN (Rambourg and Clermont, 1997). The trans cisternae show distensions

that are gradually transformed into tubular progranules at the trans face and are finally condensed into compact polymorphous granules (Clermont et al., 1993). As indicated above, the morphology of this compartment is strongly dependent on cell activity (Rambourg et al., 1993). Many other cells of the anterior pituitary gland have not been extensively studied, although some of these cell types have a spherical GC, with cis and trans side representing the outer and inner part of this sphere, respectively (Watanabe et al., 2012).

#### **ER-TO-GOLGI TRANSPORT**

Newly synthesized proteins and lipids in general, and prohormones in neuroendocrine cells, exit the ER in specific places of this compartment, called ER exit sites (ERES). These places are formed by tubular buds of different length containing a COPII coat (Sesso et al., 1994; Bannykh et al., 1996; Zeuschner et al., 2006), which assists in the deformation of ER membranes into vesicles containing membrane and soluble cargo *en route* to the GC (Barlowe et al., 1994). The COPII coat complex is formed of five soluble proteins: Sec23, Sec24, Sec13, Sec31, and Sar1 (Barlowe et al., 1994; Bickford et al., 2004). Formation of the COPII coat begins with the recruitment of the GTPase Sar1 in ER membranes. This binding depends on the activation of this GTPase by Sec12, a GEF present in ER membranes (Jensen and Schekman,2011). During this process some proteins are selectively recruited into COPII vesicles (Barlowe, 1998, 2003), whereas others enter unspecifically, a process known as bulk flow (Martínez-Menárguez et al., 1999). Sorting of the transmembrane cargo depends on Sec24 (Mancias and Goldberg, 2008). The soluble cargo present in the lumen of the ER binds to cargo receptors, such as ER-Golgi IC (ERGIC)- 53, p24, and the Erv families (Szul and Sztul, 2011). These cargo receptors cycle between the ER and Golgi and also are included in recycling COPI vesicles. The p24 family comprises 8–10 isoforms and a subset of these proteins is up-regulated, together with proopiomelanocortin, after the activation of neuroendocrine frog melanotrope cells (Strating et al., 2011). The size of COPII vesicles is regulated by ubiquitination of Sec31, allowing the formation of large COPII vesicles (Jin et al., 2012). After formation, COPII vesicles are quickly uncoated and fuse to each other to form the socalled ERGIC, IC, or vesicular-tubular clusters (VTCs), the first of these being the most used (Hauri and Schweizer, 1992; Farquhar and Hauri, 1997). This compartment was initially identified as tubule-vesicular membranes in which the cargo accumulates when cells are cultured at low temperatures (15˚C) (Saraste and Kuismanen, 1984; Schweizer et al., 1990). This compartment is located close to the GC and is also distributed throughout the cell (Lotti et al., 1992; Klumperman et al., 1998), and is associated to ERES (Bannykh et al., 1996). ERGIC is formed by vesicles and tubules, sometimes branched (Bannykh et al., 1996). Although adjacent to the ERES, ERGIC is an independent compartment and there is no continuity between them (Bannykh et al., 1998). ERGIC membranes do not have COPII coats but another type of coat complex, the COPI coat, which is involved in retrograde transport (see below). Thus, the presence of these coats can be used to discriminate between these closely related compartments. Whether the ERGIC is a stable compartment or a transitory element moving toward the GC is still a matter of debate (Ben-Tekaya et al., 2005). The fact that all the proteins associated with this compartment cycle between the ER and Golgi argues against the idea that it is a stable compartment. ERGIC-53, a type I transmembrane protein of the lectin family, is the prototypical marker of this compartment (Zhang et al., 2009). However, *in vivo* experiments showed that ERGIC-53 is located in long-lived stationary elements connected by highly mobile elements (Ben-Tekaya et al., 2005), supporting the view that ERGIC is a stable compartment. ERGIC elements or ERGIC-derived carriers move to the Golgi area along the microtubule track guided by dynein motors, where they may fuse to each other to form the CGN or, conversely, fuse with a pre-existing cisterna (Presley et al., 1997).

#### **GOLGI-TO-ER TRANSPORT**

Organelle identity is determined by its composition, and its functional integrity is due to its ability to maintain this composition despite the continuous traffic between compartments. An important mechanism involved in this process is retrograde transport. In the early secretory pathway, ERGIC is the first place where this process occurs. Here, the presence of COPI coats ensure the recycling of proteins to the ER while anterograde cargo is separated and concentrated (Scales et al., 1997; Klumperman et al., 1998; Martínez-Menárguez et al., 1999; Shima et al., 1999; Stephens et al., 2000). The function of the COPI coats in the Golgi-to-ER retrograde transport of soluble and membrane proteins has been convincingly demonstrated (Letourneur et al., 1994). However, there are retrograde routes to the ER independent of COPI (Girod et al., 1999). It is also possible that a population of COPI vesicles is involved in anterograde transport across the Golgi stack (Orci et al., 1997).

COPI-coated vesicles are found in peripheral and central ERGIC elements, at cis and lateral Golgi sides (Oprins et al., 1993) and, occasionally, at the trans Golgi side/TGN (Martínez-Menárguez et al., 1996). COPI-coated buds are observed at the lateral rims of cisternae, decreasing in number in a cis to trans direction (Ladinsky et al., 1999). COPI coats are formed of seven subunits (α, β, β', γ, δ, ε, ζ-COP) assembled in the cytosol, the coatomer, and small GTPase ADP-ribosylation factors (Arf1). Arf1 belong to the Arf family,which is made up of six members in mammals (D'Souza-Schorey and Chavrier, 2006). Arfs 1–5 have been described at the ER-Golgi interface, where they may have redundant functions. Besides, it has been suggested that different pairs of Arfs may be necessary for each transport step (Volpicelli-Daley et al., 2005). Arf1 and 4 and Arfs 3–5 are associated to the cis and trans Golgi sides, respectively (Donaldson and Jackson, 2011). Arf 4 and 5 interact with calcium-dependent activator for secretion (CASP), regulating the formation of neuroendocrine secretory granules (Sadakata et al., 2010). Arf1-GDP is recruited to membranes by p23, a member of the p24 family (Gommel et al., 2001), and, once there, it is activated by a GEF. GTP-bound Arf1 is able to recruit coatomer *en bloc* to membranes. At the ERGIC and Golgi membranes, GBF1 (Golgi-associated BFA-resistant protein) is the major GEF for ARF1 during COPI vesicle formation and is the target of the drug brefeldin A (Kawamoto et al., 2002; Garcia-Mata et al., 2003), while Big1 and Big2 (BFA-inhibited GEF) are described as GEFs for Arf1 in the TGN and endosomes (Ishizaki et al., 2008). ArfGAPs stimulate the GTP hydrolysis of Arf, which

has low intrinsic GTPase activity (Inoue and Randazzo, 2007). It has been postulated that this reaction triggers uncoating (Tanigawa et al., 1993); however, ArfGAP1 has also been described as a coatomer component and also as a sensor of membrane curvature, so the role of this protein is under discussion (Shiba and Randazzo, 2012). ArfGAP1, phosphatidic acid generated by phospholipase D and BARS (brefeldin A-ribosylated substrate) have been implicated in COPI vesicle fission (Yang et al., 2008).

Many transmembrane proteins transported into COPI vesicles bear di-lysine motifs at their C-terminus (Letourneur et al., 1994), including ERGIC-53 (Schindler et al., 1993). The coatomer subunits, α- and β'-COP, directly bind this signal (Jackson et al., 2012). The p24 family of proteins is recruited by direct interaction of their cytoplasmic tail, which contains phenylalanine- and di-lysine-based signals, with γ-COP (Nickel et al., 1997; Béthune et al., 2006). Not all cargo incorporated in COPI vesicles has this sorting signal so that the cargo may require adaptors. This is the case with glycosylation enzymes (Popoff et al., 2011). One example of an adaptor for soluble proteins is the KDEL receptor, a transmembrane protein mostly present in the cis-Golgi and ERGIC, which interacts with the coatomer through a di-lysine motif in the cytosolic tail, whereas the luminal part interacts with soluble proteins bearing a K-D-E-L sequence found in many ER-resident proteins (Semenza et al., 1990; Majoul et al., 2001). In this way, soluble proteins that have escaped from the ER are retrieved.

As indicated above, it is possible that there are several subpopulations of COPI vesicles with different compositions and locations, each carrying out its specific functions (Moelleken et al., 2007). Indeed, γ and ζ-COP present two isoforms localized differently along the GC, enabling the existence of four types of coatomer, which may act in different routes and types of cargo (Popoff et al., 2011). Thus, this coat conforms versatile vesicles that may play different roles at the ER-Golgi interface (Orci et al., 1997; Shima et al., 1999; Malsam et al., 2005).

#### **INTRA-GOLGI TRANSPORT**

Once cargo has reached the GC, the manner in which it moves through the stack remains controversial, and two classical models have been proposed: vesicular transport and cisternal maturation. The first proposes that Golgi cisternae are static so that the cargo must use vesicles to move. The second model proposes that cisternae are dynamic structures that move from the cis to trans Golgi sides. Thus, according to this model, the entire cisterna is the carrier. The postulated roles of COPI vesicles in vesicular and cisternal maturation models are completely different, since they are regarded as being responsible for anterograde and retrograde transport, respectively.

Launched by Palade (1975), the vesicular transport model postulates that cisternae are stable compartments. Thus, cargo must leave one cisterna and move to an adjacent one in order to advance through the stack. This process is mediated by COPI vesicles. For decades this model was widely accepted since it was based on new experimental data, especially the analysis of cell-free systems and immunomicroscopical studies (Rothman and Fine, 1980; Orci et al., 1986; Barlowe et al., 1994). It provided a good explanation for the well-known compartmentalization of the Golgi resident enzymes (Roth and Berger, 1982; Dunphy et al., 1985). Besides, it was strongly supported by the discovery of COPI and COPII vesicles which, it was postulated, act sequentially in the early secretory pathway (Rothman and Wieland, 1996). It was clearly demonstrated that COPII vesicles are involved in ER exit (Barlowe et al., 1994), while COPI were identified in *in vitro* assays as being responsible for anterograde transport between cisternae (Orci et al., 1986). However, this model was questioned when a clear role for COPI vesicles in retrograde transport was demonstrated (Cosson and Letourneur, 1994; Letourneur et al., 1994). Even so, the model remained in force even when two types of COPI vesicle were seen to be involved in anterograde and retrograde transport in the Golgi stack (Orci et al.,1997). The weakest point of this model is that it does not explain how large cargo is transported.

A cisternal maturation model was postulated by Grasse (1957) based on early electron microscopy observations, but it was not until the early 1990s that it was re-considered (Bonfanti et al., 1998; Glick and Malhotra, 1998). According to this model, cisternae are formed at the cis-side of the GC by fusion of ER-derived membranes and then these newly formed cisternae move from the cis to the trans side. This model fits very well with the observation that the Golgi resident enzymes are not strictly compartmentalized through the Golgi stack (Nilsson et al., 1993; Rabouille et al., 1995). Furthermore, this model is able to explain the transport of large cargo, such as procollagen (300 nm rigid rod) (Bonfanti et al., 1998) or algal scales (up to 1.5–2µm) (Melkonian et al., 1991), which do not fit within 50–60 nm vesicles. During cisternal progression, Golgi resident enzymes must be packed into COPI vesicles and transported backwards. Thus, Golgi enzymes are not lost but recycled in a trans to cis direction, maintaining the polarity of the organelle (Glick et al., 1997). Immunocytochemical (Martinez-Menárguez et al., 2001; Mironov et al., 2001) and proteomic (Gilchrist et al., 2006) analyses of these vesicles showed that they are mostly devoid of anterograde transport markers but enriched in glycosylation enzymes. However, other studies on COPI vesicle composition provided conflicting results. Thus, the role of COPI vesicles remains controversial (Cosson et al., 2002). Life cell imaging studies in yeast involving direct real time visualization of cisternal maturation strongly support this model (Losev et al., 2006; Matsuura-Tokita et al., 2006).

Since neither of these models explains all the experimental data, a combination of the same models (dual model) as well as new models (rapid partitioning, kiss-and-run) have been proposed (reviewed in Glick and Luini, 2011). The cisternal maturation model seems to be more efficient, given that it seems easier to transport anterograde cargo using a large carrier (the cisterna) than to use many small vesicles between adjacent cisternae and repeat this process several times until reaching the TGN. However, these models are not mutually exclusive and a combination of both might serve (Pelham and Rothman, 2000). Cisternae may move slowly, whereas vesicles transport anterograde and retrograde cargo more rapidly. Large molecules may use the cisternal maturation mechanism, while small molecules can be transported using vesicular transport (Orci et al., 2000; Pelham and Rothman, 2000). In addition, other mechanisms might operate in this transport step as several new models postulate. According to the rapid partitioning model, the GC operates as a single compartment but contains processing and export domains of differing lipid composition, which allow the specific retention of resident enzymes and cargo proteins (Patterson et al., 2008). The cargo associates with these domains and then leaves the compartment from every level. However, this model has difficulty in explaining many previous observations including the well-known polarized distribution of glycosylation processing enzymes, the progression of procollagen and other cargo across the stack or the role of COPI vesicles. The kiss-and-run model, meanwhile, proposes that two cisternae may fuse to each other through narrow tubules, connecting their lumens transiently, and allowing the transit of anterograde and retrograde cargo before disconnection (Mironov and Beznoussenko, 2012). Given that specific retrograde transport is not necessary in this model, COPI might be involved in the fission process. It is also not clear how all the experimental data available concerning the role of COPI coats in retrograde transport fit this model.

Despite the abundance and development of Golgi-associated tubules, as described above, most models of intra-Golgi transport do not include a role for these elements. Tubules may act as intermediate transport carriers, alone or with vesicles, and in anterograde and retrograde transport (Griffiths, 2000; Mironov et al., 2003; Marsh et al., 2004; Trucco et al., 2004; Martínez-Alonso et al., 2005, 2007; Vivero-Salmerón et al., 2008). In fact, it has recently been found that a similar mechanism involving COPI and lipid-modifying enzymes may regulate the formation of both types of carrier in the GC (Yang et al., 2011). With some exceptions, intercisternal tubular connections are scarce in control cells but increase when secretory activity is stimulated (Marsh et al., 2004; Trucco et al., 2004), supporting the view that tubules may be involved in anterograde transport when there is an excess of cargo. In fact, it has been postulated that all secretory compartments are connected by tubules and the cargo moves along this pathway like food through the gut (Griffiths, 2000). Supporting this idea, although unusual, a Golgi stack formed by a single cisterna arranged helically and direct connections between the Golgi cisternae and the ER have been described (Tanaka et al., 1986). In order to clarify the role of tubules, it is first necessary to determine their composition, which is a difficult task. The number of tubules can be increased by the use of the fungal drug brefeldin A. Nowadays it is well accepted that brefeldin A-induced tubules are involved in Golgi-to-ER transport. Although this system is artificial, it is believed that brefeldin A intensifies a process that occurs naturally (Lippincott-Schwartz et al., 1990). Tubules can also be enhanced by lowering the temperature (Martínez-Alonso et al., 2005, 2007). These tubules exclude anterograde and retrograde cargo but recruit Golgi resident enzymes and a specific set of Rab and SNARE proteins involved in intra-Golgi transport (Martínez-Alonso et al., 2005, 2007). Thus, these induced tubules may be indicative of the recycling mechanisms of Golgi enzymes postulated by the cisternal maturation model.

#### **CONCLUDING REMARKS**

The GC has fascinated scientists for more than a century. Although it is without doubt the most photographed cell structure, it still retains most of its mystery. What is the reason for its beautiful architecture? Why does it form a ribbon in most cells? What are that the functions of the tubular networks? Do vesicles and tubules have specific functions? Is there a single mode of intra-Golgi transport? Do COPI vesicles take part in anterograde transport? Is the ERGIC a real compartment? These and many other questions remain unanswered,which, in itself, is surprising, given the intense research and the large number of research groups working in the field. It is to be hoped that new approaches and research models will help fill the gaps in our knowledge of this beautiful organelle.

#### **REFERENCES**


#### **ACKNOWLEDGMENTS**

The work in this review was supported by grants from the Ministerio de Ciencia e Innovación (Spain) Consolider COAT CSD2009-00016) and Fundación Séneca de la Comunidad Autónoma de la Región de Murcia (04542/GERM/06) to José A. Martínez-Menárguez and a grant from the University of Valencia (UV-INV-AE11-41831) to Mónica Tomás.

regulated secretory pathway in PC12 cells. *Biochem. J.* 443, 387–396.


anterior pituitary gland. *Trans. N. Y. Acad. Sci.* 23, 346–351.


visualized in living yeast. *Nature* 441, 1002–1006.


proteins by exclusion from COPIcoated vesicles. *Cell* 98, 81–90.


apparatus and plasma membrane. *J. Cell Biol.* 148, 45–58.


and the secretory pathway: the case of rab18 in neuroendocrine cells. *Front. Endocrinol.* 2:1. doi:10.3389/fendo.2011.00001


imaged by freeze-etch electron microscopy: views of different membrane coatings involved in tubulation versus vesiculation. *Cell* 75, 123–133.


functions, regulation and relations with diseases. *Histol. Histopathol.* 24, 1193–1204.

**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: 17 January 2013; accepted: 14 March 2013; published online: 28 March 2013.*

*Citation: Martínez-Alonso E, Tomás M and Martínez-Menárguez JA (2013) Morpho-functional architecture of the Golgi complex of neuroendocrine cells. Front. Endocrinol. 4:41. doi: 10.3389/fendo.2013.00041*

*This article was submitted to Frontiers in Neuroendocrine Science, a specialty of Frontiers in Endocrinology.*

*Copyright © 2013 Martínez-Alonso, Tomás and Martínez-Menárguez. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.*

REVIEW ARTICLE published: 14 August 2013 doi: 10.3389/fendo.2013.00101

# Role of adaptor proteins in secretory granule biogenesis and maturation

#### **Mathilde L. Bonnemaison<sup>1</sup> , Betty A. Eipper 1,2 and Richard E. Mains <sup>2</sup>\***

<sup>1</sup> Department of Molecular, Microbial and Structural Biology, University of Connecticut Health Center, Farmington, CT, USA <sup>2</sup> Department of Neuroscience, University of Connecticut Health Center, Farmington, CT, USA

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Juan Ramon Peinado, University of Castilla la Mancha, Spain J. David Castle, University of Virginia Health System, USA

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

Richard E. Mains, Department of Neuroscience, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-3401, USA e-mail: mains@nso.uchc.edu

In the regulated secretory pathway, secretory granules (SGs) store peptide hormones that are released on demand. SGs are formed at the trans-Golgi network and must undergo a maturation process to become responsive to secretagogues. The production of mature SGs requires concentrating newly synthesized soluble content proteins in granules whose membranes contain the appropriate integral membrane proteins. The mechanisms underlying the sorting of soluble and integral membrane proteins destined for SGs from other proteins are not yet well understood. For soluble proteins, luminal pH and divalent metals can affect aggregation and interaction with surrounding membranes.The trafficking of granule membrane proteins can be controlled by both luminal and cytosolic factors. Cytosolic adaptor proteins (APs), which recognize the cytosolic domains of proteins that span the SG membrane, have been shown to play essential roles in the assembly of functional SGs. Adaptor protein 1A (AP-1A) is known to interact with specific motifs in its cargo proteins and with the clathrin heavy chain, contributing to the formation of a clathrin coat. AP-1A is present in patches on immature SG membranes, where it removes cargo and facilitates SG maturation. AP-1A recruitment to membranes can be modulated by Phosphofurin Acidic Cluster Sorting protein 1 (PACS-1), a cytosolic protein which interacts with both AP-1A and cargo that has been phosphorylated by casein kinase II. A cargo/PACS-1/AP-1A complex is necessary to drive the appropriate transport of several cargo proteins within the regulated secretory pathway. The Golgi-localized, γ-ear containing, ADP-ribosylation factor binding (GGA) family of APs serve a similar role.We review the functions of AP-1A, PACS-1, and GGAs in facilitating the retrieval of proteins from immature SGs and review examples of cargo proteins whose trafficking within the regulated secretory pathway is governed by APs.

**Keywords: regulated secretory pathway, maturation, cargo, AP-1, GGA, PACS-1, prohormone**

#### **THE REGULATED SECRETORY PATHWAY**

Neuroendocrine cells synthesize, process, and store peptide hormones so that they are available for secretion upon demand (1). These professional secretory cells devote as much as half of their total protein synthesis to the production of a single hormone (2). The regulated secretory pathway allows intracellular storage of peptide hormones until an external stimulus triggers exocytosis of the secretory granules (SGs) that contain the peptides. Neuroendocrine tumors and metabolic disease are linked to defects in hormone secretion, observed via an increase in circulating hormone levels due to impaired intracellular storage or cellular response. The alterations which result in loss of storage and secretagogue responsiveness are poorly understood.

Peptide hormones are first synthesized as inactive precursors. The signal peptide found at the N-terminus of the preprohormone is recognized by signal recognition particle, which stops translation and directs entry of the nascent preprohormone into the lumen of the endoplasmic reticulum (3); removal of the signal peptide by signal peptidase yields the prohormone. Most proteins which are going to be secreted undergo this step. The endoplasmic reticulum is also the site at which disulfide-bond formation and N-linked glycosylation occur (**Figure 1**) (3, 4, 5). Professional secretory cells have developed specialized sensing mechanisms to avoid triggering the endoplasmic reticulum stress pathway, which can lead to cell death; for example, increased expression of Stress-associated Endoplasmic Reticulum Protein 1 (SERP1) prevents endoplasmic reticulum stress in the anterior pituitary and pancreas (6).

After exiting the endoplasmic reticulum, prohormones are transported to the Golgi apparatus, where additional posttranslational modifications such as oligosaccharide maturation and phosphorylation can occur (**Figure 1**) (5, 7). When they reach the *trans*-Golgi network (TGN), prohormones, and their processing enzymes are concentrated into granules budding from the TGN; these structures presumably represent newly forming SGs (1). These new SGs are immature and must undergo a maturation process before they are capable of secreting peptide hormone in response to secretagogue. Maturation involves remodeling of the immature SG membrane by removal of non-regulated secretory proteins and excess membrane; this process involves clathrincoated vesicles mediated by adaptor proteins (APs) (**Figure 1**)

pathway proteins are red. Examples of post-translational modifications that occur in each organelle are shown. The time-scale is representative of the time spent by hormones and hormone processing enzymes in each organelle of the regulated secretory pathway. Until newly synthesized proteins reach the trans-Golgi network (TGN), proteins from both secretory pathways share the same compartments. Secretory proteins are sorted into secretion of mature SG content. The maturation process involves remodeling of immature SG membranes, generating the constitutive-like secretory pathway (blue); the fate of the vesicles leaving immature SGs is not clear; return to an endosomal compartment from which proteins can be secreted in a constitutive-like fashion or recycled to the TGN is shown. ER, endoplasmic reticulum.

(8, 9), acidification of the lumen, aggregation of content proteins and, at least in some cell systems, fusion of immature SGs (10, 11). The final post-translational modifications needed to generate bioactive peptide hormones occur in immature and mature SGs (12, 13). Mature SGs, which appear dense in the electron microscope and during sucrose density gradient centrifugation, can contain more than 300 mg/ml protein, largely peptide hormones (14). Unlike constitutively secreted proteins, which are found in the extracellular compartment within minutes after exit from the TGN, it takes about 90 min for peptide hormones to go from the TGN to mature SGs (**Figure 1**) (15, 16). In addition, mature SG content can be stored for many days before being secreted into the extracellular compartment in response to a stimulus (**Figure 1**) (17, 18); proteins and peptides stored in the regulated secretory pathway are released at a low rate (basal secretion) even in the absence of secretagogue (19, 20).

The study of immature SGs remains a challenge due to their transient role as intermediates between the TGN and mature SGs. Morphologists describe immature SGs as vacuoles found in close proximity to the TGN which contain dense material surrounded by a loose membrane with a partial clathrin coat, while biochemists distinguish immature SGs from mature SGs by their inability to respond to secretagogue or by their release of incompletely processed newly synthesized products (14,21). Early studies using <sup>35</sup>SO<sup>4</sup> to label sugars and Tyr residues demonstrated high K <sup>+</sup>/Ca2+-stimulated release of <sup>35</sup>SO4-labeled SG components as soon as 15 min after synthesis (16); it is not clear how to relate <sup>35</sup>SO<sup>4</sup> labeling to biosynthetic or endocytic trafficking. In the end, what controls and triggers the formation and maturation of immature SGs remains unclear. Both soluble and membrane proteins destined for the regulated secretory pathway must enter immature SGs when they exit the TGN, but how the trafficking of soluble and membrane proteins is coordinated is still under debate.

#### **FORMATION OF IMMATURE SGs**

Proteins destined for the regulated secretory pathway are sorted in the TGN and in immature SGs. Although the sorting mechanisms are not completely understood, the diverse biophysical and biochemical properties of soluble and membrane proteins suggest that they are targeted to the regulated secretory pathway through

**FIGURE 2 | Immature SG formation at the TGN**. **(A)** Granins and prohormones aggregate in response to the high calcium, mildly acidic conditions in the TGN. Several soluble proteins interact with cholesterol or cholesterol-rich membranes; their sorting into immature SGs depends on this interaction. **(B)** Membrane proteins identified in immature SGs and discussed in the main text are shown; their cytosolic tails can interact with adaptor proteins. The SNARE and synaptotagmin proteins that enter immature SGs

(synaptotagmin IV, VAMP4, syntaxin 6) are replaced during the maturation process. Proteins known to enter immature SGs are drawn approximately to scale (except for the vacuolar proton pump, V-ATPase, and VMATs). PC, prohormone convertase; C, catalytic domain; P, P domain; PTP, protein-tyrosine phosphatase; PHM, peptidylglycine α-hydroxylating monooxygenase; PAL, peptidylglycine α-amidating lyase; CI-MPR contains 15 domains; CPD contains three similar domains; Cys, cysteine-rich domain.

different mechanisms. The TGN is a cellular crossroad; departing proteins can enter vesicles targeted to endosomes, lysosomes, endoplasmic reticulum, or the plasma membrane (1, 22). One of the first studies demonstrating sorting of regulated secretory proteins at the TGN was performed using a cell-free system from PC12 cells, a neuroendocrine tumor cell line: vesicles budding from the TGN contained either heparin sulfate proteoglycan, a soluble protein of the constitutive secretory pathway, or secretogranin II, a soluble protein of the regulated secretory pathway (23).

#### **VIEW FROM THE LUMINAL SIDE**

The mildly acidic pH (pH 6.4) and high calcium (1–10 mM) environment of the TGN can induce aggregation of selected proteins (e.g., secretogranin II, chromogranin B, and prolactin) destined for the regulated secretory pathway, resulting in their segregation from the constitutive secretory pathway (**Figure 2A**) (24–26). Other regulated SG proteins bind specific lipids in the TGN, resulting in their sorting and entry into immature SGs; prohormone convertase 1/3 and prohormone convertase 2 interact with lipid rafts and secretogranin III binds to cholesterol (**Figure 2A**) (27–29). Indeed, chromogranin A, which enhances prohormone

aggregation, interacts at the TGN with secretogranin III, and thus with cholesterol-rich membranes (**Figure 2A**) (28, 30). If the interaction of chromogranin A with secretogranin III is blocked, chromogranin A is not sorted correctly (31). Finally, a role for receptor-mediated sorting of regulated secretory proteins exiting the TGN has been considered. Carboxypeptidase E was proposed as a prohormone sorting receptor because it interacts with the Nterminal region of proopiomelanocortin (POMC), which was previously reported to serve as a sorting domain (15, 32). This conclusion is controversial because the sorting of proinsulin, luteinizing hormone, and follicle stimulating hormone does not depend on carboxypeptidase E (33, 34). Although the sorting of cargo upon binding to a receptor is an attractive concept, SG protein sorting appears to involve multiple processes. Carboxypeptidase E was recently shown to interact with phogrin, a SG membrane protein of the Insulinoma Associated protein 2 (IA-2) family; this interaction involves the pro-region of the luminal domain of phogrin and mature carboxypeptidase E (**Figure 2B**). When one binding partner is missing, the other does not accumulate in SGs, instead localizes to the perinuclear region; the sorting of carboxypeptidase E and phogrin at the TGN is inter-dependent (35).

#### **VIEW FROM THE CYTOSOLIC SIDE**

#### **Sorting signals contributed by SG membrane protein trafficking**

Membrane proteins cannot aggregate as extensively as soluble proteins. The identification of trafficking signals in the cytosolic domains of endocytic cargo led to the postulate that the cytosolic domains of SG membrane proteins would carry signals to ensure their entry into immature SGs (**Figure 2B**). Indeed, deletion or mutation of the cytosolic domain of phogrin results in a decrease in its entry into SGs (36, 37). Similar observations were made for peptidylglycine α-amidating monooxygenase 1 (PAM-1) (**Figure 2B**). PAM-1 is a bifunctional enzyme catalyzing the amidation of glycine-extended peptides, rendering them bioactive. Exogenous expression of a truncated PAM-1 protein lacking its cytosolic domain resulted in its inefficient storage in SGs. Metabolic labeling revealed that 20–40% of the newly synthesized truncated PAM protein entered the regulated secretory pathway, but endocytic trafficking and SG re-entry of the truncated PAM-1 protein were eliminated (38). The cytosolic domain of PAM-1, which is highly phosphorylated, interacts with several cytosolic proteins (39). Two sites in the cytosolic domain of PAM-1 are phosphorylated by casein kinase II (CKII) (**Table 1**); both sites were mutated to Asp (PAM-1/TS/DD), to try to mimic phosphorylation, or to Ala (PAM-1/TS/AA) to prevent phosphorylation. When expressed in AtT-20 corticotrope tumor cells, exogenous PAM-1/TS/DD entered immature SGs more efficiently than PAM-1, while PAM-1/TS/AA did not enter immature SGs efficiently and was degraded (40). This study suggests that phosphorylation of the cytosolic tail of PAM-1 enhances its entry into immature SGs.

Similar conclusions were reached in studies of two different neurotransmitter transporters, Vesicular Monoamine Transporter 2 (VMAT2) and Vesicular Acetylcholine Transporter (VAChT). Using the pH gradient established by the vacuolar proton pump, VMAT2, and VAChT translocate monoamines and acetylcholine, respectively, from the cytosol into the lumen. VMAT2 and VAChT enter two different types of vesicles in PC12 cells: VMAT2 is preferentially targeted to SGs while VAChT is found in synaptic-like

microvesicles (**Figure 2B**) (41). The cytoplasmic domain of the VAChT contains a di-leucine motif with adjacent conserved Glu and Ser residues. Phosphorylation of this Ser by protein kinase C or mutation to a phosphomimetic residue results in a preference for VAChT entry into SGs, rather than into synaptic-like microvesicles (41). Additionally, VMAT2 contains two conserved Glu residues upstream of its di-leucine-like motif (**Table 1**, in blue); mutation of these Glu residues into Ala results in accumulation of VMAT2 in synaptic-like microvesicles (41).

These studies suggest that the sorting of membrane proteins into the regulated secretory pathway requires cytosolic signals and/or luminal/transmembrane signals. The negatively charged region within the cytosolic domain of membrane proteins is probably required for efficient sorting and entry into SGs. Since the cytosolic domain is involved in their sorting, cytosolic proteins must come into play in the formation of immature SGs and protein sorting in the regulated secretory pathway.

#### **ADP-ribosylation factor 1 is required for SG biogenesis**

In the 1990s, the cytosolic protein ADP-ribosylation factor 1 (Arf1), a member of the ADP-ribosylation family (Arf), was shown to promote the formation of immature SGs using cell-free systems from PC12 and GH<sup>3</sup> cells, two neuroendocrine cell lines (42, 43). Arf proteins, which belong to the Ras superfamily of small GTPases, were originally identified as necessary for the ADPribosylation reaction catalyzed by cholera toxin (44). Despite their names, the cellular function of Arfs does not involve ADP ribosylation, but rather membrane trafficking (45). Arf proteins are divided into three classes: class I (Arf1, Arf2, and Arf3), class II (Arf4 and Arf5), and class III (Arf6). Class I and class II Arfs are present at the Golgi and their main function is to regulate Golgi trafficking, while Arf6 functions at the plasma membrane and in the endocytic pathway. Arfs exist in two states: GDPbound Arf is cytosolic and inactive while GTP-bound Arf is membrane-associated and active (**Figure 3**) (45, 46). All Arfs contain an N-terminal myristoylated amphipathic helix, which


Protein, host, and ID are shown. Acidic clusters are in red. CKII sites are in bold. AP-1 and AP-1-like sites are in blue and GGA sites are underlined in green. TMD, transmembrane domain; COOH, C-terminus of the protein; NH2, N-terminus of the protein. The gray shade and italic font was to help the reader distinguish COOH or TMD or NH<sup>2</sup> from the amino acid sequence.

allows their association with membranes. This region is buried in a hydrophobic pocket when GDP is bound, but becomes accessible to membranes when GTP replaces GDP, triggering conformational changes in the switch 1 and switch 2 regions which surround the nucleotide binding site (45–47). A resident TGN protein, cargo, or lipid may interact with GDP-bound Arf1, bringing it near the TGN (**Figure 3**). In the context of COPI vesicles, which are involved in retrograde transport from the Golgi to the endoplasmic reticulum, p23, a transmembrane resident Golgi protein, is thought to recruit GDP-bound Arf (48).

The removal of GDP and binding of GTP is catalyzed by a guanine nucleotide exchange factor (GEF) (**Figure 3**) (45, 46). The two Arf GEFs identified at the TGN are BIG1 and BIG2 (brefeldin A inhibited GEFs). Interestingly, GBF1, a GEF known to be involved in the GDP/GTP exchange of Arf4 and Arf5 in the *cis*-Golgi compartment, has recently been identified on the TGN membrane (49). BIG1 and BIG2 are recruited to the TGN membrane upon binding to GTP-bound Arf4 and Arf5, whose nucleotide exchange was mediated by GBF1 (49). Both BIGs contain a Sec7 domain, which is the active site for the nucleotide switch and the target of brefeldin A, a fungal product. Treatment of neuroendocrine and exocrine cells with brefeldin A blocks the formation and maturation of immature SGs but does not alter mature SG exocytosis (50, 51).

Once nucleotide exchange has occurred, GTP-bound Arf recruits APs, enzyme modifying lipids, and effectors to the membrane before reacting with a GTPase activating protein (GAP) that promotes hydrolysis of GTP to GDP and release of Arf from the membrane (**Figure 3**). The Arf-GAP family is composed of 24 members, each with a GAP domain essential for its activity on Arf (45). GAP activity is modulated by the presence of coat proteins previously recruited by the GTP-bound Arf. When COPI is bound to Arf1, GAP activity is increased (52). Golgi-localized, γ-adaptin ear containing, Arf-binding (GGA) 3 is a coat protein involved in TGN-to-endosome transport. In contrast, when Arf1 recruits GGA3, it blocks the hydrolysis of GTP-bound to Arf1 (53).

Using the cell-free PC12 system, Arf1 was shown to enhance the formation of SGs and constitutive vesicles (42). Moreover, when Arf1 is bound to membranes isolated from PC12 cells, it recruits a set of different proteins, including the adaptor protein 1A (AP-1A) (54, 55). In non-endocrine secretory systems, such as rhoptries in *Toxoplasma gondii* (56), glue granules in *Drosophila* (57), and Weibel–Palade bodies in endothelial cells (58), AP-1A is required for the formation of SGs. Although it is not clear if AP-1A is required for SG formation in neuroendocrine cells, there is evidence that AP-1A is required for SG maturation (**Figure 4A**).

#### **Membranes and SG formation: phosphatidylinositol-4-phosphate and cholesterol**

The appropriate lipid composition is essential for normal TGN function and SG biogenesis. The TGN membrane is enriched in phosphatidylinositol-4-phosphate (PI4P). In mammalian cells, three enzymes that synthesize PI4P from phosphatidylinositol (PI) have been found at the Golgi, PI4KIIα, PI4KIIIα, and PI4KIIIβ; PI4KIIα (59) and PI4KIIIβ (60) have been implicated in export of material from the Golgi to the plasma membrane. PI4KIIα resides primarily in the TGN, endosomes, and SGs (61–64). Indeed, PI4KIIα was found on adrenal chromaffin granule membranes (65–67), cellugyrin-positive Glucose Transporter four vesicles (68), *Drosophila* salivary gland glue granules (69), synaptic

in red), along with the CKII binding site in the FBR. AP-1 and AP-3 interact with the same site in the FBR while GGA3 interacts with a different site. The FBR is also responsible for interactions with cargo. **(C)** Autoinhibition mechanism: the autoinhibition domain binds FBR, preventing PACS-1 interaction with cargo. When CKII phosphorylates Ser<sup>276</sup> of the autoinhibition domain, it inhibits the internal binding and promotes cargo interaction with PACS-1. PP2A, protein phosphatase 2A. **(D)** Each GGA is composed of four domains: VHS (pink), GAT (green), hinge (bar), and GAE (purple). Regulatory domain (Continued)

#### **FIGURE 4 | Continued**

and sites of protein-protein interaction are shown. Autoinhibition sequence is shown (CKII phosphorylation site is in bold and the DXLL motif is underlined in green). VHS, Vps27p, Hrs, Stam; GAT, GGA, and TOM; GAE, γ-adaptin ear. **(E)** Autoinhibition mechanism: when

vesicles (70), and immature SG membranes from PC12 cells (71). PI4KIIα is recruited to membranes upon palmitoylation of a central CCPCC motif (62). Deletion of the CCPCC motif or prevention of its palmitoylation resulted in loss of PI4KIIα perinuclear localization and loss of PI4KIIα activity (62). PI4KIIα palmitoylation is required for kinase activity and for localization to lipid rafts at the TGN but its targeting motif to Golgi membranes remains to be determined (63, 72). Lipid rafts are enriched in cholesterol and reduction of endogenous cholesterol levels resulted in loss of PI4P synthesis by PI4KIIα (63, 73). The membranes of neuroendocrine SGs contain high levels of cholesterol (74, 75). A decrease in the cholesterol level in AtT-20 cells inhibited the formation of constitutive secretory vesicles and SGs (76). Since PI4KIIα requires cholesterol for activity and membrane localization (63) and cholesterol is required for SG formation, this suggests that PI4KIIα is required for SG biogenesis.

#### **MATURATION OF SGs**

Once immature SGs are formed, they are not responsive to secretagogues and must undergo a maturation process to gain this ability. SG maturation involves a decrease in lumenal pH, SG fusion, and membrane remodeling.

#### **pH DECREASE**

The lumen of the TGN and the lumen of immature SGs are comparable in pH (pH 6.3), while mature SGs are more acidic (luminal pH 5.5). Secretogranin II processing in PC12 cells exogenously expressing prohormone convertase 2 occurs in immature SGs but not in the TGN (77). Although similar in lumenal pH, the TGN and immature SGs are two distinctly different compartments; the proteolytic processing of some prohormones and granins starts in immature SGs, not in the TGN. The decrease in pH during SG maturation is necessary for full processing of prohormones and granins because some hormone processing enzymes, such as PAM-1, exhibit maximal catalytic activity at pH 4.5–5 (78).

#### **SG FUSION**

Maturation of SGs can be characterized by a decrease or an increase in size. Analysis of immature and mature SGs in PC12 cells and mammotrophs revealed an increase in SG size during maturation (16, 79). An elegant biochemical assay revealed that the increase of SG size in PC12 cells was due to homotypic fusion of immature SGs (10). SG fusion involves syntaxin 6, a target (t)-SNARE [Soluble NSF (*N*-ethylmaleimide-Sensitive Factor) Associated Protein (SNAP) Receptor] protein, and synaptotagmin IV (80, 81). The SNARE complex is involved in membrane fusion and is composed of two t-SNAREs, located on the target membrane, and one vesicle (v)-SNARE on the vesicle membrane. Members of the synaptotagmin family are associated with the vesicular membrane and regulate membrane fusion with the SNARE complex. Syntaxin 6 phosphorylated by CKII, the autoinhibition domain binds VHS preventing cargo interaction. When the autoinhibition is dephosphorylated, no internal binding occurs allowing cargo interaction with GGA. The single letter code for amino acids was used; CHC, clathrin heavy chain; CKII, casein kinase II.

localizes to the TGN, endosomes, and immature SGs (82,83), while synaptotagmin IV is largely present at the Golgi and in immature SGs (84, 85). Blocking the fusion event decreases prohormone convertase 2 processing and activity, resulting in impaired SG maturation (80). Although this concept is attractive, there is no evidence of SG fusion in other types of neuroendocrine cells. SG fusion of a different type is clearly observed during compound exocytosis; mature SGs fuse together to promote rapid release of SG content (86).

#### **MEMBRANE REMODELING**

The reduction of SG size during maturation may be explained by membrane remodeling. The presence of a clathrin coat on patches of immature SG membrane revealed egress of material in clathrin-coated vesicles, mediated by the AP, AP-1A (8, 9, 54). As a result, proteins like vesicle associated membrane protein 4 (VAMP4) (87, 88), furin (89), and both mannose 6-phosphate receptors (MPRs) (83, 90–92), each of which is known to have a canonical AP-1A binding site in its cytosolic domain, are found in immature SGs but not in mature SGs. Although the fate of the material retrieved from immature SGs is not entirely clear, there are lines of evidence suggesting that it can be recycled to the TGN or secreted in a process called constitutive-like secretion (**Figure 1**). The study of constitutive-like secretion is difficult to monitor since it refers only to the non-stimulated secretion of regulated SG proteins as they traverse immature SGs; without a means of assessing the time since synthesis, it is impossible to distinguish basal from constitutive-like secretion. Nevertheless, constitutivelike secretion has held many scientists'attention due to its potential link to cancer and metabolic disease. Much of our knowledge of constitutive-like secretion comes from studying retrieval of material from immature SGs. The cytosolic machinery involved in this process contains the APs, AP-1A, and GGAs, and their partner phosphofurin acidic cluster sorting protein 1 (PACS-1). The discovery of an interaction between the cytosolic tail of membrane SG proteins retrieved during SG maturation with these cytosolic proteins has expanded our understanding of membrane remodeling during SG maturation. Each cytosolic component essential for maturation is presented below, followed by the description of SG membrane proteins trafficking involving AP-1A, GGAs, or PACS-1.

#### **CYTOSOLIC MACHINERY**

#### **ADAPTOR PROTEIN 1A The adaptor protein family**

The adaptor protein family includes five cytosolic heterotetrameric complexes: AP-1 (γ/β1/µ1/σ1), AP-2 (α/β2/µ2/σ2), AP-3 (δ/β3/µ3/σ3),AP-4 (ε/β4/µ4/σ4), and AP-5 (ζ/β5/µ5/σ5) (93–96). In mammals, several isoforms have been reported for subunits of AP-1, AP-2, and AP-3: AP-1 has two γ subunits (γ1 and γ2), two µ subunits (µ1A and µ1B), and three σ subunits (σ1A, σ1B, and σ1C); AP-2 has two α subunits (α1 and α2); AP-3 has two β subunits (β3A and β3B), two µ subunits (µ3A and µ3B); and two σ subunits (σ3A and σ3B) (93, 97–101). Every subunit is ubiquitously expressed except for µ1B, which is found exclusively in polarized epithelial cells, and β3B and µ3B, which are only found in neurons and neuroendocrine cells (99–101). Each adaptor protein carries transmembrane proteins on a defined intracellular route: AP-1 brings cargo between the TGN and endosomes and removes material from immature SGs; AP-2 is an important player in clathrin-mediated endocytosis; AP-3 carries cargo to lysosomes and lysosome-related organelles; AP-4 transports cargo from the TGN to the plasma membrane or endosomes; AP-5 is found on late endosomal membranes.

The cytosolic domains of cargo proteins present motifs which are recognized by adaptor protein complexes. For example, the Tyr sorting motif (YXXΦ, where X is any residue and φ is a hydrophobic residue) is recognized by the µ subunit of all adaptor protein complexes, with the possible exception of AP-5, which has not yet been studied (**Figure 4A**) (102–105). The di-leucine sorting motif [(D/E)XXXL(L/I)] interacts at the interface of two subunits: γ/σ1 in AP-1, α/σ2 in AP-2, and ε/σ3 in AP-3 (**Figure 4A**) (106, 107); AP-4 and AP-5 have not been shown to interact with the di-leucine motif. The γ/α/δ/ε/ζ and β1-5 subunits contain a large N-terminal trunk domain that associates with the other three subunits of the complex, followed by a hinge region and a C-terminal ear domain, which can interact with cytosolic proteins (**Figure 4A**) (108–111). The hinge region of the β subunits of AP-1, AP-2, and AP-3 is capable of binding the terminal domain of the clathrin heavy chain *in vitro*, while the β subunits of AP-4 and AP-5 lack such a motif (**Figure 4A**) (112–114). Although the β3 subunit of AP-3 can bind clathrin, purified clathrin-coated vesicles lack AP-3, suggesting that AP-3 works independently of clathrin (115). In addition, there is also evidence that the γ and α hinge domains of AP-1 and AP-2 interact with clathrin (**Figure 4A**) (116, 117).

#### **AP-1A recruitment to membranes depends on Arf1 and PI4P**

Adaptor protein 1A (γ/β1/µ1A/σ1) accumulates at the TGN, which is enriched in PI4P. *In vitro* binding assays revealed that purified AP-1A interacts preferentially with PI4P (**Figure 4A**) (59). Indeed, structural studies on the AP-1A core confirmed the presence of a PI4P binding site within the γ subunit (118). Reduction of PI4KIIα levels in a monkey kidney cell line decreased the TGN content of PI4P and disrupted the TGN localization of AP-1A (59). However, a recent study on development of the *Drosophila* salivary gland revealed no change in the localization of AP-1 in PI4KIIα null flies, suggesting that PI4P synthesis in flies can be carried out by another PI4K (69). Although the presence of PI4KIIα on the membranes of immature SGs suggests that PI4P can be formed there, there is no evidence that AP-1A recruitment to immature SGs requires PI4P in neuroendocrine cells.

Brefeldin A treatment of intact cells produces a diffuse cytoplasmic localization of AP-1A instead of the normal perinuclear, membrane-associated localization. However, pretreatment with GTPγS, a non-hydrolyzable analog of GTP, prevented the membrane dissociation of AP-1A upon addition of brefeldin A in permeabilized normal rat kidney cells (119). Indeed, AP-1A is

recruited to membranes upon the binding of GTP-bound Arf1 to the trunk region of its β1 and γ subunits (**Figure 4A**) (55, 120– 122). Studies on PC12 cells also show that AP-1A recruitment to immature SG membranes requires Arf1 (54, 55). Interestingly, AP-1A cannot be recruited to mature SG membranes in PC12 cells, in agreement with morphological studies indicating that AP-1A/clathrin coats are not found on mature SGs (55). A recent structural study of AP-1A in complex with GTP-bound Arf1 revealed that Arf1 can change the conformation of µ1A without the presence of PI4P or cargo; the addition of cargo promotes this conformational change (120). This shift in µ1A conformation is thought to promote the association of µ1A with cargo.

In addition to Arf1 and PI4P, AP-1A phosphorylation affects its localization; cytosolic AP-1A is phosphorylated on its β1 subunit, while membrane-associated AP-1A is phosphorylated on its µ1A subunit (123). Phosphorylation of the β1 subunit prevents clathrin binding, while phosphorylation of µ1A may promote conformational changes in the subunit that enable it to interact with cargo (118, 124). The phosphorylation status of AP-1A is thus essential for protein trafficking.

#### **PHOSPHOFURIN ACIDIC CLUSTER SORTING PROTEIN 1 – AN ENHANCER OF AP-1A FUNCTION**

Adaptor protein 1A is not the only AP involved in the retrieval of material from immature SGs. Furin, a type I transmembrane endoprotease, cleaves peptides after the last Arg of an R-X-[K/R]-R motif; its discovery was a key advance contributing to the identification of the prohormone convertases that are essential to SG function. Furin is concentrated in the TGN and travels between the TGN, endosomes, and plasma membrane (125). In cells containing a regulated secretory pathway, furin was identified in immature SGs but not in mature SGs (89). The removal of furin from immature SGs involves AP-1A and clathrin-coated vesicles (89). Its cytoplasmic domain contains a Tyr-based sorting motif and an acidic cluster, which includes two Ser residues, both of which are substrates for CKII (**Table 1**) (126, 127). The Tyr motif found in the cytosolic domain of furin mediates TGN exit and plasma membrane internalization; the acidic cluster is important for retrieval of furin from endosomes and immature SGs (89, 127, 128). Mutation of the two CKII phosphorylation sites in furin to Ala (to prevent phosphorylation) resulted in the accumulation in mature SGs in a corticotrope tumor cell line (89,129). Since AP-1A does not bind to acidic cluster, another interactor was sought. A search for proteins with high affinity for the phosphorylated acidic cluster through yeast two-hybrid screens led to the discovery of PACS-1 (127).

Phosphofurin acidic cluster sorting protein 1 is a ubiquitously expressed cytosolic protein comprised of four domains: at its N-terminus is an atrophin-1-related region (ARR), with limited homology to atrophin-1; this region is followed by a furin binding region (FBR), the middle region (MR), and the C-terminal region (CTR) (**Figure 4B**) (127). The functions of the ARR and the CTR are unknown. The FBR of PACS-1 interacts with acidic clusters in cargo proteins, with APs (AP-1A, AP-3, and GGA), and with CKII (**Figure 4B**) (127, 129, 130). The MR contains an autoinhibition domain, which is composed of an acidic cluster with a CKII phosphorylation site (**Figure 4C**); when this region is not phosphorylated, the MR binds the FBR, resulting in inhibition of

PACS-1 binding to its cargo (131). When the MR is phosphorylated by CKII, it reduces MR affinity for FBR, allowing the FBR to bind to acidic clusters in cargo proteins (130, 131). Furin has been identified in a heterotrimeric complex with AP-1A and PACS-1 (129). Expression of a PACS-1 mutant capable of binding the furin acidic cluster, but not AP-1A, does not alter the distribution of AP-1A, instead resulting in the accumulation of furin in mature SGs (129). This demonstrates that removal of furin from immature SGs is a cooperative process involving both AP-1A and PACS-1 (**Figure 5A**).

#### **GGAs – ANOTHER FAMILY OF ADAPTOR PROTEINS ESSENTIAL FOR SG MATURATION**

Golgi-localizing, γ-adaptin ear homology domain, ARF-binding proteins (GGAs) were discovered by several groups interested in different aspects of the proteins. Boman et al. screened for Arf3 binding partners using a yeast two-hybrid system, while Hirst et al. used a bioinformatic approach to look for proteins with homology to the ear region of the γ subunit of AP-1 (109, 132). These studies, combined with the fact that the proteins discovered concentrated in the TGN area, explain the origin of the name GGA (109, 133).

The GGA family contains three members in humans (GGA1, GGA2, and GGA3), two in *Saccharomyces cerevisiae* (Gga1p and Gga2p), and one in *Drosophila melanogaster* and *Caenorhabditis elegans* (109, 132, 134). Each member is ubiquitously expressed and acts as a monomeric clathrin adaptor (132, 135). From their N-terminus to their C-terminus, they are composed of a VHS [vacuolar protein sorting 27 (Vps27), hepatocyte-growth factorreceptor substrate (Hrs), signal-transducing adaptor molecule (Stam)] domain, a GAT [GGA and TOM (Target of Myb)] domain, a hinge, and a GAE (γ-adaptin ear) domain (134) (**Figure 4D**).

The VHS domain was identified based on its role in the trafficking of yeast Vps27p and for its homology to two other monomeric endosome-localized clathrin APs, Hrs, and Stam (109, 132, 136). A yeast two-hybrid screen revealed that the cytosolic domains of proteins containing either a tyrosine sorting motif (YXXΦ) or a dileucine sorting motif [(D/E)XXXL(L/I)] were not recognized by any GGA proteins, but were specific to adaptor protein complexes (126, 137). The VHS domain of GGAs recognize a motif called acidic cluster-di-leucine motif (DXXLL), which is different from the adaptor protein di-leucine sorting motif [(D/E]XXXL(L/I)] (**Figure 4D**) (134, 137).

Target of Myb1 (TOM1), a protein involved in the trafficking of ubiquitinated proteins, contains a GAT domain similar to the GAT domain in GGAs. Indeed, both yeast Gga1p and mammalian GGAs bind monoubiquitin via their GAT domain (134, 138–141). This interaction is required for the sorting of some ubiquitinated cargo proteins in *S. cerevisiae*, such as Gap1 and Arn1 (134, 138, 142, 143). The GAT domain interacts with GTP-bound Arf1 and Arf3 and the C-terminus of GAT binds PI4P; as for AP-1A, both binding sites are required for proper membrane localization of GGA in cells (**Figure 4D**) (132, 133, 141). Indeed, brefeldin A treatment results in the dissociation of GGAs from TGN membranes and in their accumulation in the cytosol, as seen for AP-1A (109, 132, 133, 135). Interestingly, association of the GAT domain of GGA3 to Arf1 blocks GAP proteins from acting on GTP-bound Arf1, thus stabilizing it on TGN membranes (53).

The hinge region between the GAT and the GAE domains is predicted to lack secondary structure (134) and contains a clathrin binding motif which is necessary for the recruitment of clathrin *in vitro* and in cells (**Figure 4D**) (53, 109). While the GAE domain of GGA1 also interacts with clathrin heavy chain *in vitro*, the GAE domains of GGA2 and GGA3 are unable to do so (53). GGAs are important for the recruitment of clathrin to membranes; overexpression of GGAs increases the TGN localization of clathrin while overexpression of truncated GGA1 lacking the hinge and GAE domains reduces the TGN level of clathrin (53). Moreover, *in vitro* studies revealed that the GGA1-hinge region and the γ subunit of AP-1 interact (**Figures 4A,D**) (108, 144). In addition, the hinge region of both GGA1 and GGA3 contains a DXXL[L/M] motif, which can bind the VHS domain when the surrounding Ser residue has been phosphorylated by CKII (**Figures 4D,E**) (108, 145); this internal binding results in autoinhibition of the AP by loss of cargo binding (108, 145, 146). The autoinhibition is removed when the Ser residue is dephosphorylated by a protein phosphatase 2A-like enzyme (146). *In vitro* phosphorylation by CKII was not observed for GGA2, suggesting that GGA2 is controlled differently (108, 126). Although the autoinhibition concept is attractive, a recent structural report refutes the existence of such a motif within GGA1 and possibly within GGA3 as well (147). These conflicting results are currently unclear, but emphasize the fact that our knowledge of GGAs is incomplete.

The GAE domain has homology to the ear domain of γ-adaptin of AP-1 (109, 132, 133, 135, 148). It interacts with several cytosolic proteins including rabaptin-5, a Rab4/Rab5 effector involved in membrane fusion at the endosomes (109, 149, 150) and γ-synergin (148).

In neuroendocrine cells, reducing the levels of GGA3 or overexpressing a dominant negative form of GGA1 resulted in impaired egress of syntaxin 6, cation-independent MPR (CI-MPR), and VAMP4 from immature SGs and in decreased proteolytic processing of prohormone convertase 2 and secretogranin II (151). Thus GGAs, AP-1A, and PACS-1 are all key components in the maturation of SGs in neuroendocrine cells.

#### **DO GGAs AND AP-1A WORK TOGETHER?**

Despite differences in their cargo recognition motifs, GGAs and AP-1A are both recruited to TGN membranes in an Arf1 and PI4P dependent manner, interact with clathrin and share common cargoes, such as CI-MPR (**Table 1**). Using transmission electron microscopy, vesicles budding from the TGN were seen to contain one AP or both GGA2 and AP-1A (108). A novel technique developed in mammalian cells to rapidly inactivate AP-1A or GGA2 revealed that AP-1A is the main AP required for trafficking between the TGN and endosomes; when AP-1A is inactivated, GGA2 is no longer present in clathrin-coated vesicles (152). In addition, GGA2 is thought to be required for proper localization of rabaptin-5, a group of lysosomal hydrolases, both types of MPRs and sortilin in clathrin-coated vesicles. Other substrates, such as Arf1, carboxypeptidase D (CPD), furin, the copper transporters (ATP7A and ATP7B), or the SNARE proteins require AP-1A for localization in clathrin-coated vesicles (152). In agreement with these data, mutation of the GGA binding site in the cytosolic tail of the CI-MPR decreases its amount in clathrin coats at the TGN, suggesting that GGAs help recruit specific cargo with APs and clathrin (108). However, a recent study in yeast using fluorescently tagged proteins suggested that the Gga proteins come into play first, and then recruit other proteins along with AP-1A after being released from the membrane (153). The yeast study also suggested that AP-1 and Gga proteins work mainly in separate vesicles. This divergence between yeast and mammals is unclear and may be linked to changes during evolution. Nevertheless, these studies do not provide information on the mechanism which governs the egress of material at the immature SG. In addition, yeast do not have a regulated secretory pathway, limiting their use in understanding SG maturation.

#### **REAL LIFE EXAMPLES**

Most of the integral membrane proteins found in post-TGN membranes contain multiple signaling motifs in their cytosolic tails. The study of protein retrieval from immature SGs was facilitated by our growing knowledge of APs and the motifs they recognize. While our understanding of the rules governing retention vs. retrieval of membrane proteins from immature SGs is not yet complete, detailed studies of the interactions of several immature SG cargo protein cytosolic tails with multiple APs reveals many of the key features (**Figure 5A**; **Table 1**).

#### **SNARE PROTEINS**

Similar observations to furin were made with the v-SNARE component, VAMP4 which contains an AP-1A binding motif (dileucine sorting motif) and an acidic cluster surrounding a CKII phosphorylation site (**Table 1**) (87). VAMP4 accumulates at the TGN, on endosomes and on immature SG of PC12 and AtT-20 cell lines (87, 154). As for furin, overexpression of PACS-1 mutant unable to interact with AP-1A but capable of interacting with VAMP4 resulted in accumulation of VAMP4 in mature SG (87). Interestingly, VAMP4 accumulates in mature SGs only if both its AP-1A and its PACS-1 binding motifs are mutated; if either binding motif is remaining, VAMP4 is efficiently removed from immature SGs (87). This is different from furin which accumulated in mature SGs when the CKII sites were mutated. The AP-1A/PACS-1/clathrin sorting system seems to work differently with VAMP4 and furin.

In addition, VAMP4 interacts with syntaxin 6, which binds synaptotagmin IV (80, 154). All three proteins are retrieved from immature SGs (83, 84, 154). Although *in vitro* studies failed to show binding between syntaxin 6 and AP-1A (87), observation of immature SGs using electron microscopy showed that syntaxin 6 is retrieved in AP-1A and clathrin containing vesicles (83). It remains unclear whether synaptotagmin IV interacts with AP-1A, but since the removal of VAMP4 and syntaxin 6 has been linked with AP-1A, it seems likely that synaptotagmin IV removal is also dependent on the AP-1A/PACS-1/clathrin sorting system due to its indirect binding with VAMP4. Interestingly synaptotagmin IV is known to inhibit calcium triggered exocytosis (84), suggesting that its removal is necessary to make mature SG responsive to secretagogues. Since synaptotagmin IV plays a critical role in SG fusion which is an essential step in SG maturation, these studies would suggest that SG fusion occurs before membrane remodeling.

#### **THE MANNOSE 6-PHOSPHATE RECEPTORS**

Mannose 6-phosphate receptors transport soluble lysosomal hydrolases from the TGN to endosomes (137, 155). MPRs are found in immature SGs and are removed during maturation in clathrin-coated vesicles that contain AP-1A (83). Two MPRs have been described: MPR300, the CI-MPR, and MPR46, the cationdependent MPR (CD-MPR). Much has been learned by studying the multiple trafficking signals in both MPRs. Both MPRs contain Tyr sorting motifs (YXXΦ) and an acidic cluster-di-leucine motif (DXXLL) next to a CKII phosphorylation site (**Table 1**). Early studies revealed that the Tyr motif of CI-MPR is recognized by AP-2 but not by AP-1A (156). CD-MPR contains two tyrosine motifs: the first tyrosine motif is located next to the transmembrane domain, and may not be a signaling motif because of the hindrance caused by the membrane (**Table 1**, Y <sup>211</sup>QRL). The second tyrosine motif is found away from the membrane (**Table 1**, Y256RGV) and is implicated in the protein internalization (157).

The DXXLL motif is recognized by the VHS domain of GGAs (137, 150). Interestingly, the DXXLL motif found on the cytosolic tail of the CI-MPR binds the VHS domain of all GGAs while the DXXLL motif of the CD-MPR interacts with the VHS domain of GGA1, interacts weakly with GGA3 and does not interact with GGA2 (137, 150). Mutational analyses revealed that the residues surrounding the motif are important for establishing GGA specificity (137). Down-regulation of GGA3 or inhibition of GGA1 results in the accumulation of CI-MPR in mature SGs, suggesting that retrieval of CI-MPR from immature SGs requires GGAs (151).

Independent *in vitro* studies focusing on either CD-MPR or CI-MPR showed that removal of the di-leucine motif from either receptor had little effect on its ability to recruit AP-1A. In CI-MPR, elimination of the CKII phosphorylation site abolished recruitment of AP-1A (90, 127). Indeed, binding of AP-1A to CI-MPR requires prior phosphorylation of its cytosolic domain by CKII (91, 158). These features suggested that PACS-1 might be involved in the trafficking of CI-MPR. In non-neuroendocrine cells, reducing the cellular level of PACS-1 or overexpressing a PACS-1 mutant unable to interact with acidic clusters resulted in accumulation of CI-MPR in endocytic compartments, suggesting that PACS-1 recruitment is necessary for MPR retrieval from the endosomes

to the TGN (127, 129). AP-1A binding to CD-MPR requires the acidic cluster; experiments aimed at understanding the role of phosphorylation at the Ser CKII site in CD-MPR in its binding of AP-1A have yielded conflicting results (**Table 1**, Ser268) (91, 159). It is not yet known whether MPRs are retrieved from immature SGs in a PACS-1 dependent manner.

GGA3 and PACS-1 interact with the CI-MPR at sites that are overlapping, but not identical (**Table 1**) (130). The FBR region of PACS-1 binds CKII,enhancing its kinase activity (130). This results in phosphorylation of the MR of PACS-1, releasing the FBR for cargo interaction (**Figure 4C**). In addition, CKII phosphorylates GGA3, decreasing its affinity for cargo and thus making space for PACS-1 to bind (**Figures 4E** and **5B**) (130). Although this study did not focus on immature SGs, it provides a mechanism of why GGA and AP-1A/PACS-1 are necessary for SG maturation.

#### **CARBOXYPEPTIDASE D DOES NOT CONTAIN AN AP-1A BINDING MOTIF**

The exopeptidase CPD, a type I transmembrane protein, accumulates in the TGN and cycles between the TGN and plasma membrane via endosomes (160). In AtT-20 cells, CPD is found in immature, but not in mature SGs suggesting that it is removed during SG maturation (161). The isolated luminal domain of CPD primarily enters the constitutive secretory pathway, suggesting that signals in its cytosolic tail play an essential role in CPD entry into immature SGs and in the TGN retention of CPD (161). The CPD cytosolic tail contains an acidic cluster with two CKII phosphorylation sites (TDT), but lacks classical AP and GGA binding motifs (**Table 1**). *In vitro* binding assays revealed that the CPD cytosolic tail binds to AP-1A and AP-2 only when its two CKII phosphorylation sites are phosphorylated or have been mutated to phosphomimetic residues (EDE) (**Table 2**) (162). Deletion of the C-terminus of the CPD tail, which contains the CKII phosphorylation sites does not abolish AP-1A or AP-2 binding (**Table 2**) (162), suggesting a role for additional CPD/AP-1A/2 interaction motifs. A tyrosine-like motif (**Table 2**, F1332HRL) located upstream of the acidic cluster is thought to be important for TGN export and endocytosis and is necessary for AP-1A and AP-2 binding (162). In addition, the C-terminus of CPD may bind another cytosolic protein which would prevent APs to interact. This unidentified protein probably loses its affinity for the C-terminus of CPD when the CKII sites are phosphorylated or removed, allowing APs to interact. Phosphorylation of CPD by CKII promotes AP-1A binding, suggesting that PACS-1 plays a role in its trafficking. The trafficking of CPD at immature SGs probably results from the coordinated work of multiple cytosolic proteins including AP-1A and PACS-1.

#### **VMAT2 SUGGESTS THE EXISTENCE OF ADDITIONAL ADAPTOR PROTEINS**

The monoamine transporter, VMAT2, accumulates in mature SGs in neuroendocrine cells. Its cytosolic domain contains an acidic cluster with two CKII sites at the C-terminus; upon CKII phosphorylation, this site interacts with PACS-1 (163). As for furin, expression of VMAT2 in which the CKII sites have been replaced with phosphomimetic mutations promotes its retrieval from immature SGs. Mutation of these same two residues into Ala to prevent phosphorylation promotes VMAT2 accumulation in mature SGs (163); interestingly, expression of VMAT2 lacking the acidic cluster/CKII

#### **Table 2 | Summary of mutational analysis of CPD tail.**


Results adapted from (162). TMD, transmembrane domain; COOH, C-terminus of the protein. CKII phosphorylation site is in bold, the acidic cluster in red and the AP-1A-like site in blue. Gray marks TMD and COOH

domain resulted in its egress from immature SGs (163). Unlike furin, which accumulates in the TGN at steady state, VMAT2 concentrates in mature SG raising the question of VMAT2 retention mechanism. These findings suggest that additional cytosolic proteins are involved in the sorting of SG proteins at the immature SG and may explain the retention of other membrane proteins in SGs during maturation. Indeed the amidating enzyme, PAM, remains in SGs during maturation, and presents an acidic cluster with two CKII phosphorylation sites (39). Although direct binding between PAM cytosolic domain and PACS-1 has not been shown, it is possible that PAM interacts with the same cytosolic proteins which bind VMAT2 for its retention in SG.

#### **PHOGRIN**

Phogrin is a transmembrane protein that spans the membrane of immature and mature SGs. The cytosolic domain of phogrin contains a protein-tyrosine phosphatase (PTP) domain preceded by a 105 residue-long domain and followed by 10 residues at its C-terminus (**Figure 2B**; **Table 1**). The PTP domain of phogrin is an active phosphatidylinositol (PI) phosphatase; its enzymatic activity is thought to play an essential role in its ability to regulate peptide hormone secretion (164). Both domains surrounding the PTP domain are involved in the correct sorting of phogrin in the regulated secretory pathway (36, 37). The 105 residue stretch close to the transmembrane domain contains a tyrosine-extended motif (**Table 1**, Y655QEL) and the 10 residues at the C-terminus present a di-leucine-like motif (**Table 1**, E <sup>993</sup>EVNAIL). Both motifs are required for entry of phogrin into SGs and internalization at the plasma membrane (36, 37). Interestingly, both motifs are involved in the binding of phogrin to AP-1A and AP-2. Mutation of one of these motifs decreases the binding strength of APs for phogrin but does not abolish the interaction (36, 37). Although phogrin binds AP-1A via unconventional motifs, it still remains to be determined what prevents this transmembrane protein from being removed during SG maturation.

#### **SORTILIN AND PRO-BDNF**

Sortilin, a type I transmembrane protein, binds the pro-region of brain-derived neurotrophic factor (pro-BDNF) and transports it to SGs for processing (165). As for CPD, expression of the luminal domain of sortilin results in its entry into the constitutive secretory pathway, suggesting that entry into SGs is mediated by the cytosolic tail (166). A recent study revealed that interaction of sortilin with Huntingtin associated protein 1 (HAP-1) and pro-BDNF must occur for pro-BDNF entry into SGs and subsequent proteolytic processing (167). In the absence of HAP-1, sortilin, and pro-BDNF are targeted to lysosomes. Interestingly, the cytosolic tail of sortilin contains both a tyrosine motif and an acidic clusterdi-leucine motif, classical AP, and GGA binding sites, respectively (**Table 1**). Both motifs are implicated in sortilin internalization (168). As expected, the VHS domain of GGA2 interacts with the acidic cluster-di-leucine motif of sortilin (168). However, the tyrosine motif is recognized by the retromer complex, not by AP-1A (169). The AP-1A binding motif on sortilin remains to be elucidated. Both GGA2 and AP-1A were shown to be important for the TGN-to-endosome transport of sortilin. It is possible that HAP-1 interacts with sortilin to prevent GGA and AP-1A binding and mediates the entry of sortilin and pro-BDNF into SGs. Although sortilin contains AP binding motifs, it is not clear if sortilin is retrieved from immature SGs. One could assume that once in immature SGs, sortilin can release pro-BDNF and be retrieved during SG maturation.

#### **CONCLUSION AND FUTURE DIRECTIONS**

Little is known about the cytosolic machinery involved in sorting membrane proteins at the TGN for their entry into immature SGs. A recent study in *Drosophila* suggested that AP-3, the AP transporting proteins to lysosomes, was involved in sorting proteins into the regulated secretory pathway at the TGN level (170). Unfortunately, it is not known whether this is also true in the mammalian neuroendocrine system. In addition, studies on the trafficking of individual proteins raises many questions regarding the cytosolic machinery necessary for their retrieval or retention during SG maturation. It remains to be explained how a membrane protein which contains an AP (AP-1A, PACS-1, or GGAs) binding motif can remain in SGs during maturation. One possibility is that other cytosolic proteins bind the tails of membrane proteins, preventing their interaction with APs. Recent proteomic studies on purified SGs are an excellent source of information and may help identify additional proteins involved in the egress of membrane proteins (171, 172). Another possibility is a binding site within the cytosolic tail of membrane proteins which could prevent AP interaction; a post-translational modification, such as phosphorylation, can abolish this internal interaction while promoting the interaction with APs. Lastly, the membrane of the immature SG may play an important role in concentrating membrane proteins to be removed during maturation. Katsumata et al. addressed this question using the rat parotid gland as a source of SGs. The immature SG membrane revealed spatial segregation of membrane proteins: VAMP2 and Rab3D were in GM1a rich microdomains, while syntaxin 6 and

VAMP4 were not. Since VAMP2 and Rab3D remain in mature SGs while syntaxin 6 and VAMP4 are removed from immature SGs, membrane rearrangement could explain how the right proteins are sorted for retrieval of material at the immature SG (173). The trafficking of individual proteins found in immature SGs can differ based on the cell type or animal model. It is evident that professional secretory cells share a common mechanism, and

#### **REFERENCES**


that the differences they present are probably required for their physiological function.

#### **ACKNOWLEDGMENTS**

This work was supported by National Institutes of Health Grants DK32948 and DK32949. We thank Dr. Vishwanatha Kurutihalli and Dr. Yan Yan for thoughtful comments on the manuscript.

in mouse pituitary tumor cells. *J Cell Biol* (1982) **95**:559–66. doi:10.1083/jcb.95.2.559


dense core secretory granule sorting domain from the PC1/3 protease. *Proc Natl Acad Sci U S A* (2009) **106**:7408–13. doi:10.1073/ pnas.0809576106


switching. *Cell* (1998) **95**: 237–48. doi:10.1016/S0092- 8674(00)81754-7


post-Golgi compartment. *J Biol Chem* (2003) **278**:5343–52. doi:10. 1074/jbc.M208291200


complex with clathrin. *Science* (1998) **280**:431–4. doi:10.1126/ science.280.5362.431


of the furin proprotein convertase: localization to the TGN and recycling from the cell surface. *EMBO J* (1994) **13**:18–33.


**275**:7176–83. doi:10.1074/jbc.275. 10.7176


ligand-binding site of GGA1/3 VHS domains by an internal acidic cluster-dileucine motif. *Proc Natl Acad Sci U S A* (2002) **99**:8072–7. doi:10.1073/pnas.082235699


separates from GM1a-rich membrane microdomain during granule maturation. *Biochem Biophys Res Commun* (2007) **357**: 1071–7. doi:10.1016/j.bbrc.2007. 04.053

174. Youker RT, Shinde U, Day R, Thomas G. At the crossroads of homoeostasis and disease: roles of the PACS proteins in membrane traffic and apoptosis. *Biochem J* (2009) **421**:1–15. doi:10.1042/ BJ20081016

**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: 10 June 2013; accepted: 31 July 2013; published online: 14 August 2013. Citation: Bonnemaison ML, Eipper BA and Mains RE (2013) Role of adaptor proteins in secretory granule biogenesis and maturation. Front. Endocrinol. 4:101. doi: 10.3389/fendo.2013.00101 This article was submitted to Frontiers in Neuroendocrine Science, a specialty of Frontiers in Endocrinology.*

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

### The cortical acto-myosin network: from diffusion barrier to functional gateway in the transport of neurosecretory vesicles to the plasma membrane

#### **Andreas Papadopulos, Vanesa M. Tomatis, Ravikiran Kasula and Frederic A. Meunier \***

Queensland Brain Institute, The University of Queensland, St Lucia Campus, Brisbane, QLD, Australia

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Nicolas Vitale, Centre National de la Recherche Scientifique, France Jose-Maria Trifaro, University of Ottawa, Canada

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

Frederic A. Meunier, Queensland Brain Institute, The University of Queensland, St Lucia Campus, QBI Building #79, St Lucia, QLD 4072, Australia e-mail: f.meunier@uq.edu.au

Dysregulation of regulated exocytosis is linked to an array of pathological conditions, including neurodegenerative disorders, asthma, and diabetes. Understanding the molecular mechanisms underpinning neuroexocytosis including the processes that allow neurosecretory vesicles to access and fuse with the plasma membrane and to recycle post-fusion, is therefore critical to the design of future therapeutic drugs that will efficiently tackle these diseases. Despite considerable efforts to determine the principles of vesicular fusion, the mechanisms controlling the approach of vesicles to the plasma membrane in order to undergo tethering, docking, priming, and fusion remain poorly understood. All these steps involve the cortical actin network, a dense mesh of actin filaments localized beneath the plasma membrane. Recent work overturned the long-held belief that the cortical actin network only plays a passive constraining role in neuroexocytosis functioning as a physical barrier that partly breaks down upon entry of Ca2<sup>+</sup> to allow secretory vesicles to reach the plasma membrane. A multitude of new roles for the cortical actin network in regulated exocytosis have now emerged and point to highly dynamic novel functions of key myosin molecular motors. Myosins are not only believed to help bring about dynamic changes in the actin cytoskeleton, tethering and guiding vesicles to their fusion sites, but they also regulate the size and duration of the fusion pore, thereby directly contributing to the release of neurotransmitters and hormones. Here we discuss the functions of the cortical actin network, myosins, and their effectors in controlling the processes that lead to tethering, directed transport, docking, and fusion of exocytotic vesicles in regulated exocytosis.

**Keywords: cortical actin, myosin, regulated exocytosis, cdc42, phosphoinositides, secretory vesicles**

#### **INTRODUCTION**

Regulated exocytosis relies on the timely fusion of secretory vesicles or granules (SVs/SGs) with the plasma membrane. For this to occur, SVs need to be mobilized, translocated, docked, and primed at the plasma membrane. Translocation, docking/priming, and fusion of SGs rely on dynamic changes in the cortical actin network, a dense mesh of filamentous actin underneath the plasma membrane (**Figures 1A–C**) that is controlled by actin effectors and myosin motor proteins. The thick actin ring of the cortical actin network can be visualized in chromaffin cells by staining actin with a variety of methods ranging from classical immunofluorescence to phalloidin (covalently linked to fluorophores), a fungal alkaloid that preferentially binds actin filaments (**Figures 1B,C**). More recently, the development of lifeact-GFP, a 17-residue peptide from *S. cerevisiae* that selectively binds to actin without affecting neuroexocytosis (1, 2), has allowed the probing of the dynamic changes occurring during stimulation of exocytosis on the cortical actin network by time-lapse imaging (**Figures 1C,D**). Following secretagogue stimulation the cortical actin ring fragments, coinciding with a decrease in cortical F-actin labeling (**Figure 1B**). This process is Ca2+-dependent and involves actinsevering proteins such as scinderin (3–6). Although actin reorganization helps vesicles reach the plasma membrane (7), F-actin also serves as an anchoring point for SGs and provides tracks for their directed motion toward fusion sites (8). Molecular motors associated with F-actin, such as myosins (9), are involved in additional functions (2, 10).

In nerve terminals, actin is a well-known modulator of neurotransmitter release. Actin is involved in synaptic vesicle mobilization as well as axonal vesicle trafficking and synaptic plasticity (11). It is the most abundant cytoskeletal protein in synapses and is highly enriched in dendritic spines, whose formation is initiated by dendritic filopodia formation (12–15), an actin-driven process facilitated by the action of myosin X (16, 17). Neurotransmitter release at central synapses is regulated by actin and depolymerization of F-actin by latrunculin A was found to transiently enhance neurotransmitter release indicating a restraining role of F-actin in active zones (18).

#### **NEW ROLES FOR ACTIN IN EXOCYTOSIS**

The cortical actin network plays an important and well-described role during vesicle exocytosis (5, 7, 9, 10), and in recent years new functions for actin and its associated proteins have emerged (2, 9, 10, 20–24). Ca2+-dependent reorganization and remodeling of the cortical actin network help vesicles move toward the plasma membrane by partial disassembly of the cortical layer (**Figure 1B**) (3,

**(A)** Electron micrograph of a bovine chromaffin cell region attached to the thermanox support. Note the presence of a filamentous cortical region that is devoid of SG. Bar, 1µm [adapted from Ref. (19)]. **(B)** Confocal images showing the mid section of bovine chromaffin cells expressing lifeact-RFP and

6). At the same time, this remodeling provides tracks that extend further toward the center of the cell allowing the mobilization of SGs from the reserve pool (25) to their docking and fusion sites at the plasma membrane (4, 26, 27). Ca2<sup>+</sup> regulates the cortical F-actin disassembly in chromaffin cells via two pathways (28, 29). The first involves stimulation-induced influx of extracellular Ca2<sup>+</sup> through Ca2<sup>+</sup> channels and results in activation of scinderin and ensuing F-actin severing. The second pathway is triggered by Ca2<sup>+</sup> release from intracellular stores (30) and can be induced in the absence of secretagogue stimulation, by phorbol esters (3). Here actin disassembly is achieved through protein kinase C (PKC) activation followed by myristoylated alanine-rich C kinase substrate (MARCKS) phosphorylation that inhibits its F-actin-binding and cross-linking properties (28). The cortical actin network provides a layered structure that retains 2–4% of the total vesicles in close proximity to the cell surface that contribute to the burst of catecholamine release at the onset of stimulation (26, 31, 32). Indeed the majority of SGs in the vicinity of the plasma membrane are tethered to the cortical actin network (6), and newly arriving vesicles are also caught in this dense mesh of F-actin (33). Other studies

counter stained with FITC-conjugated phalloidin in the presence or absence of nicotine (50µM). **(C)** Maximum intensity projection of the footprint of a chromaffin cell. **(D)** TIRF images showing actin lengthening in a chromaffin cell expressing lifeact-GFP (pseudocolor) after the addition of PI3Kδ inhibitor IC87114. **(B–D)** Adapted from Ref. (2).

point to the existence of F-actin cages that organize the SNARE proteins SNAP25 and syntaxin-1 as well as L- and P/Q-type calcium channels, creating sites in the cortical actin network where SGs fuse preferentially (34). Consistent with these data, studies using total internal reflection fluorescence (TIRF) microscopy revealed that vesicle motion becomes restricted in the vicinity of the plasma membrane (35, 36). Interestingly, both actin depolymerization (37) and N-WASP- and Cdc42-dependent actin polymerization (**Figure 1D**) potentiate exocytosis (2, 38). While these results may appear contradictory, such opposing role for actin is not unlikely. Partial actin depolymerization helps SGs to cross the actin layer that acts as barrier, and the remaining (10) as well as newly forming actin fibers provide tracks for vesicles to reach the plasma membrane (2, 38). The balance between actin polymerization and depolymerization is likely regulated by scinderin acting as a molecular switch capable of inducing both actin polymerization and depolymerization (39). An important link connecting membranes and actin during exocytosis is the glycerophospholipid phosphatidylinositol 4,5-bisphosphate (PIP2). Although it is only a minor component of cellular membranes, microdomains, and clusters of PIP<sup>2</sup> play a crucial role in exocytosis. PIP<sup>2</sup> is known to control actin polymerization by modulating the activity and targeting of actin regulatory proteins (40). PIP<sup>2</sup> involvement in SNARE-mediated exocytosis, i.e., its Ca2+-dependent interaction with synaptotagmin-1 and syntaxin, has been described in numerous studies (41–44). Decreased levels of PIP<sup>2</sup> in the brain and impairment of its synthesis in nerve terminals lead to early postnatal lethality and synaptic defects in mice, including decreased frequency of miniature currents, enhanced synaptic depression, and a smaller ready release pool of synaptic vesicles, delayed endocytosis, and slower recycling kinetics (45). The formation of PIP<sup>2</sup> microdomains at syntaxin-1A clusters with docked SGs seems to be required for Ca2+-dependent exocytosis (46). Both PIP<sup>2</sup> and syntaxin-1A have been found in punctate nanoclusters in isolated PC12 cell plasma membrane sheets, and similar PIP<sup>2</sup> clusters in PC12 cells have been reported to link synaptotagmin-1 and syntaxin-1A, thus providing a platform for SV recruitment (46, 47). Likewise, the clustering of syntaxin-1A in model membranes has been shown to be modulated by PIP<sup>2</sup> (48). PIP<sup>2</sup> also plays a role in regulated exocytosis by controlling several proteins involved in modifying the actin cytoskeleton (40), as well as stimulating actin polymerization (49). PIP<sup>2</sup> binds scinderin in a Ca2<sup>+</sup> and pH-dependent manner (50). PIP<sup>2</sup> binding inhibits scinderin-induced actin depolymerization (51, 52), as well as the ADF/cofilin actin-severing activity (53) thereby promoting actin polymerization. A transient increase in PIP<sup>2</sup> levels is sufficient to promote the mobilization and recruitment of SVs to the plasma membrane via Cdc42-mediated actin reorganization (2). PIP<sup>2</sup> therefore links exocytosis and the actin cytoskeleton by coordinating the actin-based delivery of SVs to the plasma membrane (2). Likewise, decreasing PIP<sup>2</sup> levels in neuroendocrine cells by either ATP depletion or sequestering PIP<sup>2</sup> rapidly reduces the amount of cortical F-actin (54). In a similar study, nanomolar interaction of HIV-1 transcriptional activator with PIP<sup>2</sup> was found to prevent the actin reorganization necessary for bringing SVs to the plasma membrane and severely impaired neurosecretion in PC12 and chromaffin cells (55). Another actin-binding protein that PIP<sup>2</sup> has been found to interact with is vilin, with PIP2-vilin association inhibiting actin depolymerization and enhancing actin cross-linking (56). The interplay of Rho GTPases such as Cdc42, RhoA, and Rac with PIP<sup>2</sup> and other actin regulatory proteins controls Ca2+-regulated exocytosis in chromaffin cells (9, 22). Other small GTPases implicated in regulated secretion in neurosecretory and endocrine cells are Arf6 (57), Rab27A (58) as well as RalA and Rab3A. RalA has not only been shown to tether insulin granules to R- and L-type calcium channels (59) but also binds to the exocyst complex and regulates filopodia formation linking morphological changes and regulated exocytosis (60). RalA, which is present in GLUT4 vesicles in adipocytes, also interacts with the exocyst complex and its activation is required for insulin-stimulated GLUT4 trafficking. Impairment in the function of RalA in these cells attenuated insulin-stimulated glucose transport. RalA also interacts with Myo1C acting as a cargo receptor for this motor protein (61). In addition RalA has been found to control SG exocytosis in PC12 cells by interacting with phospholipase D1. It is activated during exocytosis and the expression of a constitutively active mutant was found to enhance neuroexocytosis whereas expression of an

inactive mutant or silencing resulted in reduced secretion (62). Of the four homologs (A/B/C/D) Rab3A is the best characterized (63). Rab3A is involved in the late steps of exocytosis. Early studies showed that Rab3A is associated with SG in bovine chromaffin cells and rat PC12 cells (64, 65). Overexpression of Rab3A mutant proteins defective in either GTP hydrolysis or in guanine nucleotide-binding inhibited exocytosis (66). Similarly the perfusion of Rab3A and various guanine nucleotides into chromaffin cells resulted in delayed catecholamine secretion suggesting a negative regulatory role in secretion (67). Rab3A plays a role in vesicle priming, where it is involved in Munc13-1 activation and interacts with Munc18-1 to regulate priming and fusion (68). Furthermore, Munc13 and Rab3A localize in the acrosomal region in human sperm, where they stimulate acrosomal exocytosis and play an important role in membrane docking (69). In human spermatozoa Rab3A and Rab27 act in a cascade that regulates dense core granule exocytosis (70). Rab3 interaction with Munc18 has also been shown to regulate SG density at the periphery of PC12 cells (71) and Rab3 guanine cycling is required for Munc18-dependent SG docking (72). However, the high level of redundancy between the four Rab3 isoforms makes it difficult to fully assess their individual contributions and the lack of an obvious exocytic phenotype in double and triple knock-out animals points to a regulatory but not essential role of Rab3A in exocytosis (73).

A number of new functions are now being attributed to the interplay between actin and various myosins. Non-muscle myosin II, and the unconventional myosins 1c/e, Va, and VI are involved in different stages during regulated exocytosis of SGs.

#### **MYOSINS**

Myosins are a 17-member superfamily of actin-based molecular motor proteins (74) that are involved in many aspects of eukaryotic cell functions, including cell movement, establishment of cell shape and polarity (75–80), and vesicular trafficking (61, 81). Myosin function is not limited to that of a molecular motor, as myosins also regulate actin polymerization, serve as molecular anchors (33), and even play a role in signal transduction (82, 83). All myosins contain a heavy chain with a conserved ∼80 kDa N-terminal catalytic domain that includes the ATPase activity and actin-binding regions (**Figure 2**) (84). This domain is followed by an α-helical neck region containing one or more IQ motifs that allow binding of light chains and calmodulin (CaM). The C-terminal myosin tail contains cargo/membranebinding domains, kinase activity, and/or mediates heavy chain dimerization depending on the myosin class (**Figure 2**) (83, 85).

#### **MYOSIN I**

Myosin I (**Figure 2**) is a single-headed membrane-associated protein that is expressed in all eukaryotic cells (84). Although there is currently no evidence for myosin I involvement in neurosecretion two isoforms of the human myosin 1C gene (86), myosin 1C (Myo1C), and myosin 1E (Myo1E) have been implicated in regulated exocytosis. All members of this unconventional myosin family interact with actin through their catalytic head domain (87). Myo1C is also capable of binding phosphoinositides (88) (**Figure 3**), thereby linking the actin cytoskeleton to the plasma membrane (89). Myo1C is recruited to GLUT4-containing vesicles

Ref. (83).

that undergo regulated exocytosis in 3T3-L1 adipocytes in an insulin-dependent manner, and is involved in their transport to the plasma membrane (**Figure 3**) (61, 81). In addition, Myo1C also tethers GLUT4-containing vesicles to the cortical actin network

with SG through cysteine string proteins and myosin Va binds to MyRIP

membrane/cargo-binding domains (orange). The small insert of myosin VI,

manner. Myosin II also regulates size and duration of fusion pore opening.

(**Figure 3**) underneath the plasma membrane in response to insulin (90), and promotes GLUT4 insertion to the plasma membrane by fusion (91), thereby regulating glucose uptake in adipose and muscle tissue (92). Myo1C is required for vascular endothelial growth factor receptor-2 (VEGFR2) delivery to the cell surface and for angiogenic signaling (93). VEGF stimulation promotes the recruitment of VEGFR2 to Myo1C and its delivery to the cell surface (93).

In *Xenopus* oocytes Myo1E, the only long-tailed myosin 1 class motor protein has been found to rapidly relocate from the cytosol to cortical SGs upon secretagogue stimulation and to bind to cysteine string proteins, components of cortical SGs that mediate vital steps in regulated exocytosis (94) by interacting with SNAP25 and the calcium sensor synaptotagmin 9 in pancreatic β-cells (95). While cortical granule exocytosis is enhanced by overexpression of Myo1E it is inhibited by injection of Myo1E antibodies (94). Myo1E has also been implicated in the recruitment of several actin-binding proteins leading to N-WASP recruitment and Arp2/3-mediated actin polymerization (**Figure 3**) (96).

#### **MYOSIN II**

Class II myosins are most abundant in muscle cells where their main function is to generate mechanical force. Non-muscle cells also contain a subset of myosin II molecules with distinct functionality. They all consist of two heavy chains (230 kDa), two regulatory light chains, and two essential light chains (**Figure 2**). In addition to actin cross-linking, bundling, and contractile properties, myosin II is known to regulate actin polymerization and is therefore linked to a great number of functions in eukaryotic cells including motility, adhesion (97), and regulated exocytosis (24, 98, 99). Non-muscle myosin II has been implicated in vesicle transport through the actin cytoskeleton (**Figure 3**). Expression of an inactive non-phosphorylatable regulatory light chain mutant myosin II fused to GFP drastically impairs granule mobility and influences actin dynamics, similar to blebbistatin treatment (100).

There is mounting evidence that myosin II is involved in controlling fusion pore dynamics and release kinetics (**Figure 3**). Expression of non-phosphorylatable regulatory light chain mutant myosin II that produces an inactive protein alters single vesicle fusion kinetics and slows fusion pore expansion (23, 24). Similarly, the release kinetics of fluorescently tagged tissue plasminogen activator and brain-derived neurotrophic (BDNF) factor are prolonged following overexpression of a wild-type form of the myosin II regulatory light chain and shortened by overexpression of a dominant-negative form (101). The use of a green fluorescent pH-sensitive protein (pHluorin) targeted inside the SVs revealed that the altered kinetics of release were caused by changes in the duration of fusion pore opening. Additional evidence indicates that myosin II affects catecholamine release by directly controlling the size of the fusion pore and the duration of its opening (20). Actin cortex disassembly elicited by high frequency stimulation promotes full fusion of SGs – an effect blocked by pharmacological inhibition of myosin II or myosin light chain (MLC) by preventing the fusion pore dilation (102). Inhibition of either actin polymerization with cytochalasin D or myosin II function with blebbistatin also slowed fusion pore expansion and increased its lifetime, suggesting that the interplay between actin and myosin II can accelerate catecholamine release (20). Similar results indicating that myosin II activity maintains an open fusion pore were obtained in exocrine pancreatic cells where myosin II (blebbistatin) and MLC (ML-9) inhibition did not alter the

number of fusion events but resulted in a decreased fusion pore lifetime (103).

It has been suggested that myosin II contractility could also help to squeeze secretory cargo out of vesicles surrounded by an actin coat once they are connected to the plasma membrane through a fusion pore (104). Fusion pore opening and closing might not be enough to release large cargo from SVs and myosin II might provide an active extrusion mechanism (104). The direct involvement of MLC and myosin II was also observed in GLUT4-containing vesicle fusion following insulin-stimulated glucose uptake in 3T3- L1 adipocytes. Only active phosphorylated myosin II was recruited to GLUT4 vesicles in an activity-dependent manner. Interestingly, insulin specifically stimulates the myosin IIA isoform via MLC kinase phosphorylation of MLC (105, 106). Myosin II inhibition also increases the distance of SGs from the plasma membrane, and promotes the retraction of the cytoskeleton, suggesting its involvement in the final approach of vesicles toward the plasma membrane (107).

Myosin II involvement in integrin-mediated cell adhesion and exocytosis has been linked to changes in cell adhesion properties (108, 109). Glucose stimulation of pancreatic β-cells promotes the remodeling of integrin focal adhesions and phosphorylation of focal adhesion kinases and myosin II (108, 109). As myosin II is one of the main substrates of Rho-kinase 1/2, which stimulates myosin–actin interactions and induces reorganization of the actin cytoskeleton, this activity could modulate SG translocation and cargo release in response to secretagogue stimulation.

#### **MYOSIN VA**

Myosin Va (**Figure 2**) has been implicated in exocytosis and vesicle movement to the cell periphery. In melanocytes, in a complex with Rab27a and melanophilin, myosin Va regulates melanosome transport to the plasma membrane (110, 111). In pancreatic βcells myosin Va also functions in the transport to and retention of insulin granules at the cortical actin network under stimulated conditions as well as their secretion (112–114). In neurosecretory cells, myosin Va is associated with SGs and plays distinctive roles during SG exocytosis (25, 115). Firstly, it assists the membrane remodeling required for SG maturation by promoting the removal of the transmembrane protein furin from maturing SGs (116). Secondly, in a complex with the SG-associated small GTPase Rab27 and its effector MyRIP, myosin Va regulates the interaction of SGs with the cortical actin network (**Figure 3**) (58). This complex has been implicated in exocytosis of SGs by modulating the transport of SGs and their retention in the cortical actin network on their way to the plasma membrane (117, 118). The interaction between myosin Va and MyRIP facilitates the dissociation of SGs from microtubules, enhancing their directed motion and the probability of SG docking to the plasma membrane (118). As a conventional processive molecular motor, myosin Va moves selective cargo along actin filaments (117). This feature strongly supports the key role of this protein in the translocation and tethering of SGs to the cell periphery. Blocking myosin Va function reduces the immobilization periods of SGs thereby decreasing the density of docked SGs near the plasma membrane and their exocytosis (117, 119). In resting conditions, myosin Va forms a stable complex with synaptic vesicle membrane proteins, synaptobrevin II, and synaptophysin (120). This complex is rapidly disassembled upon Ca2<sup>+</sup> increase in either intact nerve endings or*in vitro* assays (120). In chromaffin cells, influx of Ca2<sup>+</sup> dissociates myosin V from chromaffin vesicles supporting a role for Ca2<sup>+</sup> in the regulation of transient interactions between myosin V and its cargo (25). Furthermore, when an antibody against myosin V head was introduced in permeabilized chromaffin cells after a first stimulation of 40 s, the secretory response to a second stimulation several minutes after the first one, was greatly reduced. This points to a role for myosinV in providing SVs for the refilling of the release-ready pool following stimulation (25). The role of Ca2<sup>+</sup> as a regulator of the interaction between myosin V and its cargo has also been demonstrated in melanosomes and *Xenopus* egg extracts (121). Released CaM activates CaM kinase II (CaMK-II),a myosinVa binding partner (122). CaMK-II activation leads to myosin Va phosphorylation and the release of melanosomes from F-actin (121). Similarly, microinjection of CaM antibodies into chromaffin cells resulted in reduced catecholamine output in response to stimulation (123). Ca2+-regulated phosphorylation of myosin Va is believed to represent a universal mechanism that regulates the association between myosin Va and its cargo. These observations suggest that by regulating the interaction between myosin V and SGs, Ca2<sup>+</sup> could also control the association between SGs and actin during SG mobilization in the cortical region (124). Importantly, the Ca2+-regulated attachment/release of myosin Va from SGs could be finely coordinated by other molecular motors, such us myosin VI (33). This cooperative model would allow a highly organized and controlled mechanism that regulates SG transport, retention, and anchoring and ultimately SG fusion with the plasma membrane.

#### **MYOSIN VI**

Another member of the myosin family, myosin VI is critical for SV recruitment to the cortical actin network (**Figure 2**). The cellular functions of myosin VI are attributed to its unique ability to generate movement from the plus to the minus end of actin filaments. Myosin VI has an additional unique 53 aa insert, the "reverse gear," between the motor domain and the neck region that has been predicted to be responsible for this exceptional inverted movement directionality (**Figure 2**) (125, 126). Interestingly, this insert binds CaM even though it does not contain a recognizable IQ-CaM motif (127). The tail domain region is the most variable amongst the myosin VI isoforms. Four alternatively spliced isoforms are generated due to the presence of a large insert (21–31 aa), a small insert (9 aa), no insert, or both inserts in this domain (**Figure 2**) (128, 129).

The function of myosin VI depends on the ability of its cargobinding domain (CBD) region to interact with different binding partners that target myosin VI to specific cellular compartments (130). MyosinVI undergoes cargo-mediated dimerization a potential regulatory pathway for all myosins (131, 132). Myosin VI has been linked to clathrin- and non-clathrin-mediated endocytosis, as well as maintenance of Golgi organization and cell polarity. The large and no insert isoforms are the main isoforms mediating these functions (128, 133–137). Myosin VI has also been implicated in autophagy (138), stereocilia maintenance (139), spermatid individualization (140–142), nuclear transcription (143), and cell–cell contacts (144, 145). Evidence of a role of myosin VI in secretion were highlighted by Warner et al. (146) using immortalized cells from Snell's waltzer mice, a strain of myosin VI knock-out mice (146–148). Immortalized fibroblastic cells from these mice have a reduced Golgi complex (∼40% smaller in comparison with that in normal cells) that is accompanied by a similar reduction in constitutive secretion (146). The down-regulation of myosin VI expression using small interfering RNA selectively reduces the secretion of prostate-specific antigen and vascular endothelial growth factor in the prostate cancer cell line LNCaP (149). Myosin VI together with its binding partner optineurin, regulates the final stage of constitutive exocytosis by mechanically controlling the formation of the fusion pore between the SV and the plasma membrane in HeLa cells (150). Less is known about the role of myosin VI in the nervous system (151). Myosin VI is widely and highly expressed in the brain; it is found in synapses and enriched at the postsynaptic density (151). In hippocampal neurons, myosin VI forms a complex with α-amino-3 hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR), AP-2 and synaptic-associated protein 97 (SAP-97), and mediates AMPAR clathrin-mediated endocytosis. Importantly, myosin VI function underpins hippocampal neurons synapses and dendritic spines formation (151). Other work supports the role of myosin VI in neurotransmission by demonstrating that myosin VI; together with its binding partner GIPC1 is necessary for BDNF-TrkB-mediated synaptic plasticity (152).

Myosin VI has a very slow rate of release of ADP from its nucleotide-binding pocket, which therefore slows the dissociation of myosin VI from actin (153–155). Studies carried out in Snell's waltzer mice have shown that myosin VI allows the formation, maturation, and function of sensory hair cells by mediating the attachment of membrane compartments to the F-actin cytoskeleton (148). Together these lines of evidence point toward the possibility that myosin VI could regulate neuroexocytosis by anchoring/recruiting SVs to the actin network before they undergo fusion with the plasma membrane. Although little is known about the precise molecular mechanism(s) underpinning this role, the function of myosin VI in regulated in exocytosis in PC12 cells has been questioned (156). However, *Drosophila* mutants lacking myosin VI display altered neuromuscular junction morphology and synaptic vesicle localization resulting in impaired synaptic plasticity (157). Myosin VI could therefore mediate the mobilization of synaptic vesicles from different functional pools, by a yet to be elucidated mechanism. We recently described a novel role for the myosin VI small insert isoform (**Figure 2**) in regulated exocytosis in PC12 cells (33). Using purified SGs in a pull-down approach followed by mass spectrometry, we identified myosin VI as a cytosolic protein that interacts with SGs in a Ca2+-dependent manner. We found that myosin VI maintains an active pool of SGs near the plasma membrane by tethering them to the cortical actin network (**Figure 3**). This allows the replenishment of the pool of SGs near the plasma membrane and is key to sustaining exocytosis during long periods of stimulation (33). Interestingly, we found that c-Src phosphorylation in a DYD motif located in the CBD of myosin VI small insert is one of the mechanisms controlling its function in regulated neuroexocytosis (33). The mechanisms that target myosin VI to SGs and the regulation of the isoform specific tethering function still need to be elucidated.

There are several other members of the myosin family that could potentially be involved in regulated secretion, including myosin X, a motor protein found predominantly at the tip of filopodia of many cell types including neurons (16, 17). Filopodia are important precursors for dendritic spine and synapse formation but more work is needed to assess whether neurosecretion can occur in these structures.

#### **CONCLUSION**

Understanding the detailed roles of myosins and other accessory proteins in regulated exocytosis is challenging. Although a great deal is known about the involvement of these proteins and their effectors during the different stages of secretion, there is still no comprehensive model of the interplay of the different myosin isoforms, e.g., the transition from myosin Va-mediated directed

#### **REFERENCES**


transport to myosin VI-dependent recruitment to the cortical actin network. Common pathways that are shared by other cellular functions, such as adhesion or migration should also be explored further. Future work should therefore aim at combining *in vitro* techniques with live cell microscopy experiments in order to explore the complex interplay between the different myosin molecular motors during neuroexocytosis. In particular, it will be necessary to address the nature of the pathways, which coordinate and control myosin functions in order to achieve such precise spatio-temporal trafficking of SVs en route to fusion with the plasma membrane.

#### **ACKNOWLEDGMENTS**

The authors thank Rowan Tweedale for critical reading of the manuscript.

actin concentrations in brain dendritic spines and postsynaptic densities. *Proc Natl Acad Sci U S A* (1982) **79**:7590–4. doi:10.1073/ pnas.79.23.7590


Involvement of Rho GTPases in calcium-regulated exocytosis from adrenal chromaffin cells. *J Cell Sci* (1999) **112**(Pt 24):4763–71.


*Mol Biol Cell* (2004) **15**:520–31. doi:10.1091/mbc.E03-06-0402


*Mol Biol* (2013) **20**:679–86. doi:10. 1038/nsmb.2570


*Chem* (2004) **279**:3096–110. doi: 10.1074/jbc.M308878200


clathrin-mediated endocytosis. *Mol Biol Cell* (2012) **23**:2891–904. doi:10.1091/mbc.E11-04-0383


(2003) **116**:3453–61. doi:10.1242/ jcs.00669


**178**:529–40. doi:10.1083/jcb. 200612042


single-headed myosin-VI move processively. *Biophys J* (2006) **90**:3643–52. doi:10.1529/biophysj. 105.075721


**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: 15 August 2013; accepted: 05 October 2013; published online: 21 October 2013.*

*Citation: Papadopulos A, Tomatis VM, Kasula R and Meunier FA (2013) The cortical acto-myosin network: from diffusion barrier to functional gateway in the transport of neurosecretory vesicles to the plasma membrane. Front. Endocrinol. 4:153. doi: 10.3389/fendo.2013.00153*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

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

### Munc13-1 translocates to the plasma membrane in a Doc2B- and calcium-dependent manner

#### **Reut Friedrich<sup>1</sup> , Irit Gottfried<sup>1</sup> and Uri Ashery 1,2\***

<sup>1</sup> Department of Neurobiology, Wise Faculty of Life Sciences, Tel Aviv University, Tel Aviv, Israel

<sup>2</sup> Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Nicolas Vitale, Centre National de la Recherche Scientifique, France Luis M. Gutiérrez, University Miguel Hernandez, Spain

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

Uri Ashery, Department of Neurobiology, The George S. Wise Faculty of Life Sciences, Tel Aviv University, Ramat Aviv, Tel Aviv 69978, Israel e-mail: uriashery@gmail.com

Munc13-1 is a presynaptic protein activated by calcium, calmodulin, and diacylglycerols (DAG) that is known to enhance vesicle priming. Doc2B is another presynaptic protein that translocates to the plasma membrane (PM) upon elevation of internal calcium concentration ([Ca2+]<sup>i</sup> ) to the submicromolar range, and increases both spontaneous and asynchronous release in a calcium-dependent manner. We speculated that Doc2B also recruits Munc13-1 to the PM since these two proteins have been shown to interact physiologically and this interaction is enhanced by Ca2+. However, this calcium-dependent co-translocation has never actually been shown. To examine this possibility, we expressed both proteins tagged to fluorescent proteins in PC12 cells and stimulated the cells to investigate the recruitment hypothesis using imaging techniques.We found that Munc13-1 does indeed translocate to the PM upon elevation in [Ca2+]i, but only when co-expressed with Doc2B. Interestingly, Munc13-1 co-translocates at a slower rate than Doc2B. Moreover, while Doc2B dislocates from the PM as soon as the [Ca2+]<sup>i</sup> returns to basal levels, Munc13- 1 dislocates at a slower rate and a fraction of it accumulates on the PM. This accumulation is more pronounced under subsequent stimulations, suggesting that Munc13-1 accumulation builds up as some other factors accumulate at the PM. Munc13-1 co-translocation and accumulation was reduced when its mutant Munc13-1H567K, which is unable to bind DAG, was co-expressed with Doc2B, suggesting that Munc13-1 accumulation depends on DAG levels. These results suggest that Doc2B enables recruitment of Munc13-1 to the PM in a [Ca2+]<sup>i</sup> -dependent manner and offers another possible Munc13-1-regulatory mechanism that is both calcium- and Doc2B-dependent.

**Keywords: Munc13, Doc2B, calcium, translocation, phorbol ester**

#### **INTRODUCTION**

Munc13-1 is a key player in the final stages of exocytosis. It is a relatively large protein (1,735 aa) with several distinct domains (1). Its Mun domain (aa 859–1531) is responsible for the crucial role of Munc13-1 in exocytosis: the conversion of syntaxin to its open form. This open form of syntaxin interacts with SNAP-25 to form a heterodimer; the heterodimer then interacts with synaptobrevin to form the SNARE complex, which holds the vesicle in close proximity to the plasma membrane (PM) and enables efficient fusion. Munc13-1 has an active C1 domain [aa 567–616 (1, 2)] and has been shown to translocate to the PM upon TPA/PMA stimulation (3). Munc13-1 also has three active C2 domains (4–8), and it interacts with members of the double-C2 domain (Doc2) protein family via residues 851–1461 (3).

Doc2A and B are high-affinity sensors of internal calcium concentration ([Ca2+]i), containing tandem C2 domains that are responsible for their Ca2+-dependent PM targeting and that promote priming and fusion (9, 10). Doc2A interacts with Munc13-1 through its Munc13-interacting domain (Mid, aa 13–37), which is highly conserved in Doc2B [92%, (3)]. The interaction between Munc13 and Doc2 has been demonstrated in both cell-free and intact cell systems. In the yeast two-hybrid system, Munc13-1 and Munc13-2 were shown to interact with both Doc2A and Doc2B (3). Co-immunoprecipitation of Munc13-1 with Doc2A from PC12 cells was markedly enhanced when the cells were stimulated by TPA or high K<sup>+</sup> in the presence of extracellular calcium (3). A growth hormone assay in PC12 suggested that this interaction has a physiological role; overexpression of the Doc2-interacting domain of Munc13 reduced the Ca2+-dependent exocytosis from PC12 cells, and co-expression with Doc2 suppressed this reduction (3). However, expression of Mid alone in PC12 cells had no effect on the number of docked vesicles (11).

The physiological relevance of this interaction was then further tested in neurons. Introduction of the Mid peptide into presynaptic neurons of cholinergic synapses reversibly inhibited synaptic transmission evoked by action potentials (12). In contrast, the scrambled Mid peptide did not inhibit synaptic transmission. Recordings from the calyx of Held revealed that presynaptic loading of a synthetic Mid peptide significantly attenuates phorbolester (PE) induced synaptic potentiation, whereas the scrambled Mid peptide had no effect (13).

Imaging experiments revealed that Doc2B translocates to the PM when it is co-expressed with Munc13-1 following PE application (14). Doc2A and Munc13-4 relocated to the cell periphery together with the secretory lysosome marker CD63 upon Ag stimulation in a calcium-free medium (15). This interaction depended on the Mid domain of Doc2A although C2B also seemed to play a role (15). Both Doc2B and Munc13-1 are expressed in chromaffin cells (10, 16). Based on the fact that Doc2B translocates to the PM upon calcium elevation and interacts with Munc13-1, and on the above data, we hypothesized that Doc2B will efficiently recruit Munc13-1 to the PM in a [Ca2+]i-dependent manner. We found that Munc13-1 co-translocates to and accumulates at the PM when co-expressed with Doc2B in a calcium-dependent manner.

#### **MATERIALS AND METHODS**

#### **CELL LINES**

PC12 cells (a generous gift from Dr. Nicolas Vitale, CNRS-UPR, Strasbourg, France) were grown in DMEM (Biological Industries, Beit Haemek, Israel) supplemented with glucose (4,500 mg/l) and l-glutamine (Gibco-BRL), and containing 5% (v/v) fetal calf serum, 10% (v/v) horse serum, and 100 U/ml penicillin/streptomycin (Biological Industries). The cells were split regularly to maintain confluence and were kept in a 37°C, 5% CO2-humidified incubator. Transfection was performed using Lipofectamine™ 2000 (Invitrogen) according to the manufacturer's instructions at a 2:3 (DNA:reagent) ratio. Cells were imaged 16–32 h after transfection.

COS-7 cells were grown in DMEM supplemented with 10% fetal calf serum, 2 mM l-glutamine, and 100 U/ml penicillinstreptomycin. The cells were split regularly to maintain confluence and were kept in a 37°C, 5% CO2-humidified incubator. Transfection was performed using Jet-PEI (PolyPlus Transfection, New York, NY, USA) according to the manufacturer's instructions at a 1:2 (DNA:reagent) ratio.

#### **DNA CONSTRUCTS**

All plasmids encoding fluorescently labeled Doc2B and the Doc2B–glutathione-*S*-transferase (GST) plasmids were a generous gift from Dr. Alexander J. Groffen (Vrije Universiteit, Amsterdam, The Netherlands) and the control GST construct was a generous gift from Prof. Joel Hirsch (Tel Aviv University, Israel). The sequences of all constructs were verified by automated DNA sequencing. Munc13-1 plasmids are a kind gift from Prof. N. Brose (Max-Plank institute, Gottingen, Germany).

#### **WESTERN BLOTTING**

Western blot experiments were performed according to standard procedures. In general, protein extracts (for overexpression experiments ∼10µg protein, for endogenous experiments ∼100µg protein) were loaded on an SDS-polyacrylamide (8 or 11%) gel and electrophoresed with a constant current of 30 mA for each gel. The proteins from the gel were transferred to a nitrocellulose membrane by electroblotting at a constant current of 400 mA for 1 h. The membrane was incubated in blocking solution (5% milk powder) overnight at 4°C with gentle agitation. After five washes, we incubated the membrane with primary rabbit anti-Doc2B antibody for 1 h at room temperature (diluted 1:500 in 1% BSA with 0.05% azide). The membrane was washed five times and incubated with the relevant secondary horseradish peroxidase-conjugated antibody for 45 min at room temperature (diluted 1:15,000 in milk). Then the membrane was washed six times, and detection was performed using enhanced chemiluminescence solution (Pierce) and exposure to Super RX film (Fuji).

#### **CO-TRANSLOCATION EXPERIMENTS**

For the translocation experiment, PC12 cells were plated on glass coverslips, transfected with different combinations of plasmids (see Results for details) encoding fluorescently tagged proteins – Doc2Bwt–mRFP, Doc2BD218,220N–mRFP, Doc2BMid– mRFP,Munc13-1wt–EGFP,Munc13-1H567K–EGFP, or EGFP alone using Lipofectamine 2000. Imaging was performed 16–32 h posttransfection. Cells were perfused constantly with external solution and excited using a high K<sup>+</sup> solution (70 mM) as described in Groffen et al. (9). The imaging setup consisted of an Olympus IX-70 inverted microscope with a 60× total internal reflection fluorescence (TIRF) objective (Olympus), a TILL Photonics TIRF condenser (Gräfelfing, Germany), two solid-state lasers (Laser Quantum, Stockport, UK) emitting at 473 and 532 nm, an Andor Ixon 887 EMCCD camera (Belfast, Northern Ireland), and a dualview beam-splitting device (Optical Insights, Roper Bioscience, Tucson, AZ, USA). Time-lapse images were taken every 300 ms. The equipment was controlled byMetamorph software (Molecular Devices, Downingtown, PA, USA), which was also used to perform the analysis. Confocal images were recorded using a 63× objective of Zeiss LSM 510 META microscope equipped with 30 mW 488 nm Argon laser and 15 mW 561 nm DPSS laser.

#### **RESULTS**

#### **Doc2B AND Munc13-1 CO-TRANSLOCATE UPON ELEVATION OF [Ca2**+**]<sup>i</sup>**

To investigate our hypothesis that co-expression of Doc2Bwt and Munc13-1 results in co-translocation of both proteins to the PM upon [Ca2+]<sup>i</sup> elevation, we expressed Munc13-1 fused to EGFP in PC12 cells together with Doc2Bwt fused to mRFP. The cells were subjected to three short (20 s) KCl applications (with relaxation times between each application) which caused elevation of [Ca2+]<sup>i</sup> and translocation of Doc2B and Munc13-1 to the PM (**Figure 1**). In the absence of calcium, the translocation of both proteins was abolished suggesting it is calcium dependent (Figure S1 in Supplementary Material). Munc13-1 translocated to the PM upon elevation in [Ca2+]<sup>i</sup> , but only when co-expressed with Doc2Bwt (**Figures 1** and **2**). When Munc13-1 was expressed in PC12 cells without Doc2B it did not translocate to the PM upon high K<sup>+</sup> stimulation (Figure S2 in Supplementary Material). To establish a more quantitative connection between Doc2Bwt translocation and that of Munc13-1wt, we repeated this experimental protocol using TIRF microscopy, focusing on the fluorescence of the cell's lower PM (**Figure 2A**). Co-translocation of Munc13-1wt and Doc2Bwt was clearly seen in the TIRF plane (**Figure 2A**). Examining the translocation kinetics revealed that Munc13-1wt translocates to the PM at a slower rate than Doc2Bwt (**Figure 2B**; *n* = 13). Moreover, Munc13-1wt translocation, at least in the second and third application, peaked when Doc2Bwt was already in its dislocating phase (**Figure 2B**). Munc13-1 dislocation also seemed to occur at a slower rate than that of Doc2B. It is interesting to note that although Doc2Bwt fluorescence returned to its initial level after each application, Munc13-1wt started to accumulate at the PM after the first application, though this became more apparent after

**FIGURE 1 | Co-translocation of Doc2Bwt and Munc13-1wt**. Confocal images of a PC12 cell co-expressing Munc13-1wt-EGFP (left) and Doc2Bwt-mRFP (middle). Merged images presented on the right. **(A)** At basal calcium both proteins have a cytoplasmic distribution. **(B)** At the initial phase of the calcium elevation (2–10 s after KCl application), translocation of Doc2B occurs, but Munc13-1 is still mainly cytoplasmic. **(C)** Munc13-1wt translocation is seen only at later stages of the experiment when Doc2B concentration on the membrane is higher (10–20 s after KCl application).

the second and third application. It is also interesting to note that although the translocation of Munc13wt to the PM appeared rather uniform in the confocal images (**Figure 1**), it showed a patchy appearance in TIRF images (**Figure 2A**). This patchy pattern was not unique to Munc13-1wt's Doc2Bwt-dependent translocation; a similar dotted staining pattern appeared in PE-induced translocation of Munc13-1wt (**Figure 3**). As PE mimics the interaction of Munc13-1 with diacylglycerol (DAG), it is possible that when on the PM, Munc13-1 interacts with DAG.

We have previously shown that a mutated form of Doc2B– Doc2BD218,220N is constitutively associated with the PM. We therefore examined the effect of Doc2BD218,220N on Munc13- 1 wt's distribution in the cell. Co-expression of these two proteins resulted in the constant translocation of both Doc2BD218,220N and Munc13-1wt (**Figure 4**). As in **Figure 2**, Munc13-1wt showed a patchy pattern at the PM.

#### **CO-TRANSLOCATION OF Munc13-1 AND Doc2B DEPENDS ON THE Doc2B–Munc13-1 INTERACTION AND ON Munc13-1s C1 DOMAIN**

To determine whether the Mid domain of Doc2B is responsible for the co-translocation of Doc2B and Munc13-1, we repeated the translocation experiment with Doc2B harboring a scrambled Mid domain (Doc2BMid), and examined if this mutation abolishes translocation of Munc13-1wt. Scrambling the Mid domain of Doc2B disrupted the interaction with Munc13-1wt [Figure S3 in Supplementary Material; (12)]. Although the Mid mutation disrupted most of Munc13-1wt's translocation, some degree of translocation still existed (**Figure 5**). The translocation was barely detected in the first KCl application but in later applications,

**FIGURE 2 | Co-translocation of Doc2Bwt–mRFP and Munc13-1wt–EGFP in the TIRF plane**. **(A)** TIRF images of a PC12 cell expressing Munc13-1wt–EGFP (upper panel) and Doc2Bwt–mRFP (lower panel). From left to right: basal state, beginning of translocation, maximum translocation, and back to basal state. **(B)** Quantification graph of the translocation kinetics observed in the cell's lower PM. Red line indicates Doc2Bwt–mRFP, green line indicates Munc13-1wt–EGFP (n = 13). Yellow arrows indicate KCl application.

panel) after PE application. Munc13-1wt translocated to the PM while Doc2Bwt did not. **(B)** TIRF images of a PC12 cell co-expressing Munc13-1wt–EGFP (upper panel) and Doc2Bwt–mRFP (lower panel) after PE application. Munc13-1wt translocated to the PM while Doc2Bwt did not. Note the patchy nature of Munc13-1wt 's translocation at the PM.

some Munc13-1 showed translocation and accumulation on the PM. These results suggest that the Mid domain is important for Munc13-1 co-translocation; however, Munc13-1 also accumulates slowly at the PM without the interaction with Doc2B Mid domain, possibly via its C1–DAG interaction.

The C1 domain of Munc13 is important for its membraneattachment ability via its interaction with DAG (17, 18), and we therefore examined whether the C1 domain is also important for

**FIGURE 4 | A PC12 cell co-expressing Doc2BD218,220N–mRFP and Munc13-1wt** . **(A)** Epi-fluorescence images of a PC12 cell co-expressing Doc2BD218,220N–mRFP (left) and Munc13-1wt–EGFP (right). **(B)** TIRF images of a PC12 cell co-expressing Doc2BD218,220N–mRFP (left) and Munc13-1wt–EGFP (right).

Doc2B-induced Munc13-1 translocation and accumulation. We used a Munc13-1 mutant that does not bind DAG (Munc13- 1 H567K) and cannot translocate to the PM upon PE stimulation, and examined whether Doc2B could induce Munc13- 1 H567K translocation. Munc13-1H567K displayed lower translocation ability than Munc13-1wt (**Figure 6** compared to **Figure 2**) and did not accumulate on the PM like its wild-type counterpart. Thus, it seems that the initial translocation of Munc13- 1 depends on interaction with Doc2B and soon after, the C1–DAG interaction determines the accumulation of Munc13-1 at the PM.

#### **DISCUSSION**

Munc13-1 is a key player in the synapse; the activity of this multidomain protein is tightly regulated by many factors, including calmodulin (5, 19, 20), DAG (18, 21), and Ca2<sup>+</sup> in a PIP<sup>2</sup> dependent manner (7). We show here that Doc2B recruits Munc13-1 to the PM in a Ca2+-dependent manner, thereby suggesting another possible Munc13-1-regulatory mechanism.

The Doc2 family of proteins interacts with the Munc13 family of proteins primarily via the Mid domain, located within the Nterminal domain of Doc2 (3, 12, 14, 15, 22). Doc2 translocates to the PM upon elevation of [Ca2+]<sup>i</sup> . Munc13-1, on the other hand, does not translocate to the PM upon calcium elevation. However, when co-expressed with Doc2B, Munc13-1 co-translocates to the PM following Doc2B. This translocation could not be detected with endogenous Doc2B and Munc13-1 as these proteins are expressed in low levels in neuroendocrine cells (10, 16, 19) and such translocation, if occurs might be undetectable under these conditions. Munc13-1's translocation is most likely mediated through the Mid domain as reflected by its reduction upon mutation in the Mid domain. When the elevation of calcium is brief, Munc13-1 translocation is reversible, dislocating back to the cytosol after Doc2B dislocation. However, following repeated stimulations, Munc13-1 starts to accumulate at the PM even after Doc2B has dislocated back to the cytosol. This accumulation was abolished when a Munc13-1 mutant that does not bind DAG (Munc13-1H567K) was co-expressed with Doc2B suggesting that the C1–DAG interaction determines Munc13-1's accumulation at the PM. It is possible that during high-frequency activity, e.g., sustained or intermittent depolarization, the level of DAG increases (23). This would cause a more stable interaction of Munc13-1 with the PM via its C1 domain, anchoring it to the PM, and enabling its catalytic activity in the fusion step (2). Thus, according to this working model, the activity of Munc13-1 at the PM depends on the stimulation's frequency and on Doc2B translocation, becoming more prominent during periods of high activity. Doc2B has been recently shown to enhance neuronal network activity by specifically increasing the firing rate within a neuronal

burst (24). Therefore, it is possible that in addition to the direct effect of Doc2B on asynchronous release, it also recruits Munc13-1 during periods of intense activity, which might also contribute to the enhanced network activity. A similar mode of action has been suggested for protein kinase C (PKC) activity (25). Many receptor stimuli induce calcium signals prior to a more persistent increase in DAG concentration. These calcium signals have only a minimal effect on conventional PKC activity in the absence of DAG. However, in the presence of DAG, each calcium spike induces a more pronounced activation cycle of conventional PKC.

Our data suggest that the interaction between Doc2B and Munc13-1 depends on more than just the Mid domain of Doc2B since a small degree of translocation was still observed when Munc13-1 was co-expressed with Doc2BMid. We hypothesize that this interaction is also dependent on the C1 domain of Munc13- 1 as it has been reported that deleting this domain increases Munc13-1's interaction with Doc2B (3). It is also possible that the C2B of Doc2B contributes to this interaction, as has been recently suggested for Doc2A and Munc13-4 (15).

A previous study described co-translocation of Doc2Bwt and Munc13-1wt upon PE stimulation (14). Such translocation was not observed here, either in the epi-fluorescence or TIRF imaging. Hence, the interaction of Munc13-1's C1 domain with DAG at the PM may compete with Doc2B binding to Munc13-1. Therefore, stimulating the cells with PE causedMunc13-1's translocation but interfered with Doc2B translocation. These findings, together with the observation that Munc13-1 reaches maximal translocation when Doc2B is already dislocating from the PM due to a decrease in [Ca2+]<sup>i</sup> , suggest that the interaction of Munc13-1 with DAG at the PM cannot occur when it is in a complex with Doc2B. Therefore, DAG binding to Munc13-1 might disrupt Doc2B binding. These findings suggest that Doc2B recruits Munc13-1 to the PM but once there, Munc13-1 associates through its C1 domain with DAG and this reduces its interaction with Doc2B. Further experiments will be needed to validate this hypothesis.

Co-expression of Doc2BD218,220N with Munc13-1wt revealed that the two proteins are being constantly translocated to the PM, yet the Doc2BD218,220N mutant could not support refilling during repeated stimulations (10). Combining these two observations

#### **REFERENCES**


**272**:16081–4. doi:10.1074/jbc.272. 26.16081


suggests a possible physiological effect of the Doc2B–Munc13- 1 interaction: this complex needs to undergo an on–off cycle from the PM to achieve a full priming effect. Another possible explanation is that Munc13-1 needs to detach from Doc2B at the PM to enable its priming effect. Taken together, our data support the hypothesis that Doc2B serves as a calcium-dependent recruitment factor for Munc13-1 whereas at the PM, Munc13-1 interacts with DAG. This provides an activity-dependent recruitment mechanism for two major synaptic proteins, Doc2B and Munc13-1.

#### **ACKNOWLEDGMENTS**

We would like to thank Mrs. Lirin Michaeli for comments and assistance on the manuscript and Prof. Alexander J. Groffen for providing DNA constructs. This work was supported in part by the Israel Science Foundation (Grant no. 1211/07 and 730/11; Uri Ashery), the BSF (Grant no. 2009279; Uri Ashery) and the German-Israeli Foundation (GIF) (Grant no. 1125-145.1/2010; Uri Ashery).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/Neuroendocrine\_Science/10.3389/ fendo.2013.00119/abstract

### **Figure S1 | Doc2Bwt and Munc13-1wt do not translocate to the PM in the**

**absence of calcium**. Epi-fluorescence images of a PC12 cell co-expressing Doc2Bwt-mRFP (left) and Munc13-1wt-EGFP (center). Merged images presented on the right. In the upper panel, the cell in its basal state. In the lower panel, the cell after application of depolarizing high K<sup>+</sup> solution without calcium (containing 0.1 mM EGTA). Note there is no evident change in the proteins distribution in the cell.

**Figure S2 | Munc13-1wt does not translocate to the PM in the absence of Doc2B**. Epi-fluorescence images of a PC12 cell expressing Munc13-1wt–EGFP, before (left) and after (right) application of depolarizing high K<sup>+</sup> solution. Note there is no evident change in the protein's distribution in the cell.

**Figure S3 | GST pull-down assay of Doc2B N-terminal with Munc13-1**. GST fusion of the N-terminal of Doc2Bwt and N-terminal of Doc2BMid binding to Munc13-1wt (Upper panel) and Munc13-1H567K (lower panel). Munc13-1wt and Munc13-1H567K show binding only to N-terminal of Doc2Bwt and not to N-terminal of Doc2BMid .

Munc13-1 homodimer to Munc13- 1/RIM heterodimer switch. *PLoS Biol* (2006) **4**:e192. doi:10.1371/ journal.pbio.0040192


neuronal activity with unique calcium-dependent and kinetic properties. *J Neurochem* (2006) **97**:818–33. doi:10.1111/j.1471- 4159.2006.03755.x


et al. Munc13-1 is a presynaptic phorbol ester receptor that enhances neurotransmitter release. *Neuron* (1998) **21**:123–36. doi:10. 1016/S0896-6273(00)80520-6


modulate spontaneous and Ca2+ evoked transmitter release via acting on both Munc13 and protein kinase C. *J Neurosci* (2008) **28**:8257–67. doi:10.1523/JNEUROSCI.0550-08. 2008


**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: 15 June 2013; accepted: 27 August 2013; published online: 17 September 2013.*

*Citation: Friedrich R, Gottfried I and Ashery U (2013) Munc13-1 translocates to the plasma membrane in a Doc2B- and calcium-dependent manner. Front. Endocrinol. 4:119. doi: 10.3389/fendo.2013.00119*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Friedrich, Gottfried and Ashery. 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.*

REVIEW ARTICLE published: 04 December 2013 doi: 10.3389/fendo.2013.00187

# CAPS and Munc13: CATCHRs that SNARE vesicles

### **Declan J. James† and Thomas F. J. Martin\*†**

Department of Biochemistry, University of Wisconsin, Madison, WI, USA

#### **Edited by:**

Stephane Gasman, Centre national de la recherche scientifique, France

#### **Reviewed by:**

Thomas Söllner, University of Heidelberg, Germany Jingshi Shen, University of Colorado Boulder, USA

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

Thomas F. J. Martin, Department of Biochemistry, University of Wisconsin, 433 Babcock Drive, Madison, WI 53706, USA e-mail: tfmartin@wisc.edu

†Declan J. James and Thomas F. J. Martin have contributed equally to this work.

#### **TRAFFICKING IN THE SECRETORY AND ENDOSOMAL PATHWAYS**

The transport of proteins and membranes in the secretory pathway is vectorial with vesicle formation in a donor compartment coupled to vesicle transport and subsequent fusion in an acceptor compartment. Vesicle delivery to an acceptor membrane involves several layers of interaction that confer targeting specificity involving tethering, docking, and priming of vesicles. These lead to SNARE pairing that mediates fusion of the vesicle with the acceptor membrane. In exocytic vesicle fusion with the plasma membrane, as well as for intracellular membrane fusion events, a diverse set of accessory (tethering and priming) factors are required to prime vesicles for fusion. Accessory factors commonly interact with vesicle and target membrane constituents that are characteristic of a membrane compartment such as Rab proteins (1, 2) and phosphoinositides (3, 4). Accessory factors also interact with SNARE proteins to promote SNARE protein complex assembly usually in association with proteins of the Sec1/Munc18 (SM) family. Accessory factors for constitutive trafficking include the tethering factor complexes Dsl1 for Golgi to ER transport, HOPS for late endosome fusion, and exocyst for exocytic fusion. Accessory factors for regulated vesicle exocytosis include the priming factors CAPS and Munc13. These exhibit sequence and structural similarity with tethering factor subunits (e.g., exocyst Sec6) (5, 6), which suggests there may be common features for these diverse accessory factors. We review aspects of tethering and priming factor function at several trafficking stations attempting to identify common operating principles.

#### **SNARE PROTEINS IN MEMBRANE FUSION**

Biochemical reconstitution studies of membrane trafficking in the Golgi led to the discovery of SNARE (soluble *N*-ethylmaleimide sensitive factor attachment protein receptor) proteins as the general machinery for membrane fusion (7). The initial identification

CAPS (Calcium-dependent Activator Protein for Secretion, aka CADPS) and Munc13 (Mammalian Unc-13) proteins function to prime vesicles for Ca2+-triggered exocytosis in neurons and neuroendocrine cells. CAPS and Munc13 proteins contain conserved C-terminal domains that promote the assembly of SNARE complexes for vesicle priming. Similarities of the C-terminal domains of CAPS/Munc13 proteins with Complex Associated with Tethering Containing Helical Rods domains in multi-subunit tethering complexes (MTCs) have been reported. MTCs coordinate multiple interactions for SNARE complex assembly at constitutive membrane fusion steps. We review aspects of these diverse tethering and priming factors to identify common operating principles.

**Keywords: CAPS (aka CADPS), Munc13, priming factors, vesicle fusion, SNAREs, multi-subunit tethering complexes**

of neuronal syntaxin-1, SNAP-25, and VAMP2 (aka synaptobrevin2) as SNARE proteins in brain membrane extracts prompted advances for understanding Ca2+-triggered vesicle fusion events for neurotransmitter and peptide secretion (8, 9). An essential role for neuronal SNAREs in regulated vesicle exocytosis was indicated by finding them to be the substrates for the *Clostridial* zinc endopeptidase toxins (10–14). The further characterization of proteins in the SNARE protein superfamily generally facilitated research on membrane trafficking throughout the secretory and endosomal pathways (15). Much current research in membrane fusion is focused on understanding how accessory factors prime vesicles for fusion by regulating SNARE complex assembly.

SNARE proteins, usually C-terminal tail-anchored membrane proteins with membrane-proximal helical SNARE motifs, are grouped into syntaxin, SNAP-25, and VAMP families based on sequence relatedness, and referred to as "Q" or "R" SNAREs based on highly conserved glutamine or arginine residues in the zero layer of the SNARE motif (e.g., syntaxin-1 as Qa-, SNAP-25 as Qbc-, VAMP2 as R-SNARE) (16). The reconstitution of SNARE proteins into liposomes demonstrated that SNAREs are sufficient for mediating membrane fusion (17). Although the details of how SNARE proteins fuse membranes is emerging, structural, and biochemical analyses indicate that SNARE proteins present on two apposed membranes form a trans-SNARE complex to pin membranes close together (18, 19). Membrane fusion ensues when trans-SNARE complexes zipper-up through coiled-coil interactions in helical SNARE motifs (16, 20). The formation of a tight coiled-coil bundle, characteristically containing 3Q (Qa, Qb and Qc, or Qa and Qbc) and 1R SNARE motifs, is coupled to membrane fusion events throughout the secretory pathway (21). For regulated vesicle exocytosis, SNARE complex assembly is thought to proceed by a two-stage process with the initial formation of heterodimeric QaQbc complexes of plasma membrane syntaxin-1 with SNAP-25 followed by the insertion of the vesicle R-SNARE VAMP2 to form heterotrimeric (RQaQbc) SNARE complexes (8). Alternative assembly pathways have been suggested (22–25).

At least 44 SNARE protein isoforms in vertebrate cells are distributed throughout membrane trafficking pathways (15). It was proposed that unique cognate SNARE pairing contributes to specificity for vesicle targeting to acceptor membranes (26, 27). However, it has been noted that SNARE pairing can be promiscuous and may not be the sole determinant of vesicle targeting specificity (28, 29). Targeting specificity is likely combinatorial consisting of multiple levels of interaction requiring accessory factors that are recruited to membranes by interactions with Rabs, phosphoinositides, and SNAREs (26). Accessory factors acting with SM proteins promote stages of SNARE complex assembly and enable specific SNARE pairing for fusion. For example, recent studies revealed differential effects of the SM proteins Munc18-1 and Munc18c for enabling fusion on cognate but not on non-cognate SNARE proteins (30, 31).

#### **SM PROTEINS IN MEMBRANE FUSION**

The SM (Sec1/Munc18) protein family consists of soluble proteins that are required for membrane trafficking (32–34). SM proteins are grouped into four highly conserved subfamilies across eukaryotes. In spite of their sequence homology, SM proteins encode a high degree of specificity for SNARE protein interactions. A common theme is SM protein-directed interaction with Q-SNAREs at exocytic (Munc18/Sec1), ER-Golgi (Sly1), endosomal-lysosomal (Vps33), and endosomal (Vps45) membrane trafficking stations (32, 35–38). The mode of Q-SNARE-binding by different SM proteins appears to differ but the SM proteins may generally function in stabilizing SNARE protein complexes (39). Studies on the interaction of the neuronal SM protein Munc18-1 with the Qa-SNARE syntaxin-1 have played a prominent role in understanding SM protein function even though Munc18-1 has unique features that distinguish it from other SM proteins (32). Munc18-1 chaperones syntaxin-1 to the plasma membrane (32, 39). At the plasma membrane, Munc18-1 stabilizes a closed form of syntaxin-1, which is unable to form heterodimeric complexes with the Qbc-SNARE SNAP-25 (40). The closed configuration of syntaxin-1 may prevent unwanted interactions with SNARE proteins as the complex traffics to the plasma membrane (41). Eliminating Munc18-1 reduces syntaxin-1 delivery to the plasma membrane, abrogates dense-core vesicle (DCV) docking, and abolishes triggered exocytosis (42).

Recent findings that Munc18-1 accelerates SNARE-catalyzed liposome fusion help to reconcile the role of Munc18-1 as a chaperone with its essential role in regulated exocytosis (30). Munc18-1 stimulates trans-SNARE complex formation and membrane fusion but does so by switching from an inhibitory to a stimulatory mode. The switch from an inhibitory to a stimulatory mode in liposome fusion for Munc18-1 requires pre-incubation with both R- and Q- SNARE proteins (VAMP2 and syntaxin-1/SNAP-25) (43), which suggests that Munc18-1 utilizes specific interaction sites on the SNARE proteins. Recent studies have identified some of these sites on VAMP2 (C-terminal) and syntaxin-1 (N- and C-terminal) (30, 31). Interestingly, Munc18c appeared to operate differently on its cognate SNAREs lacking an inhibitory mode on syntaxin-4. In addition, recognition sites on shared cognate VAMP2 differed with C-terminal sites for Munc18-1 and N-terminal sites for Munc18c (30, 31). While both Munc18-1 and Munc18c promote the assembly of cognate SNARE protein complexes, they appear to do so by distinct sets of interactions that play a role in establishing specific SNARE protein pairing.

Tethering and priming factors act with SM proteins to promote SNARE complex assembly. Munc13-1 is proposed to enhance the switching of Munc18-1 from an inhibitory to stimulatory mode for regulated vesicle exocytosis (44) as discussed below. At other fusion events, cognate SM proteins also function in concert with accessory factors. In vacuolar fusion, the SM protein Vps33 operates as a subunit of a HOPS (homotypic fusion and vacuole protein sorting) complex, a multi-subunit tethering complex that promotes trans-SNARE complex assembly (45). At other trafficking stations where SM proteins are not formally part of complexes, it is likely that tethering and priming factors cooperate with SM proteins in the assembly of SNARE complexes as noted below.

#### **TETHERING FACTORS INTEGRATE SNARE PROTEIN FUNCTION FOR CONSTITUTIVE FUSION**

Tethering is considered to be a long-range interaction of a vesicle near a target membrane independent of cytoskeletal anchoring. Tethering factors may also operate to bring vesicles into closer proximity for trans-SNARE complex assembly (docking). Tethering factors have been classified as either long coiled-coil proteins or multi-subunit tethering complexes (MTCs). Several MTCs that function at distinct membrane trafficking steps have been identified (46) where they function as important interfaces between Rabs, phosphoinositides, and SNARE proteins (47). Tethering factors are very diverse but sequence comparisons indicate a subtle relatedness among subunits of a subset of MTCs in predicted coiled coils (48). Structural analysis of several MTC subunits from COG, Dsl, exocyst, and GARP complexes indicate a homologous tertiary structure composed of an extended rod-like domain made up of helical bundles (49, 50). MTCs with this structural signature were termed members of a Complex Associated with Tethering Containing Helical Rods (CATCHR) family (50). The CATCHR homology is not restricted to MTC subunits and is found in Myo2p (Myosin V) and in CAPS and Munc13 proteins (see below) (5). In the following, we discuss a well-studied MTC (HOPS) whose subunits lacks CATCHR homology followed by two other well-studied MTCs (exocyst and Dsl1p) that contain subunits with CATCHR homology.

The HOPS complex is an MTC of ∼663 kDa comprised of six subunits Vps41,Vps11,Vps18,Vps16,Vps33, and Vps39 (**Figure 1**, upper) (45). Conserved across eukaryotes, HOPS was initially identified for its role in yeast vacuole fusion, which is a correlate of endo-lysosomal fusion in vertebrate cells (48). The HOPS complex interacts with lipids, Rabs, and SNARE proteins (45). As a complex, HOPS mediates vesicle tethering as well as membrane fusion (51). Ultra-structural analysis of purified HOPS by EM-reconstruction techniques revealed that the holo-complex is a 30 nm long molecular monolith (52, 53). A combination of biochemical and structural reconstruction indicated that its Rabinteracting subunits Vps41 and Vps39 are positioned at either end of the complex (53). This arrangement of Rab-binding domains may mediate HOPS-dependent vesicle tethering. The SM protein Vps33 and nearby Rab-binding Vps39 are juxtaposed toward the

and contains the SM protein Vps33. Lower panel: schematic of CAPS membrane where HOPS may couple vesicle tethering with SNARE complex assembly. The monolith-like HOPS would stand like Stonehenge coordinating vacuole–vacuole fusion sites arranged around a vertex ring (45). Fusion between large organelles such

as vacuoles likely requires solid tethering foundations provided by

cognate SNAREs and co-functions with the SM protein Sly1. The HOPS complex with six subunits interacts with cognate SNAREs, PI3P, and PIP2, form trans-SNARE complexes in priming **(B)**. **(iv)** Full SNARE complex zippering in response to elevations of intracellular calcium mediated by synaptotagmin and complexin triggers vesicle fusion **(C)** and contents release into the extracellular space.

the HOPS complex for coordinating large surface area membrane fusion events. HOPS also requires the phosphoinositides PI3P and PI(4,5)P<sup>2</sup> to coordinate its function (45). Overall, HOPS promotes the formation of a 3Q:1R trans-SNARE complex consisting of Vam3(Qa), Vti1(Qb), Vam7(Qc), and Nyv1(R) (54). Besides the requirement of the SM protein subunit Vps33 in SNARE complex formation, Vps16 and Vps18 interact with the soluble Qc-SNARE Vam7 and mediate a rate-limiting step for Qc-SNARE entry into a fusion-competent SNARE complex (55). The SM protein subunit Vps33 also interacts with the SNARE-binding Vps16 subunit (56), which may together with Vps18 form a SNARE-binding interface for the HOPS complex. The HOPS complex lacks the CATCHR domain homology but is an example of an MTC that integrates Rab, phosphoinositide, and SNARE interactions for tethering and compartment-specific SNARE complex assembly.

The exocyst complex, originally discovered in yeast, is a conserved multi-subunit complex ∼750 kDa composed of eight subunits Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70, and Exo84 (**Figure 1**, upper) (57). Exocyst functions in polarized constitutive exocytosis in budding yeast, plants, and vertebrate cells, and is thought to tether vesicles to exocytic sites at the plasma membrane. Ablation of exocyst components in yeast results in the mis-localization of the complex and an accumulation of vesicles within the cell interior. At the plasma membrane, the exocyst is an octomeric holo-complex but the exact pathway for assembly of the complex is under active study. Studies in yeast suggest that vesicle tethering is achieved by complex assembly initiated between vesicle-bound Sec4 (Rab)-Sec15 and plasma membranetargeted Exo70 and Sec3 subunits (58). The fully assembled exocyst complex localizes to growth cones in neurons where it plays an important role in membrane addition (59). The plasma membrane recruitment of exocyst is mediated through interactions of exocyst subunits Exo70 and Sec3 with GTPases (Rho and cdc42 family) and the phosphoinositide lipid PI(4,5)P<sup>2</sup> (60, 61).

Structural studies indicate that Sec6, Sec15, Exo70, and Exo84 subunits contain homologous CATCHR domains. These CATCHR domains mediate inter-subunit interactions to provide an elongated structure for tethering as well as interactions with other proteins (GTPases) (62). Besides a role in tethering, exocyst subunits interact with SNARE and SM proteins to control SNARE complex assembly (57). The exocyst subunit Sec6, as a dimer, interacts with the Qbc-SNARE Sec9 and inhibits the formation of Qabc acceptor SNARE complexes (63). In addition, Sec6 interacts with the SM protein Sec1 when it is part of the exocyst complex mediated by N-terminal sites in Sec6, which functionally overlap with Sec9 binding sites (64). Truncation of these N-terminal sites in Sec6 (the CATCHR domain is C-terminal) inhibits dimerization as well as Sec9 and Sec1 binding. In contrast, mutations in the Sec6 CATCHR domain disrupt exocytosis and polarized localization of exocyst but not exocyst complex formation or Sec6-Sec9 interactions (64). It is not known whether these Sec6 CATCHR domain mutations impair exocyst-Sec1 interactions. Nevertheless, it appears that Sec6 mediates interactions that anchor the exocyst to sites of polarized exocytosis (65). One speculation is that Sec6-Sec9 interactions may help stage the assembly of a Qabc-SNARE complex concomitant with arrival of the vesicle containing other exocyst components to promote trans-SNARE complex formation mediated by exocyst-Sec1 interactions. Notably the formation of a Q-SNARE complex by exocyst is similar to the HOPS recruitment of Vam7. In summary, the exocyst complex tethers through vesicle Rab (Sec4) and plasma membrane phosphoinositide and GTPase interactions, and associates with the SM

protein Sec1 and SNAREs to promote or stabilize SNARE complex assembly.

The Dsl1 complex is the smallest of the MTCs at ∼250 kDa and is composed of three subunits Dsl1, Tip20, and Sec39 (**Figure 1**, upper) (66, 67). As an essential protein complex in yeast, it is required for the fusion of Golgi-derived vesicles with the ER. Two of the three subunits have homologs in humans that are involved in retrograde trafficking pathways between the Golgi and ER (48). Structural studies of the yeast Dsl1 complex indicate that Sec39 and Tip20 subunits are bridged by Dsl1 through interactions of the CATCHR domain of Dsl1 with that of Tip20 to assemble a 20 nm structure (68). The Dsl1 subunit interacts with the COP1 coat complex and may serve as a direct link securing incoming Golgi-derived COP1 vesicles at the ER membrane (69). The Dsl1 complex interacts with ER resident SNAREs Ufe1(Qa), Sec20(Qb), and Use1(Qc) to stabilize a Qabc-SNARE complex (66, 70) in conjunction with the SM protein Sly1 (71). Interestingly, interactions with the Qb-SNARE Sec20 appear to be mediated by the CATCHR domain of the Tip20 subunit (68). The recruitment of the Qc-SNARE Use1 into SNARE complexes depends on Sec39 and Dsl1 subunits interacting with Tip20 (66). Sly1 interactions with Ufe1 and SNARE complexes are part of a larger complex that includes Dsl1 representing a complex that coordinates Golgi to ER retrograde traffic (38, 71). Overall, the Dsl1 complex links vesicle tethering to SNARE complex assembly and fusion at sites in the ER.

Tethering factors are essential coordinators linking vesicle arrival with SNARE complex assembly. Individual MTC subunits mediate a number of protein and phospholipid interactions. CATCHR domains in MTC subunits that possess them appear to generally mediate interactions with other subunits or with other proteins including SNAREs. One common theme illustrated by tethering factors at several vesicle trafficking stations is the ondemand assembly of Q-SNARE complexes with vesicle arrival. Recent studies (29) suggested that Qc-SNAREs are a particularly important determinant for selectivity in SNARE pairing for fusion. Thus, it is notable that an interaction specific for the Qc-SNARE is found for HOPS (55), and that exocyst and Dsl1 complexes bind Sec9 Qbc- or Use1p Qc- SNAREs, respectively. Overall this suggests an important role for tethering factors in promoting the entry of Qc-SNAREs into SNARE complexes to enable appropriately paired, compartment-specific fusion.

#### **PRIMING FACTORS INTEGRATE SNARE PROTEIN FUNCTION FOR REGULATED FUSION**

The release of neurotransmitters from synaptic vesicles (SVs) at the neuronal synapse or peptides from DCVs of neuroendocrine cells occurs by regulated exocytosis. Pools of SVs and DCVs are stored near the plasma membrane in various states prior to fusion (72). Traditionally, vesicles are viewed as progressing through states of tethering, docking, and priming prior to Ca2+-triggered fusion (73). Recent high-pressure freeze EM (74) and fluorescence microscopy mobility studies of vesicles (75) indicate that docking and priming may be closely linked steps. Physiological and genetic studies suggest that SNARE complex assembly occurs during priming (76, 77). SV exocytosis is strongly inhibited in mice lacking Munc13-1 (78, 79) and CAPS (80). DCV exocytosis is also

strongly impaired in cells lacking Munc13-1 or CAPS (78, 81–83). Thus, major accessory factors for SV and DCV priming are the related CAPS and Munc13 proteins. Munc18-1 is also involved in SV and DCV priming but its role in priming in cells is difficult to separate from its upstream role as a syntaxin-1 chaperone that influences vesicle docking (42, 84). It is likely that CAPS and Munc13 proteins co-function with Munc18-1 in vesicle priming.

CAPS and Munc13 proteins are related in sequence (**Figure 2**) (6, 85). CAPS and Munc13-1 exhibit ∼40% sequence similarity in a C-terminal region that contains the MHD1 homology domain. All Munc13 proteins (including Munc13-4 and BAIAP3) but not CAPS proteins also share a more C-terminal MHD2 domain. Overall the C-terminal region of CAPS and Munc13 proteins exhibits weak sequence homology to the CATCHR domain of exocyst, COG, GARP, and Dsl1 complex subunits (**Figure 2**) (6). More convincingly, crystallographic studies of a Munc13- 1(1148–1531) protein indicated strong structural similarity to the CATCHR region of the Sec6 subunit of the exocyst complex (5). This homology across diverse proteins implies an evolutionary relatedness but could also indicate a conserved functional role for the CATCHR domain. In MTCs, an inherent structural role for CATCHR domains was seen to be adapted to mediate protein–protein interactions that include SNARE-binding (50). Studies of the CAPS and Munc13 proteins indicate a role for this region in scaffolding SNARE proteins as well as for other protein interactions. In the following, we discuss membrane- and SNARE-binding features of CAPS and Munc13 proteins that have counterparts in MTCs.

#### **PRIMING FACTORS FOR REGULATED VESICLE EXOCYTOSIS: CAPS**

CAPS (aka CADPS) was discovered for its activity in regulating DCV exocytosis in neuroendocrine cells (86) and was found to correspond to the *Caenorhabditis elegans* UNC31 protein (87). Unc-31 deletion mutants exhibit a strong loss of DCV exocytosis and neuropeptide secretion (88–90) with moderate reductions in SV exocytosis and synaptic transmission (88, 89, 91, 92), which matches the conditional uncoordinated phenotype. By contrast, the phenotypes for *C. elegans* UNC13 mutants are much more severe. Unc-13 hypomorphs are paralyzed, and exhibit a strong loss of synaptic transmission with lesser impact on DCV exocytosis and neuropeptide secretion (89, 93, 94). Thus, *C. elegans* CAPS/UNC31 is essential for DCV exocytosis whereas Munc13/UNC13 plays a dominant role in SV exocytosis. In vertebrates, the requirements for vesicle priming are more complex in requiring both CAPS and Munc13 proteins. Vertebrates possess two CAPS genes (CAPS/CADPS and CAPS2/CADPS2) that control DCV exocytosis in chromaffin cells, pancreatic β cells, and neurons (83, 95–100). Munc13 proteins are encoded by five genes (Munc13-1, -2, -3, -4, BAIAP3). Munc13-1 is required for DCV exocytosis in pancreatic β cells (101, 102), which indicates that both CAPS and Munc13-1 are required for regulated insulin secretion. CAPS localizes to DCVs but not SVs in brain tissue (103). In spite of this, studies indicate that CAPS-1/2 KO mice exhibit as complete loss of synaptic transmission as reported for Munc13-1 KO mice (78, 80). Thus, it appears that CAPS and Munc13 proteins are both required for SV and DCV priming in vertebrate nervous and endocrine systems.

Recent studies have revealed CAPS to be a regulator of SNARE complex assembly. Attempts to detect direct CAPS interactions with soluble SNARE proteins were of limited success indicating only very low affinity interactions; however, recent studies revealed direct CAPS interactions with membrane-associated SNARE proteins (**Figure 1**, lower) (104–107). Liposomes containing syntaxin-1/SNAP-25 heterodimers or VAMP2 were found to retain CAPS in liposome flotation studies (104). CAPS interacted independently with either syntaxin-1 or SNAP-25 suggesting that CAPS might promote QaQbc-SNARE heterodimer formation. CAPS binding to syntaxin-1 was mediated by the membraneproximal C-terminal SNARE motif (H3) and membrane linker domain sequences of syntaxin-1 (104). CAPS interactions with N-terminal regions of the SNARE motif of VAMP2 were also detected, which suggests that CAPS might recruit VAMP2 into syntaxin-1/SNAP-25 heterodimers for RQaQbc-SNARE complex assembly. As a SNARE-binding protein, CAPS stimulated the formation of SNARE complexes on liposomes (106) and promoted VAMP2 liposome docking on supported bilayer membranes containing syntaxin-1/SNAP-25 heterodimers indicating that stable trans-SNARE complex formation had occurred. These studies utilized pre-formed syntaxin-1/SNAP-25 heterodimers, and indicated that CAPS could promote VAMP2 insertion into QaQbc-SNARE heterodimers to assemble heterotrimeric SNARE complexes (**Figure 1**, lower). The activity of CAPS in promoting SNARE complex formation was also evident in studies

of SNARE-dependent liposome fusion where CAPS markedly increased the rate and extent of fusion between donorVAMP2 liposomes and syntaxin-1/SNAP-25 acceptor liposomes (105, 106). These results also suggested that CAPS acts to promote the insertion of the R-SNARE VAMP2 into Qabc-SNARE syntaxin-1/SNAP-25 acceptors (**Figure 1**, lower) but it is not yet known if CAPS utilizes direct interactions with syntaxin-1/SNAP-25, with VAMP2, or with both to enable SNARE complex assembly. Additional studies are needed to determine the detailed mechanism of how CAPS enhances SNARE complex assembly.

A key issue is which CAPS domains mediate SNARE protein binding and SNARE complex assembly. The C-terminal region of CAPS/Munc13 proteins contains numerous α-helices, which makes standard recombinant protein analysis or yeast two-hybrid interaction studies very challenging (5, 108). We produced a set of recombinant proteins across the CAPS sequence based on secondary structure predictions and proteolysis studies, and tested these proteins for moderate-to-high affinity (0.2µM) SNAREbinding (107). Only one protein fragment corresponding to rat CAPS(859-1073) was retained in flotation studies by syntaxin-1/SNAP-25 liposomes (107). CAPS(859-1073), which brackets the MHD1 homology region (**Figure 2**), exhibited submicromolar binding affinity for SNARE proteins and effectively competed with full-length CAPS(1-1289) for binding indicating that this region contains the major SNARE-binding domain of CAPS. Further studies with protein fragments suggested that the major SNAREbinding segment may consist of a helix in the center of MHD1 that contains a VAMP2 homology region. This corresponds to the Nterminal helix of the CATCHR homology region (**Figure 2**) (5, 6, 107). These studies did not exclude the possibility that other helices within the CATCHR homology domain provide additional lower affinity SNARE-binding. Thus, this region consisting of stacked α-helices could function as a scaffold to organize helical SNARE motifs. A recent report suggested that syntaxin-1 binding by CAPS was mediated by more C-terminal sequences within CATCHR; however, these studies employed constructs in yeast two-hybrid studies that may have encoded unfolded proteins (109). Studies on CAPS are consistent with a SNARE scaffolding role for the CATCHR homology region. However, other CAPS-protein interactions have also been reported for this region (110), which could indicate a more general role for the CATCHR domain as a protein interaction domain.

For MTCs, numerous membrane interactions can be achieved by multiple subunits. As a large multi-domain protein, CAPS may instead utilize multiple domains to mediate protein and lipid interactions. CAPS exhibits low affinity but functionally significant interactions with plasma membrane PIP<sup>2</sup> via its central PH (pleckstrin homology) domain (82, 111). PIP<sup>2</sup> enhanced CAPS stimulation of SNARE-dependent liposome fusion with wild-type but not with mutant PH domain CAPS proteins (105). Inclusion of PIP<sup>2</sup> in the syntaxin-1/SNAP-25-containing acceptor liposomes was much more effective than inclusion in the VAMP2-containing donor membranes, which suggests that PIP<sup>2</sup> is an important cofactor for CAPS in acting on plasma membrane SNARE proteins (**Figure 1**, lower). PIP<sup>2</sup> may promote conformational or oligomerization changes in CAPS to enhance its SNARE interactions (111). In addition, because CAPS interacts with syntaxin-1 near its C-terminal linker that binds PIP2, this might allow CAPS to regulate the conformation of syntaxin-1 (104, 105). These results suggest a framework for understanding the actions of priming factors. CAPS utilizes two contacts with the membrane – one with membrane phospholipids via its PH domain and the other with SNAREs via its MHD1 domain – to promote the assembly of SNARE protein complexes.

CAPS localizes to DCVs (103, 112) and also interacts with plasma membrane PIP2, which could provide a trans-membrane interaction for vesicle tethering. However, such a tethering mechanism would likely be transient because of the low affinity PIP<sup>2</sup> interactions. The basis for CAPS anchoring to DCVs via Cterminal interactions (112) remains to be clarified. Possible interactions with the DCV constituents phogrin (113), VMAT (114), ARF4/5 (115, 116), and RRP17 (110) have been suggested.

Tethering factors are thought to engage in long-range capture of vesicles (tethering) at the target membrane involving distances (>20 nm) at which SNARE complexes cannot assemble. MTC tethering complexes likely bring vesicles into closer proximity to enable SNARE complex formation and docking. CAPS functions in vesicle priming to promote the assembly of SNARE complexes that bridge vesicles to the plasma membrane, which may be expected to mediate vesicle docking. Indeed, *in vitro* studies of VAMP2 liposome docking onto syntaxin-1/SNAP-25-containing membranes revealed that CAPS could promote a stable docking complex (106). EM studies in *C. elegans* also indicated that CAPS/Unc31 was required for DCV docking to the plasma membrane (88). Although DCV docking defects were not reported for chromaffin cells from CAPS KO mice, this may be attributable to the small percentage of total vesicles that are primed in these cells (83).

#### **PRIMING FACTORS FOR REGULATED VESICLE EXOCYTOSIS: Munc13**

Munc13 proteins are thought to function in vesicle priming by interacting with SNARE proteins (44). Munc13-4, a short Munc13 isoform with N- and C-terminal C2 domains (C2A and C2B, respectively) bracketing the MHD1-MHD2 region (**Figure 2**), functions in the priming as well as the maturation of lysosome-related secretory granules for fusion in secretory cells of hematopoietic origin (117). Munc13-4 appears to function as a tether for granule-plasma membrane interactions mediated by vesicle-associated Rab27 (118, 119). Recent biochemical studies revealed that Munc13-4 exhibits Ca2+-regulated SNARE interactions modulated by its C2A domain and Ca2+-dependent membrane interactions mediated by its C2B domain (120). Munc13-4 promoted the fusion of VAMP2 donor liposomes with syntaxin-1/SNAP-25 acceptor liposomes that was dependent on Ca2<sup>+</sup> and Ca2+-binding residues in each C2 domain (120). These results indicated that Ca2+-activated Munc13-4 can function similarly to CAPS by promoting the recruitment of the R-SNARE VAMP2 into Qabc-SNARE syntaxin-1/SNAP-25 acceptors for RQaQbc-SNARE complex assembly. The central MHD1-MHD2 region of Munc13- 4 with CATCHR homology may mediate SNARE-binding but this has yet to be demonstrated. For Munc13-4 as for CAPS, it appeared that anchoring the protein to the membrane (via the Ca2+-dependent C2B domain) coupled to SNARE-binding was required to promote SNARE complex assembly and liposome fusion (120).

C-terminal regions of Munc13-1 were reported to interact with N-terminal domains of syntaxin-1 in yeast two-hybrid interaction studies (121, 122). Solution binding studies with recombinant Munc13-1 protein fragments localized N-terminal syntaxin-1 binding to Munc13-1(1181–1736), which corresponds to sequences beginning in MHD1 (**Figure 2**). This interaction was proposed to counteract Munc18-1-mediated stabilization of a "closed" form of syntaxin-1 to "open" it to enable syntaxin-1/SNAP-25 heterodimer formation. However, subsequent studies reported that Munc13-1(859–1531) (termed MUN domain) failed to interact with SNARE proteins in solution but did interact with liposome-integrated SNARE protein complexes (108, 123, 124). Munc13-1(859–1531) bound to syntaxin-1/SNAP-25 or syntaxin-1/SNAP-25/VAMP2 liposomes but not to syntaxin-1 liposomes, which suggested that Munc13-1 stabilizes SNARE complexes (123). Consistent with this, other studies provided evidence that Munc13-1(859–1531) stabilized parallel conformations of syntaxin-1/SNAP-25 heterodimers (124). Recent NMR studies with Munc13-1(859–1531) and soluble SNARE proteins indicated that Munc13-1(859–1531) interacts very weakly with the C-terminal SNARE domain of syntaxin-1 and with Munc18-1-bound syntaxin-1 (125). By contrast, a structurally defined Munc13-1(1148–1531) protein fragment exhibited further attenuated SNARE protein interactions possibly because this fragment lacked more N-terminal sequences (5).

In recent studies, a Munc13-1(529–1531) protein that contained C1 and C2B domains (**Figure 2**) was shown to operate on liposomal syntaxin-1-Munc18-1 to enable Ca2+-bound synaptotagmin C2AB to promote SNARE-dependent liposome fusion (126). It was proposed that the Munc13-1 fragment catalyzed a transition of Munc18-1-bound syntaxin-1 into syntaxin-1/SNAP-25 heterodimers to serve as acceptor complexes for VAMP2. This model for Munc13-1 action retains the feature of syntaxin-1 "opening" but proposes that the Munc13-1 C-terminal domain interacts with C-terminal rather than N-terminal regions of syntaxin-1 (126). In addition, these studies with a Munc13-1 fragment plus Munc18-1 suggested a possible exchange of the Qbc-SNARE SNAP-25 into Q-SNARE complexes as an important regulated step (126).

These studies suggest a mode of action for Munc13-1 in promoting the transition of closed syntaxin-1 monomers to syntaxin-1/SNAP-25 heterodimers. Similar effects for Munc13-4 and CAPS have not yet been demonstrated. Actual differences in the mechanisms of Munc13-1 and CAPS action on SNAREs could help to account for the non-redundancy of these factors for vesicle priming. However, future study will be needed to assess whether these apparent differences result from the use of different assays for CAPS and Munc13-1 proteins. Genetic studies in *C. elegans* have found that expression of an"open"syntaxin mutant by-passes DCV docking defects in CAPS/Unc-31 mutants (88) and SV docking defects in Unc-13 mutants (127) possibly indicating that both proteins enable a transition of syntaxin-1/Munc18-1 complexes to syntaxin-1/SNAP-25 complexes. Overall, the studies on Munc13- 1 are compatible with a SNARE scaffolding role for the CATCHR homology region but the detailed mechanics of SNARE-binding remain to be worked out.

Munc13-1/2 proteins are also reported to exhibit Ca2+ dependent, high affinity PIP<sup>2</sup> interactions via a central C2B domain (128). This interaction was significant for SV exocytosis in response to high frequency stimulation rather than for responses to single action potentials. The importance of C2Bmediated phosphoinositide interactions for DCV exocytosis in neuroendocrine cells has not been determined but could play a role in recruiting cytosolic Munc13-1/2 to sites of DCV exocytosis. The adjacent DAG-binding C1 domain of Munc13-1 mediates the membrane recruitment of Munc13-1 in response to DAG, however, a functional C1 domain in Munc13-1 is not required for Ca2+-stimulated vesicle exocytosis but rather for potentiated responses (129).

In contrast to the cytoplasmic localization of Munc13-1/2 in neuroendocrine cells, Munc13-1 localizes to the active zone in synapses where it associates with at least four other active zone proteins (RIM, bassoon, aczonin/piccolo, and CAST) via its Nterminal domain (130). This molecular complex likely serves a tethering role mediated by proteins anchored to both the presynaptic membrane and to SVs (130). Interactions with SVs may be mediated by a complex of RIM and Munc13-1 with Rab3 on the vesicle (131). Studies suggest that RIM activates Munc13-1 by converting it from an inactive dimer to active monomer (42). Standard chemical fixation methods had failed to reveal decreased SV docking in neurons from Munc13-1 KO mice, but high-pressure freezing techniques with EM tomography indicated that SVs were tethered but not docked in the absence of Munc13-1 (74). Similarly, high-pressure freezing followed by low-temperature fixation in *C. elegans* revealed a requirement for UNC13 in SV docking (88). Although the studies might suggest that priming and docking are functional and morphological aspects of the same process, more studies are needed in vertebrate neurons and endocrine cells where CAPS and Munc13 are co-required for vesicle priming.

#### **SUMMARY**

At multiple trafficking stations in secretory and endosomal pathways, diverse tethering and priming factors integrate multiple protein and lipid interactions to achieve compartment-specific SNARE complex assembly for fusion. The MTCs promote SNARE complex assembly by direct interactions of MTC subunits with Q-SNAREs and collaborative interactions with SM proteins. A subset of MTC subunits utilize structurally similar CATCHR domains to mediate inter-subunit interactions as well as SNARE protein interactions. At sites of regulated vesicle exocytosis in neurons and endocrine cells, homologous CAPS and Munc13 proteins play a similar role in mediating SNARE complex assembly for vesicle priming, however they may do so by distinct mechanisms. SNAREbinding in the CAPS and Munc13 proteins appears to reside within the CATCHR domain, which may also mediate additional protein interactions. CAPS and Munc13-1 collaborate with the SM protein Munc18-1 but the details of integration remain to be worked out. Studies are needed to determine whether accessory factors and SM proteins operate in concert or sequentially to assemble SNARE complexes, and to determine how these interactions occur within the confined space of juxtaposed membranes.

#### **ACKNOWLEDGMENTS**

The Martin lab is supported by USPHS grants DK040428 and DK025861.

#### **REFERENCES**


*C. elegans* neurons. *Neuron* (2007) **56**:657–69. doi:10.1016/j.neuron.2007.09. 015


**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: 03 September 2013; accepted: 18 November 2013; published online: 04 December 2013.*

*Citation: James DJ and Martin TFJ (2013) CAPS and Munc13: CATCHRs that SNARE vesicles. Front. Endocrinol. 4:187. doi: 10.3389/fendo.2013.00187*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

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

# The functional significance of synaptotagmin diversity in neuroendocrine secretion

#### **Paanteha K. Moghadam and Meyer B. Jackson\***

Department of Neuroscience, University of Wisconsin, Madison, WI, USA

#### **Edited by:**

Stephane Gasman, Centre national de la recherche scientifique, France

#### **Reviewed by:**

Ling-Gang Wu, National Institute of Neurological Disorders and Stroke (NINDS), USA Ralf Schneggenburger, Ecole Polytechnique Fédérale de Lausanne, Switzerland

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

Meyer B. Jackson, Department of Neuroscience, University of Wisconsin, 1300 University Avenue, Madison, WI 53706-1510, USA e-mail: mbjackso@wisc.edu

Synaptotagmins (syts) are abundant, evolutionarily conserved integral membrane proteins that play essential roles in regulated exocytosis in nervous and endocrine systems. There are at least 17 syt isoforms in mammals, all with tandem C-terminal C2 domains with highly variable capacities for Ca2<sup>+</sup> binding. Many syts play roles in neurotransmitter release or hormone secretion or both, and a growing body of work supports a role for some syts as Ca2<sup>+</sup> sensors of exocytosis. Work in many types of endocrine cells has documented the presence of a number of syt isoforms on dense-core vesicles containing various hormones. Syts can influence the kinetics of exocytotic fusion pores and the choice of release mode between kiss-and-run and full-fusion.Vesicles harboring different syt isoforms can preferentially undergo distinct modes of exocytosis with different forms of stimulation.The diverse properties of syt isoforms enable these proteins to shape Ca2<sup>+</sup> sensing in endocrine cells, thus contributing to the regulation of hormone release and the organization of complex endocrine functions.

**Keywords: exocytosis, neuropeptides, dense-core vesicle, norepinephrine, insulin, calcium, fusion pores, kiss-andrun**

#### **INTRODUCTION**

Nature employs the same basic molecular machinery for the release of both hormones and neurotransmitters (1–4). Several protein families function broadly in regulated exocytosis, including SNAREs, synaptotagmins (syts), and complexins (5–7). The rich molecular diversity within these families provides a platform for variations in the release process, enabling different cell types to tune and tailor release by blending the different molecular variants of the fusion apparatus. In this way endocrine cells can optimize secretory responses elicited by widely varying signals that are unique to each system. The rates of hormone release from different cell types vary by over two orders of magnitude (8). Endocrine cells secrete an extraordinary variety of hormones by exocytosis from dense-core vesicles (DCVs), and DCVs usually co-package collections of molecules ranging in size from small catecholamines to large peptides (9, 10). The nature of the stimulus can determine which packaged molecules will be released (11, 12), and a single DCV can release both catecholamines and neuropeptides simultaneously (13). Cells also can sort different hormones to different DCVs (14, 15). The diverse forms of storage and release raise questions as to how the exocytotic machinery can be called upon to modulate release rates and enable different types of Ca2<sup>+</sup> signals to trigger the release of different substances from the same cell or even the same vesicle. One can hope to gain a better understanding of these problems by studying functional variations within the diverse families of exocytotic proteins.

Exocytosis proceeds through a sequence of distinct steps. Release can start once a fusion pore has formed to create a continuous fluid pathway from the vesicle interior to the extracellular space. The fusion pore is initially very small and can allow only small molecules such as norepinephrine to pass, but after it expands larger molecules such as chromogranins, insulin, and glucagon can escape. It is well established that DCVs of endocrine cells undergo two modes of exocytosis, kiss-and-run, and fullfusion (16–19). In kiss-and-run the pore opens transiently, and closes so that vesicles maintain their integrity as they retreat from the plasma membrane. The fusion pore formed during kiss-andrun can act as a filter to expel small molecules and retain larger molecules. The small molecules can be rapidly restored by vesicular transporters so DCVs can recycle. By contrast, in full-fusion the pore expands and the vesicle membrane collapses into the plasma membrane. After full-fusion a DCV is lost; replacing these DCVs requires the entire production sequence beginning with peptide hormone translation in the endoplasmic reticulum, DCV processing in the trans-Golgi network, and maturation (20). The choice between full-fusion and kiss-and-run thus plays a decisive role in determining the fate of the DCV as well as in selecting what molecules are released.

Ca2<sup>+</sup> triggers exocytosis by binding to syt (21–23), but Ca2<sup>+</sup> can also influence the kinetics of fusion pores in a variety of ways (24–26), thus raising the possibility that syt isoforms play roles not only in triggering exocytosis, but also in regulating release in subtle ways. Syts regulate the stability of fusion pores (26–30), and fusion pore stability in turn is intimately related to release mode, with the relative frequency of kiss-and-run versus fullfusion described quantitatively in terms of the kinetic rates of fusion pore closure and dilation (26). There are many examples of syts influencing the choice between kiss-and-run and full-fusion (26, 29, 31–33). Syts are conserved proteins with well established functions in membrane trafficking and exocytosis. They contain two C2 domains, and in syt 1, the first isoform to be characterized, each C2 domain binds two or three Ca2<sup>+</sup> ions through

interactions with key aspartate side chains (34). Syts interact with lipid membranes containing specific phospholipids including phosphatidylserine and phosphatidylinositol-4,5-biphosphate, as well as SNARE proteins. These interactions are regulated by Ca2+, but it remains unclear how binding to these targets enables syts to serve as Ca2<sup>+</sup> sensors in exocytosis (21, 22, 30).

The mouse and human genomes encode 17 syt isoforms (35), and the functional significance of syt diversity is a subject of considerable interest (22, 36, 37). Syts vary widely in their Ca2+ dependent lipid binding, with syts 1, 2, and 3 binding rapidly, syts 5, 6, 9, and 10 binding at an intermediate rate, and syt 7 binding slowly (38). The isoforms also vary in their activity in Ca2+-stimulated liposome fusion: Ca2<sup>+</sup> concentrations that trigger responses range widely between the isoforms, and many syts completely fail to confer Ca2<sup>+</sup> sensitivity on SNARE-mediated liposome fusion *in vitro* (32, 39, 40). Variations in Ca2<sup>+</sup> sensitivity and fusion pore regulation make these proteins ideal candidates for modifying the release apparatus and tuning responses as Ca2<sup>+</sup> concentrations rise and fall in distinct spatiotemporal patterns. Here, we survey relevant work on syt functions in hormone release, and where possible draw parallels between syt isoform properties, Ca2<sup>+</sup> signals, and forms of secretion.

#### **SYT FUNCTION IN ENDOCRINE CELLS**

Synaptotagmins appear broadly throughout the endocrine system, with essentially every cell type examined expressing multiple isoforms (**Table 1**). Expression varies between cell types and not all reports agree. No effort was made here to distinguish between isoforms untested versus undetected, and the number of isoforms found in endocrine cells will grow as reagents are developed and improved, and as proteomics methods advance. A large body of work supports the expression of syts 1, 4, 7, and 9 in many endocrine systems, and it is remarkable that these four molecules appear in so many different cell types. Syts 1, 4, and 7 are ancient, conserved proteins distributed widely through metazoan genomes (35). These isoforms presumably perform fundamental biological functions, and evidence is accumulating for their roles in a wide range of endocrine and non-endocrine systems.

#### **SYT 1**

Syt 1 is the most widely distributed syt isoform in nervous and endocrine systems. This low-affinity Ca2<sup>+</sup> sensor (32, 39, 41, 42) generally triggers rapid exocytosis. The very tight temporal coupling between Ca2<sup>+</sup> entry and fusion, within milliseconds, has prompted investigators to use the term "synchronous" to describe this form of release, particularly in the context of synaptic transmission. An early syt 1 knock-down study suggested that PC12 cells can secrete without syt 1 (43), but subsequent work showed that this was due to redundancy with another syt isoform of PC12 cells, syt 9 (33, 44). Overexpression of wild type syt 1 in PC12 cells left the overall time course of secretion unchanged, but overexpression of either wild type syt 1 or a number of syt 1 mutants altered fusion pore kinetics (26–28, 30, 45). Overexpressing syt 1 in PC12 cells also produced more kiss-and-run events than syt 7 and 9 (32).

Deletion of the syt 1 gene in mouse selectively abolished the initial rapid phase of exocytosis in chromaffin cells (33, 46, 47), but had no deleterious effects on slower Ca2+-triggered release. For a

#### **Table 1 | Syt isoform expression in various endocrine systems.**


In most cases the localization and expression was based on immunocytochemistry (see text).

<sup>a</sup>β-Cell lines include RINm5F, INS1, MN6, HT-T15, TC6-F7.

given concentration of Ca2+, exocytosis was much slower in chromaffin cells lacking syt 1 than in control cells (46). Mutation of a residue that reduces Ca2<sup>+</sup> binding slowed exocytosis in chromaffin cells (48). In PC12 cells syt 1 sorted preferentially to smaller DCVs (32), raising the interesting possibility that hormones packaged in smaller vesicles will be released more rapidly than hormones packaged in larger vesicles.

#### **SYT 7**

In contrast to syt 1, syt 7 acts as a high affinity Ca2<sup>+</sup> sensor (32, 39, 49). Syt 7 is more abundant on larger DCVs in PC12 cells (32).

Syt 7 overexpression in PC12 cells prolonged fusion pore lifetimes more than syt 1 overexpression (30), and favored full-fusion (32). Syt 7 knock-down in zebra fish reduced delayed synaptic release, indicating it is a slow Ca2<sup>+</sup> sensor (49). Syt 7 deletion in chromaffin cells reduced Ca2+-triggered release by 50%, also selectively impairing the slow phase of exocytosis, and deletion of both syt 1 and 7 nearly abolished Ca2+-triggered exocytosis. Thus, in chromaffin cells syt 1 and 7 are the primary Ca2<sup>+</sup> sensors for the fast and slow kinetic phases of exocytosis, respectively (47). Syt 7 gene ablation also reduced Ca2+-triggered exocytosis of insulin secretion from pancreatic β-cells and of glucagon secretion from pancreatic α-cells (50, 51). It is intriguing that syt 7 also functions in insulin responsive cells (fat and muscle), promoting glucose uptake through Ca2+-stimulated translocation of type-4 glucose transporter to the plasma membrane (52). In syt 7 knockout α cells, ω-conotoxin further inhibited glucagon secretion to baseline levels, revealing the presence of an N-type Ca2<sup>+</sup> channeldependent component of residual glucagon secretion triggered by another protein (51).

#### **SYT 9**

Syt 9 is closely related to syt 1 but exhibits intermediate Ca2<sup>+</sup> sensitivity in fusion assays (32, 39). This protein has also been referred to as syt 5 (35, 53); here syt 9 refers to a 386 amino acid isoform. It is abundant on DCVs of PC12 cells (32, 54, 55), and overexpressing it produces fusion pore lifetimes intermediate between those seen with syt 1 and 7 (30). Down-regulating syt 9 alone produced a small, insignificant reduction of fusion rate in PC12 cells, but as noted above, because of the redundancy of syt 1 and 9 as Ca2<sup>+</sup> sensors, both must be down-regulated to reduce secretion (33, 44). Silencing of syt 9 strongly inhibited insulin release from islet βcells and INS-1E cells (56). However, mice with a pancreas-specific knock-out of syt 9 had normal glucose homeostasis and showed no changes in other insulin-dependent functions (57).

#### **CA2**<sup>+</sup> **NON-BINDERS**

Slightly more than half of the mammalian syts have non-acidic amino acids at some of the positions engaged in Ca2<sup>+</sup> binding (35), and these isoforms fail to act as Ca2<sup>+</sup> sensors in liposome fusion assays (40). The best characterized of these, syt 4, harbors an evolutionarily conserved serine-for-aspartate substitution at a Ca2<sup>+</sup> ligand in the C2A domain (58). Syt 4 is widely expressed in endocrine cells and its expression rises and falls depending on electrical activity (59) and reproductive state (60). Syt 4 negatively regulates both release and Ca2+-dependent liposome fusion (28, 32, 40) and does not bind Ca2<sup>+</sup> (61). Syt 4 overexpression reduced exocytosis in PC12 cells (28, 32, 62). Although syt 4 overexpression in PC12 cells shortened the lifetimes of fusion pores capable of dilating to full-fusion (28), another form of release was enhanced in which very small fusion pores could open exclusively as kiss-and-run events with exceptionally long lifetimes. Many but not all of the effects of syt 4 overexpression were mimicked by syt 1 harboring the serine-for-alanine Ca2<sup>+</sup> ligand replacement seen in syt 4 (29). Syt 4 overexpression also favored kiss-and-run in MIN6 β-cells (31). Ablation of the syt 4 gene altered exocytosis and fusion pore properties in posterior pituitary nerve terminals. These results suggested that syt 4 reduced exocytosis in response to modest Ca2<sup>+</sup> rises but enhanced exocytosis in response to large Ca2<sup>+</sup> rises (63). Syt 4 also contributes to the maturation of DCVs (64, 65), and altering syt 4 levels changes DCV size (62, 63, 66).

The Ca2<sup>+</sup> non-binding isoforms syt 4, 8, and 13 have been detected in insulin-secreting cells (67–69). Silencing syt 4 and 13 reduced glucose stimulated insulin secretion in INS1 cells (69). Glucose stimulated expression of the syt 8 gene in human islets and syt 8 knock-down impaired both basal and evoked insulin release (70). Since syt 8 fails to stimulate liposome fusion in a Ca2+-dependent manner (40) the precise role of syt 8 in insulin secretion remains unclear. One intriguing possibility is that syt 8 increases the relative proportion of full-fusion events so that fusion pores can grow large enough to allow insulin to escape. It is likely that the Ca2<sup>+</sup> non-binders interact with other components of the release machinery in ways that remain to be elucidated. These interactions could allow syts to regulate release in ways that cannot as yet be explained in terms of their biochemical properties.

Reports vary regarding the expression of other syt isoforms in endocrine cells (**Table 1**) and little is known about their localization and functions in hormone release. Syt 2 has been reported in the hypothalamus (71) but is absent from endocrine cells, and its primary function is likely to be as a synaptic Ca2<sup>+</sup> sensor (72, 73).

#### **SPECIFICITY OF CA<sup>2</sup>**<sup>+</sup> **SIGNALING**

The differences in performance of syts as Ca2<sup>+</sup> sensors gives a special significance to the spatiotemporal character of Ca2<sup>+</sup> signals in different cell types and with different forms of stimuli. Ca2<sup>+</sup> can enter cells through a variety of routes so that the dynamics and spatial extent of changes in cytosolic Ca2<sup>+</sup> can vary enormously. This creates a scenario in which differences in Ca2<sup>+</sup> sensing properties can have a major impact on responses (74). Intracellular Ca2<sup>+</sup> rises and falls as Ca2<sup>+</sup> enters through Ca2<sup>+</sup> channels, diffuses away from these sources into the cytoplasm, binds Ca2<sup>+</sup> binding proteins, and is sequestered into stores or pumped out of the cytoplasm (75–77). The opening of one voltage gated Ca2<sup>+</sup> channel allows approximately 10<sup>3</sup> ions to enter per msec. This localized flux sets up a steep gradient to create a domain of high local Ca2<sup>+</sup> concentration. In the immediate vicinity of the channel mouth (within tens of nanometer) Ca2<sup>+</sup> can rise to >100µM, which is substantially higher than the average bulk cytoplasmic level, even under conditions of intense stimulation (75, 78). These Ca2<sup>+</sup> domains around a Ca2<sup>+</sup> channel can form and collapse rapidly (within a few ms) so that the activation of a Ca2<sup>+</sup> sensor will depend critically on the kinetics of Ca2<sup>+</sup> association and dissociation. A low-affinity sensor can detect this high local Ca2<sup>+</sup> as long as it binds with rapid kinetics. A high affinity Ca2<sup>+</sup> sensor may fail to respond if it binds too slowly. Syt 1 has properties suited for responding to transient domains with high Ca2+. Syt 7 has properties suited for responding to modest but prolonged rises in bulk Ca2<sup>+</sup> concentration (38).

The idea of localized Ca2<sup>+</sup> signals leads to two types of Ca2<sup>+</sup> concentration profile illustrated in **Figure 1**. When an endocrine cell fires at a moderate rate, Ca2<sup>+</sup> domains form and collapse around a few open Ca2<sup>+</sup> channels as shown in **Figure 1A**. The resulting brief period of high Ca2<sup>+</sup> concentration will only affect vesicles near the open channels, and will preferentially activate syt 1 over slower isoforms. Basal firing rates corresponding to the resting state of an animal (feed and breed) have been shown

to trigger norepinephrine release from chromaffin cells without triggering chromogranin release (11). This implicates exocytosis by kiss-and-run, which is preferentially triggered by syt 1 (32). By contrast, with vigorous electrical activity during stress (fight and flight) the bulk Ca2<sup>+</sup> concentration will rise to moderate levels (well under 50µM) for longer times on the order of seconds (**Figure 1B**). These signals will activate slow, low-affinity Ca2<sup>+</sup> sensors such as syt 7. This form of [Ca2+] signal has been shown to trigger full-fusion (79), with release of both norepinephrine and chromogranin (11). This can be explained by the tendency of syt 7 to trigger full-fusion preferentially (32). Different spatiotemporal patterns of cytosolic Ca2<sup>+</sup> can thus target different syt isoforms to modulate fusion kinetics. Furthermore, because syt isoforms favor different modes of exocytosis, by selectively promoting kissand-run or full-fusion, Ca2<sup>+</sup> signals that activate different syts will determine the relative release of small and large molecules. The sorting of syts to different sized vesicles may also serve as a mechanism for allowing different Ca2<sup>+</sup> signals to select different substances for release if content is also found to vary with vesicle size (32).

#### **CONCLUSION**

Studies of how different syt isoforms influence the kinetics of exocytosis are starting to resolve functionally relevant distinctions in the mechanisms by which Ca2<sup>+</sup> regulates exocytosis. More studies like these will expand our understanding of the diversity of endocrine release mechanisms, but rigorous assessments of syt isoform function in cells face a number of challenges. (1) It is important to address the expression of multiple isoforms, either by ablating endogenous proteins or varying each isoform individually with careful assessment of protein levels. Overexpression of proteins can result in high protein concentrations, possibly leading to mis-targeting and artificial functions. (2) Experiments need to address the different Ca2<sup>+</sup> sensitivities of syt isoforms. This requires measurement and control of Ca2<sup>+</sup> concentration. (3) Biophysical techniques for measuring exocytosis are very sensitive but often measure surrogates of release and detect events on different time scales. Amperometry measures release and can detect rapid processes but its greatest sensitivity is realized with biogenic amines, and is not nearly as powerful in the measurement of peptides. Furthermore, after content expulsion the closure of a fusion pore can no longer be detected, and the mode of release as kiss-and-run or full-fusion can no longer be ascertained. Capacitance measures membrane area, and single-vesicle capacitance steps provide strong evidence for kiss-and-run. Total internal reflectance microscopy of fluorescent tracers follows the fate of the vesicle content or membrane (80). Like capacitance recording, this method can reveal kiss-and-run but its time resolution is poor. Furthermore, the fate of fluorescent cargo can vary with the design of the fusion construct, raising concerns about experimental artifacts (81). Amperometry, capacitance, and total internal reflectance microscopy are very powerful techniques but they do not always agree and these differences can complicate interpretations and comparisons (62).

Given the functional versatility within the syt protein family, the studies to date probably have only scratched the surface in addressing the roles of syt isoforms in the complex and varied forms of hormone release. By selecting among the different syt isoforms and regulating their expression, and by sorting to different vesicles containing different hormones, cells can regulate the response to diverse forms of Ca2<sup>+</sup> signals. The syt isoforms present on a vesicle will determine what form of Ca2<sup>+</sup> signal will trigger fusion, how different Ca2<sup>+</sup> signals direct the choice between kiss-and-run and full-fusion, and what proportion of small and large molecules are released.

#### **ACKNOWLEDGMENTS**

This research was supported by NIH grants NS044057 and NS072905. We thank Drs. Xuelin Lou and Edwin Chapman for helpful discussions and comments.

#### **REFERENCES**


*Cell Metab* (2006) **4**:283–90. doi:10.1016/j.cmet.2006.08.011


kiss-and-run mechanism. *Nat Cell Biol* (1999) **1**:40–4. doi:10.1038/ 9012


bind to the tip of a C2-domain? *EMBO J* (1998) **17**:3921–30. doi:10. 1093/emboj/17.14.3921


*Neurosci* (2009) **12**:163–71. doi:10. 1038/nn.2252


*Chem* (2002) **277**:24499–505. doi:10.1074/jbc.M202767200


et al. Synaptotagmin III isoform is compartmentalized in pancreatic beta-cells and has a functional role in exocytosis. *Diabetes* (2000) **49**:383–91. doi:10.2337/diabetes.49. 3.383

91. Lang J, Fukuda M, Zhang H, Mikoshiba K, Wollheim CB. The first C2 domain of synaptotagmin is required for exocytosis of insulin from pancreatic beta-cells: action of synaptotagmin at low micromolar calcium. *EMBO J* (1997) **16**:5837–46. doi:10.1093/emboj/16. 19.5837

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

*Received: 23 June 2013; accepted: 31 August 2013; published online: 18 September 2013.*

*Citation: Moghadam PK and Jackson MB (2013) The functional significance of synaptotagmin diversity in neuroendocrine secretion. Front. Endocrinol. 4:124. doi: 10.3389/fendo.2013.00124*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Moghadam and Jackson. 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.*

### Lipids in regulated exocytosis: what are they doing?

#### **Mohamed Raafet Ammar, Nawal Kassas, Sylvette Chasserot-Golaz, Marie-France Bader and Nicolas Vitale\***

Institut des Neurosciences Cellulaires et Intégratives (INCI), UPR-3212 Centre National de la Recherche Scientifique, Université de Strasbourg, Strasbourg, France

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Thomas F. J. Martin, University of Wisconsin, USA Frederic A. Meunier, University of Queensland, Australia

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

Nicolas Vitale, Institut des Neurosciences Cellulaires et Intégratives (INCI), UPR-3212 Centre National de la Recherche Scientifique, Université de Strasbourg, 5 rue Blaise Pascal, 67084 Strasbourg, France e-mail: vitalen@inci-cnrs.unistra.fr

The regulated secretory pathway in neuroendocrine cells ends with the release of hormones and neurotransmitters following a rise in cytosolic calcium. This process known as regulated exocytosis involves the assembly of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins, the synaptic vesicle VAMP (synaptobrevin), and the plasma membrane proteins syntaxin and SNAP-25. Although there is much evidence suggesting that SNARE proteins play a key role in the fusion machinery, other cellular elements regulating the kinetics, the extent of fusion, and the preparation of vesicle for release have received less attention. Among those factors, lipids have also been proposed to play important functions both at the level of secretory vesicle recruitment and late membrane fusion steps. Here, we will review the latest evidence supporting the concept of the fusogenic activity of lipids, and also discuss how this may be achieved. These possibilities include the recruitment and sequestration of the components of the exocytotic machinery, regulation of protein function, and direct effects on membrane topology.

**Keywords: cholesterol, phosphatidic acids, phosphoinositides, exocytosis, membrane fusion, chromaffin cell**

#### **INTRODUCTION: A ROLE FOR LIPID IN EXOCYTOSIS?**

Release of hormones from neuroendocrine cells occurs through exocytosis, a process in which specialized vesicles (secretory granules) fuse with the plasma membrane in response to extrinsic stimuli leading to elevated cytosolic calcium. Chromaffin cells have proven to be a valuable model for unraveling the molecular mechanisms underlying this fundamental membrane fusion event (1). In these cells, secretory granules are first transported from the Golgi area to the cell periphery where they are then docked at the plasma membrane in two apparent stages: a nonprimed (fusion incompetent) and a primed (fusion competent) state. Details of the molecular machinery underlying some of these steps have been described (2). For instance, once tethered to the plasma membrane, docking of granules is mediated by soluble *N*-ethylmaleimide-sensitive factor (NSF) attachment protein receptors (SNAREs) found on granules (v-SNAREs) and on the plasma membrane (t-SNAREs). Through coiled-coil interactions, these proteins form a stable complex that provides the energy necessary to pull membranes into close proximity rendering them fusion competent (3). The mechanism of membrane fusion *per se* is however an aspect that continues to be debated. SNAREs are able to drive membrane fusion *in vitro* (4), but with relatively slow kinetics suggesting that additional factors are required for physiological membrane fusion. Recently, the SNARE accessory protein Munc18 was shown to increase the speed of SNAREmediated fusion (5). Lipids, the prime constituents of the fusing membranes, are also obvious candidates to accelerate the fusion process. The first lipid firmly demonstrated to be critical for exocytosis was phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P2], for review, see Ref. (6), and more recently fatty acids have been suggested to play an important function in neurotransmitter release from synaptic vesicles (7). Growing evidence also support a role for cholesterol and phosphatidic acid (PA) during exocytosis, for

review, see Ref. (8). Hence, many observations are in agreement with the notion that lipids contribute at different levels to exocytosis. The issues now arising concern the precise step in the dynamic exocytotic process in which a given lipid actually functions.

#### **LIPIDS ARE INVOLVED THROUGHOUT THE SEQUENTIAL STAGES UNDERLYING EXOCYTOSIS**

#### **CHOLESTEROL AND PHOSPHATIDYLSERINE FOR DEFINING EXOCYTOSIS SITES**

Cholesterol and sphingolipids cluster into discrete microdomains in cellular membranes to form lipid ordered domains. This clustering is triggered by secretagogues at the level of exocytotic sites in chromaffin cells (9). Although cholesterol depletion by methyl-β-cyclodextrin provided the initial evidence for a positive role of cholesterol in exocytosis (10), these experiments are subjected to caution because of the brought structural role of cholesterol in most cellular functions. More compelling evidence supporting a direct role for cholesterol relies on both biochemical and high-resolution imaging observations indicating that SNAREs concentrate in cholesterol-dependent clusters (11). Additionally, a cholesterol-sequestering agent, the polyene antibiotic filipin, that is supposed to have less dramatic effects on membrane structures than methyl-β-cyclodextrin, induces a dose-dependent inhibition of catecholamine secretion and the release from the plasma membrane of annexin A2 which participates in the formation and/or stabilization of GM1-PtdIns(4,5)P<sup>2</sup> enriched domains required for granule recruitment (9, 12). Altogether these findings support the notion that cholesterol is involved in the spatial definition of exocytotic sites.

In the plasma membrane, phosphatidylserine (PS) mostly resides in the inner cytoplasmic leaflet. In non-apoptotic cells, several biological functions are accompanied by a disruption of this phospholipid asymmetry resulting in the externalization of

PS on the outer leaflet of the plasma membrane. This is the case for calcium-regulated exocytosis in neuroendocrine chromaffin and PC12 cells (13, 14). However the functional importance of PS scrambling for secretion is still under debate and the precise kinetics of this translocation is not established. An interesting possibility lies in the fact that PS contributes substantially to the negative charge of the inner leaflet of the plasma membrane. Consequently, PS scrambling at exocytotic sites could modify the protein/lipid interactions occurring during either the course of exocytosis or the early phases of endocytosis, as suggested recently (15).

#### **PHOSPHOINOSITIDES FOR PRIMING SECRETORY VESICLES**

Phosphoinositides are a class of phospholipids characterized by an inositol head group that can be phosphorylated on the three, four, and five positions to generate seven distinct species key in cell signaling and trafficking. Much of the work carried out on exocytosis has focused on the role played by PtdIns(4,5)P2. Indeed a number of pioneer studies indicated that PtdIns(4,5)P<sup>2</sup> positively modulates secretion in neuroendocrine cells (16–18). Using patch clamp experiments on intact chromaffin cells and in parallel analyzing images of plasma membrane lawns, it was subsequently shown that over-expression of the kinase that generates PtdIns(4,5)P<sup>2</sup> causes an increase in the plasmalemmal PtdIns(4,5)P<sup>2</sup> level and secretion, whereas over-expression of a membrane-tagged PtdIns(4,5)P<sup>2</sup> phosphatase eliminates plasmalemmal PtdIns(4,5)P<sup>2</sup> and inhibits secretion (19). Thus, the balance between the generation and degradation rates of the plasmalemmal PtdIns(4,5)P<sup>2</sup> directly regulates the extent of exocytosis from chromaffin cell. Using the phosphatidylinositol 3-kinase inhibitor LY294002, a correlation between the level of the plasma membrane PtdIns(4,5)P<sup>2</sup> and the size of the primed vesicle pool was found (19, 20). Wen et al. (21) further demonstrated that selective inhibition of phosphatidylinositol 3-kinase delta isoform was responsible for this effect. Importantly, such an inhibition promotes a transient rise in PtdIns(4,5)P2 that was sufficient to mobilize secretory vesicles to the plasma membrane via activation of the small GTPase Cdc42 and actin polymerization. More recently, a functional link between PtdIns(4,5)P<sup>2</sup> signaling and secretory vesicle dynamics through *de novo* remodeling of the actin cytoskeleton was also described (22). These observations are consistent with a function of PtdIns(4,5)P<sup>2</sup> as an acute regulator of secretion. PtdIns(4,5)P<sup>2</sup> seems to lie in a key position controlling the size and refilling rate of the primed vesicle pools, but not the fusion rate constants *per se*. In line with this model, we recently reported that the HIV PtdIns(4,5)P2-binding protein Tat is able to penetrate neuroendocrine cells and accumulate at the plasma membrane through its binding to PtdIns(4,5)P2. By sequestering plasma membrane PtdIns(4,5)P2, Tat alters neurosecretion, reducing the number of exocytotic events without significantly affecting kinetic parameters (fusion pore opening, dilatation, and closure) of individual events (23).

Other phosphoinositides seem to act as signaling or recruitment factors to prime secretory vesicles for exocytosis. For instance, experiments carried out on permeabilized chromaffin cells reveal that PtdIns(3)P located on a subpopulation of chromaffin granules positively regulates secretion (21, 24). Extending these observations, PIKfyve kinase that produces PtdIns(3,5)P<sup>2</sup> from PtdIns(3)P on secretory granules was shown to negatively affect exocytosis (25). Hence, these studies highlight a complex regulation of neuroexocytosis by phosphoinositides, with PtdIns(4,5)P<sup>2</sup> and PtdIns(3)P being essential factors promoting ATP-dependent priming in neurosecretory cells. It is intriguing that PtdIns(3,5)P<sup>2</sup> displays an opposite effect, but reveals how fine-tuning of exocytosis by phosphoinositides could potentially control the number of vesicles undergoing priming in response to a stimulation.

#### **PHOSPHATIDIC ACID FOR FUSION**

The local formation of PA is a recurring theme in intracellular membrane traffic and its involvement in regulated exocytosis has been suggested in various models, including neuroendocrine cells (14, 26). The development of molecular tools has enabled the identification of phospholipase D1 (PLD1) as the key enzyme responsible for PA synthesis during exocytosis (14, 27). Capacitance recordings from chromaffin cells silenced for PLD1 suggest that PLD1 controls the number of fusion competent secretory granules at the plasma membrane without affecting earlier recruitment or docking steps, leading to the idea that PA acts directly in membrane fusion (28). In agreement with this concept, a molecular sensor for PA revealed local PA accumulation at the plasma membrane near morphologically docked granules at sites of active exocytosis (28).

#### **OTHER LIPIDS**

Various other lipids are suspected to take part in regulated exocytosis. Although most of them have been implicated based on *in vitro* membrane fusion assays, some have also been studied in neuroendocrine cells. For instance, diacylglycerol (DAG) increases stimulus-coupled secretion by recruiting vesicles to the immediately releasable pool through the regulation of the vesicle priming protein Munc13-1 (29). Furthermore by activating protein kinase C, DAG may modulate the phosphorylation level of various proteins contributing or regulating the exocytotic machinery, including SNAP-25 and Munc18 (30, 31). Modulating PS levels also directly affects the rate of exocytosis in PC12 cells. Although it is likely that long term provision of high level of PS (incubation with 100µM for 48 h or over-expression of PS synthase) may affect numerous key cellular functions altogether, it was shown that PS influences exocytosis by enhancing fusion pore opening and slowing fusion pore dilatation in PC12 cells (32). The kinetic changes in exocytosis resulting from elevated PS levels were shown to be unlikely the reflection of an indirect action of PS on Ca2<sup>+</sup> channels, or vesicle size or number, or the levels of some of the major exocytosis proteins, but may be the consequence of synaptotagmin binding to PS (32). Finally, arachidonic acid produced from different phospholipids by phospholipase A2 and from DAG by DAG-lipase potentiates exocytosis from chromaffin cells (33, 34).

#### **LIPIDS AS RECRUITING COMPONENTS OF THE EXOCYTOTIC MACHINERY**

Within membranes, the ability of microdomains to sequester specific proteins and exclude others makes them ideally suited to spatially organize cellular pathways. Cholesterol-enriched microdomains seem to concentrate components of the exocytotic/fusion machinery. For instance, numerous studies of the distribution of SNARE proteins in various cell types suggest that SNAREs partially associate with detergent resistant, cholesterolenriched microdomains (11). Palmitoylation appears to be the major targeting signal in these microdomains, as in the case of SNAP-25, although it is likely that other elements contribute to the enrichment of constituents of the exocytotic machinery within these cholesterol microdomains. However despite intense research there is still little known about what lipid or protein molecules are actually present at sites of exocytosis.

Up to 20 proteins potentially involved in regulated exocytosis have been reported to bind PtdIns(4,5)P<sup>2</sup> (35). These include synaptotagmin (36), syntaxin (37), Ca2+-dependent activator protein for secretion (CAPS) (38), and members of the PLD signaling pathway (16, 39, 40). Using immunogold labeling of plasma membrane sheets combined with spatial point pattern analysis, we recently observed that PtdIns(4,5)P<sup>2</sup> microdomains co-localize with SNARE clusters and docked secretory granules (12). Translocation of the PtdIns(4,5)P2-binding protein annexin A2 to the plasma membrane following cell stimulation is a hallmark of chromaffin cell exocytosis (41). Annexin A2 plays an essential role in calcium-regulated exocytosis by promoting PtdIns(4,5)P<sup>2</sup> and cholesterol-enriched domains containing SNAREs in the vicinity of docked granules (9, 12). Altogether these observation raise the notion that functional exocytotic sites defined by specific lipids such as cholesterol, GM1, and PtdIns(4,5)P<sup>2</sup> are able to recruit and sequester components to build a machine that drives fast and efficient membrane fusion (**Figure 1**).

Molecular details of how PtdIns(4,5)P<sup>2</sup> forms a platform for vesicle recruitment have recently been proposed (42). In this study, synaptotagmin-1 was shown to interact independently of Ca2<sup>+</sup> with the polybasic linker region of syntaxin-1A already associated with PtdIns(4,5)P<sup>2</sup> at the plasma membrane. When then triggered by Ca2+, this interaction allows the Ca2+-binding sites of synaptotagmin-1 to bind PS in the vesicle membrane, thereby bridging the vesicle membrane close enough for SNARE assembly and subsequent membrane fusion. Interestingly, this polybasic juxtamembrane domain of syntaxin-1A which binds PtdIns(4,5)P<sup>2</sup> has also been shown to tightly bind both PA (37) and PtdIns(3,4,5)P<sup>3</sup> (43). Like PtdIns(4,5)P2, these anionic lipids can probably recruit syntaxin-1A, and it is tempting to propose that the recruitment of syntaxin isoforms may depend on the type of lipid present. Furthermore, the fact that these lipids can be quickly converted into different forms using kinases, lipases, and phosphatases, such a protein-recruiting mechanism offers a supplementary level of control to adapt the exocytotic machinery to the physiological demands put on the cell. Finally, using superresolution optical techniques and fluorescence lifetime imaging microscopy, it was shown that distinct t-SNARE intermediate states on the plasma membrane can be patterned by the underlying lipid environment (44). Undoubtedly these high-resolution imaging techniques will be useful to determine how lipids contribute to the organization of the exocytotic platform.

#### **LIPIDS AS REGULATORS OF THE EXOCYTOTIC MACHINERY**

PtdIns(4,5)P<sup>2</sup> directly binds to a large subset of proteins from the exocytotic machinery. *In vitro* experiments revealed that at physiological concentrations PtdIns(4,5)P<sup>2</sup> regulates exocytosis by recruiting priming factors such as CAPS, which facilitate SNAREdependent liposome fusion (45). PtdIns(4,5)P<sup>2</sup> is also a necessary cofactor for PLD1 activity and PA synthesis during exocytosis (39). Finally, PtdIns(4,5)P<sup>2</sup> controls actin polymerization by modulating the activity and targeting of actin regulatory proteins. Indeed the activity of the actin-binding proteins scinderin and gelsolin, two F-actin severing proteins that are constituents of the exocytotic machinery, is regulated by PtdIns(4,5)P<sup>2</sup> (46). A transient increase in PIP2 levels is sufficient to promote the mobilization and recruitment of secretory vesicles to the plasma membrane (22). PIP2 therefore links exocytosis and the actin cytoskeleton by

coordinating the actin-based delivery of secretory vesicles to the exocytotic sites.

Diacylglycerol production through hydrolysis of PtdIns(4,5)P<sup>2</sup> by phospholipase C is mandatory for exocytosis (47). DAG is essential in the priming of exocytosis, owing to the activation of protein kinase C and Munc13, which then modulate the function of syntaxin-1A (29). It is worth to mention that up to 10 different DAG-kinases may also produce DAG from various subcellular pools of PA during exocytosis. Finally DAG is further hydrolyzed by DAG lipases to liberate fatty acids and monoacylglycerols. This pathway is essential for exocytosis as inhibition of DAG lipase blocks exocytosis (48).

Several constituents and regulators of the exocytotic machinery have also been shown to bind to PA, including small GTPases, NSF, and syntaxin-1A (49). PA directly activates these proteins, but evidence that this activation directly contributes to exocytosis remains scarce. PA is also an essential cofactor of phosphatidylinositol-4-phosphate 5-kinase, which produces PtdIns(4,5)P2, suggesting a possible positive feedback loop in the synthesis of PA and PtdIns(4,5)P<sup>2</sup> (50). Although no direct evidence in neuroendocrine systems have shown that PA directly regulates the assembly or the function of the minimal fusion machinery, *in vitro* reconstituted fusion assays with purified yeast vacuolar SNAREs do so. Indeed experiments performed in a complex reconstituted system including the SNARE chaperones Sec17p/Sec18p, the multifunctional HOPS complex including a subunit of the Sec1-Munc18 family and vacuolar lipids suggest that PA is equally essential for SNARE complex assembly and for fusion. PA has been proposed to facilitate functional interactions among SNAREs and SNARE chaperones (**Figure 1**). Interestingly, in this system, PA could not be replaced by either lipids with small headgroups, such as DAG or acidic lipids, like PS or PI (51).

Arachidonic acid has been described to directly promote the assembly of syntaxin-3 with SNAP-25 and the formation of the ternary SNARE complex (52). Interestingly, omega-3 and omega-6 fatty acids, which play important roles in human health, have be shown to recapitulate this *in vitro* effect of arachidonic acid on SNARE complex formation, suggesting that syntaxins may represent crucial targets of polyunsaturated lipids (52). In other words, polyunsaturated lipids may physiologically regulate SNARE complex assembly and thus exocytosis. Along this same line, sphingosine a releasable backbone of sphingolipids, activates vesicular synaptobrevin facilitating the assembly of SNARE complexes required for membrane fusion (53). It is however important to note that the effects of arachidonic acid and sphingosine observed in these studies are all achieved near or at the CMC value for these lipids, treatments that may also lead to membrane disorganization like detergent action.

#### **LIPIDS AFFECT MEMBRANE TOPOLOGY DURING MEMBRANE FUSION: THE CONCEPT OF FUSOGENIC LIPIDS**

The most widely accepted model for membrane fusion, the stalk pore model proposes that the merging of *cis* contacting monolayers gives rise to a negatively curved lipid structure called a stalk. The structure of this stalk depends on the composition of the *cis* monolayers (the outer leaflet of the vesicle and the inner leaflet of the plasma membrane). This model implies that cone-shaped lipids such as cholesterol, DAG, or PA, which have intrinsic negative curvatures, in the *cis* leaflets of contacting bilayers would enhance membrane fusion (54). *Vice versa* inverted cone-shaped lipids (such as PS, gangliosides, or lysophospholipids) should prevent membrane fusion in the *cis* leaflets, but promote fusion when present in the outer leaflets (54). Interestingly, GM1 was found enriched in the outer leaflet of the plasma membrane at the sites of exocytosis in stimulated chromaffin cells (9). These GM1 domains may induce positive membrane curvature in the outer leaflet (55), thereby promoting fusion (**Figure 1**). Reconstituted fusion assays and direct addition of lipids on cultured cells validate the concept that PA, DAG, and cholesterol might promote membrane fusion by changing the spontaneous curvature of membranes [reviewed in (8)]. At physiological concentrations, PtdIns(4,5)P<sup>2</sup> inhibits SNARE-dependent liposome fusion (45), most likely due to its intrinsic positive curvature-promoting properties. However, PtdIns(4,5)P<sup>2</sup> has been described to be converted from an inverted cone-shaped structure to a cone-shaped form in the presence of calcium (56). Thus, in stimulated cells, a local accumulation of PA and PtdIns(4,5)P<sup>2</sup> at granule docking sites where GM1 is in the outer leaflets may well have a synergistic effect on membrane curvature and promote fusion (**Figure 1**). In an alternate mode of changing membrane topology, synaptotagmin has been proposed to facilitate membrane fusion by phase separating PS, a process that is expected to locally buckle bilayers and disorder lipids due to the curvature tendencies of PS (57). It is worth to mention that most of lipid mentioned in this review, also have the ability to flip from one leaflet to another. How this flipping is regulated and how it affects curvature remains an unsolved issue. However, it is likely that the ability of these lipids to interact with the fusion machinery largely controls these flipping properties.

#### **CONCLUSION**

As genuine components of the exocytotic machinery, lipids have several advantages over proteins for this task: lipids can directly change the intrinsic fusion properties of membranes, recruit, and/or activate a large number of different proteins to create a local environment in which exocytosis takes place (**Figure 1**). As illustrated in this review, a given lipid can play multiple functions, acting either individually or successively or even simultaneously in concert with other lipids. At the same time, the rapid enzymatic production and degradation of lipids at exocytotic sites allows the cell to remain flexible: by changing the lipid levels, physiological function can be modified within seconds or minutes without the need for protein synthesis or degradation. Over the last decade, *in vitro* reconstituted membrane fusion combined with precise methods to quantify specific lipid species and improved molecular and pharmacological tools to manipulate cellular levels of a given lipid, have lead to a better understanding of the capacities of lipids to promote exocytosis at different steps of the process. For different kinds of vesicles in different cell types, it is likely that the local lipid environment may differentially regulate fusion pore formation, enlargement, and duration, which may in part explain the great variety of fusion kinetics observed *in vivo*. Finally, lipids could also contribute to the tight coupling between exocytosis and the early

stages of membrane retrieval and endocytosis as highlighted in a review of this issue (Houy et al., submitted). Undoubtedly, the next challenge will be to follow individual lipid dynamics at the speed of pore formation and expansion.

#### **REFERENCES**


*U S A* (2001) **98**:5619–24. doi:10.1073/pnas.091502398


#### **ACKNOWLEDGMENTS**

We wish to thank Dr. Nancy Grant for critical reading of the manuscript. This work is supported by Agence Nationale pour la Recherche (ANR-09-BLAN-0264-01 grant) to Nicolas Vitale.


**1791**:936–41. doi:10.1016/j.bbalip. 2009.02.016


catecholamine secretion requires the translocation of p36 to the subplasmalemmal region in chromaffin cells. *J Cell Biol* (1996) **133**:1217–36. doi:10.1083/jcb.133. 6.1217


of cross-linking. *Biochem Biophys Res Commun* (2012) **422**:776–9. doi: 10.1016/j.bbrc.2012.05.083


**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: 18 June 2013; accepted: 31 August 2013; published online: 17 September 2013.*

*Citation: Ammar MR, Kassas N, Chasserot-Golaz S, Bader M-F and Vitale N (2013) Lipids in regulated exocytosis: what are they doing? Front. Endocrinol. 4:125. doi: 10.3389/fendo.2013.00125*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Ammar, Kassas, Chasserot-Golaz, Bader and Vitale. 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.*

#### **Olga Gubar 1,2,3, Dmytro Morderer 1,2, Lyudmila Tsyba<sup>2</sup> , Pauline Croisé<sup>3</sup> , Sébastien Houy <sup>3</sup> , Stéphane Ory <sup>3</sup> , Stéphane Gasman<sup>3</sup> and Alla Rynditch1,2\***

<sup>1</sup> State Key Laboratory of Molecular and Cellular Biology, Institute of Molecular Biology and Genetics, Kyiv, Ukraine

<sup>2</sup> Department of Functional Genomics, Institute of Molecular Biology and Genetics, Kyiv, Ukraine

3 Institut des Neurosciences Cellulaires et Intégratives, Centre National de la Recherche Scientifique-Unité Propre de Recherche 3212, Université de Strasbourg, Strasbourg, France

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Pedro A. Jose, Georgetown University, USA Sanda Predescu, Rush University Medical Center, USA

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

Alla Rynditch, Department of Functional Genomics, Institute of Molecular Biology and Genetics, 150, Zabolotnogo Street, 03680 Kyiv-143, Ukraine e-mail: rynditch@imbg.org.ua

Intersectins (ITSNs) are a family of highly conserved proteins with orthologs from nematodes to mammals. In vertebrates, ITSNs are encoded by two genes (itsn1 and itsn2), which act as scaffolds that were initially discovered as proteins involved in endocytosis. Further investigation demonstrated that ITSN1 is also implicated in several other processes including regulated exocytosis, thereby suggesting a role for ITSN1 in the coupling between exocytosis and endocytosis in excitatory cells. Despite a high degree of conservation amongst orthologs, ITSN function is not so well preserved as they have acquired new properties during evolution. In this review, we will discuss the role of ITSN1 and its orthologs in exo- and endocytosis, in particular in neurons and neuroendocrine cells.

**Keywords: intersectin, exocytosis, endocytosis, neurons, neuroendocrine cells, Cdc42, scaffold**

#### **ITSN FAMILY OF SCAFFOLD PROTEINS**

Intersectins (ITSNs) are multifunctional scaffold proteins implicated in several cellular mechanisms, including membrane trafficking (clathrin- and caveolin-mediated endocytosis, secretagogue-evoked exocytosis) and receptor-dependent signaling (Ras-MAPK and Rho GTPase regulation, EGF receptor ubiquitylation) to name a few [extensively reviewed in (1–3)]. Based on the high involvement of ITSNs in membrane trafficking, this mini review will focus on the role of these proteins in exocytosis and endocytosis in neurosecretory cells such as neurons and neuroendocrine cells, and will discuss how ITSNs could be key players in coupling exocytosis to endocytosis.

Intersectin is highly conserved in all metazoans examined so far and ITSN orthologs have been found in nematodes (*Caenorhabditis elegans*) (4), arthropods (*Drosophila melanogaster* Dap160 – Dynamin-associated protein 160 kDa) (5), fish (*Danio rerio*) (6), amphibians (*Xenopus laevis*) (7), and mammals (*Mus musculus*, *Rattus norvegicus*, *Homo sapiens*) (8–10) (**Figure 1**). ITSN is encoded by one gene in invertebrates. Its molecular organization consists of two N-terminal EH (Eps15 homology) domains followed by a coiled-coil region and four or five SH3 (Src homology 3) domains (5). In vertebrates, two genes encode ITSN proteins (ITSN1 and ITSN2). In addition, ITSNs exist in two main isoforms generated by alternative splicing: a short form (ITSN-S) that harbors the same domain organization as invertebrate ITSN and is ubiquitously expressed and a long form (ITSN-L) which has three additional domains in its C-terminal part [tandem of Dbl (DH) and pleckstrin homology (PH) domains and a C2 domain] (10). This extension has guanine nucleotide exchange factor (GEF) properties for Cdc42 (11),

a small GTPase of the Rho family. The ITSN1-L is enriched in neurons (10, 12).

#### **ITSN1 IN NEURONAL ENDOCYTOSIS**

Multimodular ITSN1 has been shown to interact with numerous endocytic proteins, including dynamin, AP2 (adaptor protein 2), proteins from the Epsin family, and the synaptojanin phosphatase; its function in the control of endocytosis has been described in different cell types and organisms [reviewed in (1, 2)]. The role of ITSN1 in endocytosis in neurons was further demonstrated in functional assays performed in various model organisms. *C. elegans* containing deletions in the *itsn-1* gene were hypersensitive to the acetylcholine esterase inhibitor aldicarb that is widely used for unmasking neurotransmission defects. It causes rapid hypercontraction and eventual death of wild-type worms. This *itsn-1-*null aldicarb-hypersensitive phenotype could be rescued by expressing ITSN-1 in neurons, but not in muscles, indicating a pre-synaptic dysfunction in mutant animals (4). Another study revealed a decrease in the number of vesicles in neuromuscular junctions (NMJ) of *itsn-1*-null worms, implying that ITSN-1 was required for synaptic vesicle (SV) recycling (13). In both cases, the mutant worms were viable and exhibited no changes in growth and locomotor activity, implying that the nervous system functioned normally. This suggests that ITSN-1 is not essential for neuronal endocytosis in nematodes, but rather plays a regulatory role increasing the efficiency of this process. Worm ITSN-1 has been shown to form a complex with the Eps15 ortholog EHS-1 (13) and a mutation in the *ehs-1* gene also leads to SV depletion. However, *ehs-1*/*itsn-1* double mutants develop a phenotype similar to that of single

mutants, supporting the idea that these proteins act in the same process (13, 14). Intriguingly, *itsn-1* and *ehs-1* mutants have opposite sensitivities to aldicarb, indicating that they have different effects on cholinergic neurotransmission, most likely at the pre-synaptic level (4). Finally, in addition to pre-synaptic functions, ITSN-1 is required for internalization of the GLR-1 (glutamate receptor 1) subunit of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid-type glutamate (AMPA) receptors in the post-synaptic terminals of nematode command interneurons (15). Thus ITSN-1 seems to be important for both pre- and post-synaptic endocytosis in worms.

In contrast to nematodes, the loss-of-function mutation of the *ITSN* ortholog *Dap160* in *Drosophila* is lethal at late larval stages. Functional studies on the NMJ of mutant flies did not reveal significant changes in stimuli-evoked SV exocytosis; however, prolonged stimulation led to a severe decline in the excitatory post-synaptic potential (EPSP), indicating impairment of compensatory endocytosis (16, 17). As in nematodes, the number of vesicles in the active zones decreased in the NMJ of mutant *Drosophila* (16). In agreement with a function of ITSN in tethering endocytic proteins to the sites of endocytosis, the loss of Dap160 leads to mislocalization of dynamin, endophilin, and synaptojanin, resulting in decreased levels of these proteins in periactive synaptic zones (16, 17). Interestingly, overexpression of Dap160 also leads to a decrease in the level of synaptojanin, a 5<sup>0</sup> -phosphoinositide phosphatase involved in late stages of endocytosis, in NMJ synaptic terminals (18), indicating that Dap160/ITSN1 functions as a scaffold in endocytosis and is required to maintain endocytic proteins in synaptic terminals at levels necessary for efficient retrieval of SVs.

Studies on lamprey giant reticulospinal synapse have confirmed the data obtained in invertebrates concerning the involvement of ITSN1 in SV endocytosis. Perturbation of ITSN1 function by injecting either anti-ITSN1 antibodies or the SH3C domain into the pre-synaptic terminal resulted in an accumulation of clathrin-coated pits and a reduction in SV number, suggesting a defect in compensatory endocytosis (19, 20). It has also been shown that ITSN1 binds the AP2 clathrin adaptor both in lamprey and mammalian brain, suggesting that ITSN1 participates in the early steps of clathrin-mediated SV endocytosis. Moreover, such an interaction prevents ITSN1 from associating with synaptojanin 1 (20). These findings imply that ITSN1 temporally regulates SV endocytosis.

The data regarding the role of ITSN1 in SV endocytosis in mammals remain contradictory. ITSN1 along with Eps15 have been shown to form a complex with FCHo (Fer/Cip4 homology domain-only) proteins, that promotes the initiation of clathrincoated pit formation – the step required for clathrin-mediated endocytosis of different cargo in various cell types, including SV endocytosis in neurons (21). ITSN1 isoforms have also been identified as members of the synaptotagmin I-associated endocytic complex in the synaptosomal fraction of rat brain, and were subsequently shown to co-localize with other members of this complex in the pre-synaptic transmitter-release face of the giant calyx-type synapse of the chick ciliary ganglion (22). Knockout of *ITSN1* in mice slows down SV endocytosis in neurons, although the animals are viable (23). However, another group found that SV endocytosis remains normal following ITSN1 knock-down in rat primary hippocampal neurons. This was demonstrated by the absence of a decrease in the uptake of FM4-64 membrane dye after KCl stimulation (24). Moreover, using capacitance measurements of plasma membrane in knockout mouse calyx of Held synapse, Sakaba and co-workers have recently shown that ITSN1 is not essential for endocytic SV recovery after pulse stimulation (25). Furthermore, several authors have reported that ITSN1 is mostly localized post-synaptically in hippocampal neurons, (24, 26), although some data indicating a pre-synaptic localization in mammalian synapses also exist (20, 25). Finally, recent electrophysiological studies did not reveal abnormalities in synaptic transmission either in ITSN1 knockout mice or in ITSN1/ITSN2 double knockout mice using a wide range of stimulation protocols, indicating that under the conditions tested SV recycling is not perturbed (27). Therefore it seems possible that in mammals, in contrast to invertebrates, ITSN1 acts predominantly postsynaptically; however, we cannot totally exclude it is an accessory protein in SV recycling.

Thus the role of ITSN in neuronal endocytosis apparently has not been conserved during evolution. In some groups of organisms (e.g., arthropods) ITSN is indispensable for normal SV recycling, whereas in higher evolved organisms it has become non-essential probably due to the development of alternative or compensatory mechanisms. In lower organisms, these functions are fulfilled by the short isoform, whereas in higher organisms (e.g., vertebrates) ITSN1-L is the likely candidate as it is the main isoform in neurons.

#### **ITSN1 FUNCTION IN EXOCYTOSIS IN NEUROENDOCRINE CELLS**

There is much less evidence about the implication of ITSN in exocytosis. One of the first studies reported the interaction of ITSN1 with t-SNAREs [target Soluble NSF Attachment Protein (SNAP) Receptor] SNAP-23 and SNAP-25, but no functional consequence was demonstrated (9). A new specific function of ITSN1-L as GEF for Cdc42 in exocytosis was described in neuroendocrine cells. In regulated exocytosis, remodeling of the dense actin cortical network is an important step that is controlled by small GTPases. The key players in this process are the Rho family GTPases [reviewed in (28, 29)]. In the PC12 rat pheochromocytoma secretory model, Cdc42 was shown to be activated near the plasma membrane during exocytosis, where it recruits Neural Wiskott–Aldrich syndrome protein (N-WASP) and induces actin polymerization (30). ITSN1-L appeared to be an ideal candidate for Cdc42 activation at docking sites for secretory granules, because it is a specific GEF for Cdc42 and at the same time binds to its effector N-WASP (11). This leads to local polymerization of actin, thereby facilitating exocytosis. ITSN1-L was observed to co-localize with exocytic sites in PC12 and primary bovine chromaffin cells (31). Moreover, silencing of ITSN1 (as well as of Cdc42) significantly inhibits regulated exocytosis in PC12 cells, whereas overexpression of the C-terminal part of ITSN1-L (DH-PH-C2 domains) promotes exocytosis and peripheral actin polymerization in neuroendocrine cells (31, 32). These results were confirmed in *ITSN1*-null mice where exocytosis is also reduced in chromaffin cells (23). Finally, ITSN1 was very recently shown to regulate the replenishment of the fast-releasing SV pool in mouse calyx of Held synapse, possibly together with dynamin 1 and as a GEF for Cdc42 (25). Thus ITSN1-L displays novel properties in membrane trafficking distinct from those of ITSN1-S that were probably acquired during evolution.

#### **ITSN AS A PLAUSIBLE LINK BETWEEN EXO- AND ENDOCYTOSIS**

Since ITSN1, especially the isoform ITSN1-L, is strongly implicated in both endo- and exocytosis it is an ideal protein for the coupling of these processes. Regulated exocytosis is always followed by compensatory endocytosis, which is also Ca2+-dependent and is stimulated together with exocytosis (33). Thus the presence of a calcium-binding C2 domain in ITSN-L may be a key for its implication in both exo- and endocytosis. Moreover, in studies on the chick ciliary ganglion ITSN1-S and ITSN1-L have been reported to be associated with the CaV2.2 calcium channels at transmitter-release sites (34).

In neurons, the active zone is highly structured to provide maximum secretory efficiency. The protein matrix, which consists of the cytoskeleton and scaffold proteins, ensures efficient SV docking, whereas electron-dense projections from this matrix serve to tether and maintain the ready-to-use SV pool (35). It is surrounded by a periactive zone that serves for SV recycling by endocytosis. A tight connection and coordination of SV release and recycling is provided by scaffolds [reviewed in (36)]. ITSN1 is a perfect candidate to fulfill this role. During synaptic activity, Dap160 has recently been shown to shuttle between active and periactive zones, ensuring the delivery of dynamin to the periactive zone for efficient bulk membrane retrieval (37). In the lamprey giant reticulospinal synapse, ITSN1 is also localized in the active zone of resting synapses, but re-localizes with its major

#### **FIGURE 2 | Hypothetical model of ITSN1 coupling exo- and endocytosis in neuroendocrine cells**. SNAP-25 recruits ITSN1-L to the SNARE complex (VAMP2, SNAP-25, syntaxin 1) at the exocytic sites where ITSN1-L facilitates regulated exocytosis by activating Cdc42. Immediately after vesicle fusion and hormone release, ITSN1 promotes clathrin-mediated endocytosis of the

granule membrane via AP2, and facilitates later stages of endocytosis together with dynamin and synaptojanin. ITSN1 then stays in the vesicle cluster (probably with dynamin) to be recruited back to the membrane for the next release cycle. ITSN1, total ITSN1(both short and long isoforms); ITSN1-L, long isoform of ITSN1.

binding partner dynamin to the periactive zone upon stimulation (19). These results support the idea that ITSN1 acts as a scaffold coupling vesicle fusion and internalization events.

Although to date there is currently no direct evidence that ITSN1 plays a role in compensatory endocytosis in neuroendocrine cells, such data do exist for other cell types (including neurons). Based on these data and the presence of ITSN1 partners in neuroendocrine cells, we propose the following hypothetical model of ITSN1 functioning in exo- and endocytosis in these cells (**Figure 2**). ITSN1 may function in both the release and retrieval of the dense-core granules. SNAP-25 could recruit ITSN1-L to exocytic sites where the latter activates Cdc42 thereby inducing actin polymerization and facilitating the late stages of secretion (9, 31, 32). After release, ITSN1-L remaining on the vesicle membrane could directly promote assembly of the endocytic complex, because the vesicle membrane has been shown to be preserved as an entity during the compensatory endocytosis in neuroendocrine

#### **REFERENCES**


*Chem* (1998) **273**(47):31401–7. doi: 10.1074/jbc.273.47.31401


cells (38). Otherwise, ITSN1 could be recruited *de novo* and induce clathrin coat assembly via AP2 (20). Then ITSN1, in turn, can recruit dynamin to provoke vesicle scission (8, 9). After vesicle dissociation from the membrane, AP2 could be replaced by synaptojanin 1, which takes part in the uncoating and dephosphorylation of PIP<sup>2</sup> (phosphatidylinositol 4,5-bisphosphate) (20). The vesicle is then refilled and can be tethered ready for another cycle of exocytosis or added to the releasable vesicle pool. ITSN1 may accompany the vesicle throughout this cycle and remain in the releasable pool as it does in neurons (19). Thus ITSN1 could be a multipurpose player providing the crossroad between exo- and endocytosis in neurosecretory cells and an important regulator of these processes.

#### **ACKNOWLEDGMENTS**

We thank Dr. A.-L. Haenni and Dr. N. Grant for helpful discussions and comments on the manuscript.


the clathrin adaptor complex AP2. *Proc Natl Acad Sci U S A* (2010) **107**(9):4206–11. doi:10.1073/pnas. 0911073107


PS, Bader MF, et al. Intersectin-1L nucleotide exchange factor regulates secretory granule exocytosis by activating Cdc42. *EMBO J* (2006) **25**(15):3494–503. doi:10. 1038/sj.emboj.7601247


presynaptic nerve terminal. *Physiology (Bethesda)* (2004) **19**:262–70. doi:10.1152/physiol.00014.2004


**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: 01 July 2013; accepted: 09 August 2013; published online: 27 August 2013.*

*Citation: Gubar O, Morderer D, Tsyba L, Croisé P, Houy S, Ory S, Gasman S and Rynditch A (2013) Intersectin: the crossroad between vesicle exocytosis and endocytosis. Front. Endocrinol. 4:109. doi: 10.3389/fendo.2013.00109*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Gubar, Morderer, Tsyba, Croisé, Houy, Ory, Gasman and Rynditch. 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.*

#### **Sébastien Houy, Pauline Croisé, Olga Gubar, Sylvette Chasserot-Golaz, Petra Tryoen-Tóth,Yannick Bailly, Stéphane Ory, Marie-France Bader and Stéphane Gasman\***

Institut des Neurosciences Cellulaires et Intégratives (INCI), Centre National de la Recherche Scientifique (CNRS UPR 3212), Université de Strasbourg, Strasbourg, France

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Silvio O. Rizzoli, European Neuroscience Institute Göttingen, Germany Corey Smith, Case Western Reserve University, USA

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

Stéphane Gasman, Institut des Neurosciences Cellulaires et Intégratives (INCI), Centre National de la Recherche Scientifique (CNRS UPR 3212), Université de Strasbourg, 5 rue Blaise Pascal, Strasbourg 67084, France e-mail: gasman@inci-cnrs.unistra.fr

Although much has been learned concerning the mechanisms of secretory vesicle formation and fusion at donor and acceptor membrane compartments, relatively little attention has been paid toward understanding how cells maintain a homeostatic membrane balance through vesicular trafficking. In neurons and neuroendocrine cells, release of neurotransmitters, neuropeptides, and hormones occurs through calcium-regulated exocytosis at the plasma membrane. To allow recycling of secretory vesicle components and to preserve organelles integrity, cells must initiate and regulate compensatory membrane uptake. This review relates the fate of secretory granule membranes after full fusion exocytosis in neuroendocrine cells. In particular, we focus on the potential role of lipids in preserving and sorting secretory granule membranes after exocytosis and we discuss the potential mechanisms of membrane retrieval.

**Keywords: exocytosis, compensatory endocytosis, membrane lipids, neuroendocrine cells, chromaffin cells**

#### **INTRODUCTION**

Mammalian cells exhibit complex and dynamic patterns of intracellular membrane traffic between various organelles. Although much has been learned concerning the mechanisms of vesicle transport and vesicle fusion at donor and acceptor compartments, relatively little attention has been paid to understanding how organelle homeostasis is maintained. This aspect is particularly important in neurosecretory cells in which intense membrane trafficking and mixing occurs between secretory vesicles and the plasma membrane during secretion of various transmitters, peptides, and hormones. Calcium-regulated exocytosis, i.e., fusion of secretory vesicles with the plasma membrane results in the merging of these two membrane compartments, hence triggering an increase in plasma membrane surface and loss of identity. As a consequence, exocytosis must be coupled to a compensatory endocytotic process allowing the plasma membrane to recover its integrity and the granule membrane to be recycled. In neurons, molecular mechanisms of synaptic vesicle recycling and coupling with exocytosis have been intensively studied, but is still debated [for reviews see (1–3)]. However, the equivalent process for large dense core granules in neuroendocrine cells remains largely unexplored.

In neuroendocrine cells, secretion can occur through different modes of exocytosis depending on the physiological demand (see **Figure 1** for details). The "kiss-and-run" mode allows only the release of catecholamines and other small molecules through a narrow fusion pore (4, 5). During "cavicapture" (granule cavity capture) mode, expansion of the fusion pore triggers partial release of the small proteins (6–9). During kiss-and-run and cavicapture processes, the granule shape remains almost intact, whereas during "full collapse" or "full fusion" exocytosis, granules lose their round shape, flatten out in the plane of the plasma membrane leading to the merging of these two compartments and the complete release of the granular content (10–12). Whereas the molecular mechanisms of the various exocytotic modes (secretory vesicle recruitment, docking, priming, and fusion processes) have been largely explored, how granule and plasma membranes maintain their composition and recover their integrity after full fusion exocytosis is poorly known. Two types of retrieval have been described after full fusion exocytosis: clathrin-mediated endocytosis and bulk endocytosis. Bulk endocytosis occurs during elevated secretory activity when clathrin-mediated endocytosis is unable to fully compensate the large increase in membrane surface. To rapidly reverse this excess of plasma membrane, bulk endocytosis internalizes large invaginations of plasma membrane, which then form endosomal-like compartments. Bulk endocytosis has been described in several reviews (13, 14). Here, we focus on potential mechanisms that allow neuroendocrine cells to compensate full fusion exocytosis of large dense core vesicles through clathrin-mediated endocytosis.

#### **THE FATE OF THE SECRETORY GRANULE MEMBRANE AFTER FUSION**

For a long time, it was believed that after full fusion exocytosis vesicular components diffuse into the plasma membrane and are subsequently randomly internalized. This model implies that both the secretory vesicle and plasma membranes lose their identities and that exocytosis is not directly coupled to endocytosis. In contradiction with this model, synaptic activity in neurons results neither in the overall dispersion of vesicle components in the plasma membrane nor in the enrichment of

**neuroendocrine cells**. Hormone sorting and large dense core granule biogenesis occurs at the trans-Golgi network. Mature granules either constitute the reserve pool or are recruited to the plasma membrane as a readily releasable pool. Large proteins (blue dots), small neuropeptides (black dots), and small molecules like catecholamines (red dots) can be released differentially according to the exo-endocytosis mode. During "kiss-and-run" mode, only small molecules are released through a narrow fusion pore, whereas cavicapture (granule cavity capture) allow the partial release of small neuropeptides (7). Note that for these two modes, retrieval of intact granules

plasma membrane components in synaptic vesicles (2). Similarly, despite full fusion exocytosis in neuroendocrine cells, granules, and plasma membranes seem to maintain their specific protein composition.

Early evidence for exo-endocytosis coupling came from morphological studies in the 80s suggesting that large dense core granule membrane-bound components could be retrieved after exocytosis (10, 15). At the same time, Geisow and co-workers observed an important increase in the number of coated pits containing secretory granule components in secretagogue-stimulated chromaffin cells (16). Later on patch-clamp and imaging studies suggested a fast temporal coupling between exocytosis and endocytosis processes (17, 18).

Using electron microscopy of cultured chromaffin cells, our group has recently described clustering of secretory granule proteins on the plasma membrane after full fusion exocytosis, arguing against the idea that granule components are dispersed in the plasma membrane (12). The group of Holz recently confirmed that chromaffin granule markers remain associated after fusion is easily conceived as the granule shape remains almost intact. During full fusion exocytosis, the intra-granular contents are all released and the granule membrane collapses into the plasma membrane. This membrane incorporation is compensated by clathrin-mediated endocytosis that specifically retrieves granule membrane piece by piece [see **Figure 2** and (19)]. After uncoating, the endocytic granule membrane reaches early endosome where granule components remain clustered (12), before being re-maturated at the Golgi network. During intense exocytotic activity, bulk endocytosis supports clathrin-mediated endocytosis by internalizing large plasma membrane invaginations that most likely follow the lysosomal degradation pathway.

(19). Additionally, we have shown that these granule-bound proteins are subsequently internalized through vesicles devoid of plasma membrane makers (12). In other words, granule membranes are maintained together as"microdomains"after exocytosis and are subsequently recaptured without intermixing with the plasma membrane. How do neuroendocrine cells preserve granule membrane integrity after full collapse and precisely sort granule membrane lipids and associated proteins?

#### **LIPIDS AS CENTRAL ORGANIZERS OF EXO-ENDOCYTOSIS COUPLING?**

#### **CREATING MEMBRANE DOMAINS TO PRESERVE GRANULE IDENTITY**

The preservation of secretory granule identity after fusion with the plasma membrane implies that both proteins and lipids do not diffuse in the plasma membrane. Lateral segregation of membrane lipids would in this case represent an obvious sorting mechanism. Interestingly, we and others have shown that exocytosis requires several types of lipid remodeling processes (described in the present "Research Topic" by Amar et al. (28)). Some of these processes might contribute to prevent granular component from diffusing. For instance, secretagogue-evoked stimulation of chromaffin cells triggers the formation of lipid raft microdomains at the plasma membrane enriched in ganglioside GM1, cholesterol, and phosphatidylinositol 4,5 bisphosphate [PI(4,5)P2]. Such lipid rafts correspond to membrane areas stabilized by the presence of cholesterol within a liquid-ordered phase in which lateral diffusion of proteins and lipids is limited, resulting in the clustering of specific components (20). In neuroendocrine and various other secretory cells, lipid raft formation is necessary for the spatial organization of the exocytotic machinery including SNARE proteins (21–23). As a consequence, it is tempting to imagine that this lipid confinement at the exocytotic sites would help to prevent granular lipids and proteins from diffusing after secretory granules fusion with the plasma membrane.

Of particular interest, the formation of membrane domains corresponding to exocytotic sites is regulated by annexin-A2 (21, 24), a calcium- and phospholipid-binding protein involved in both exo- and endocytosis (25). Annexin-A2 has been described on clathrin-coated vesicles in the adrenal gland (26). The protein displays two typical YXX∅ endocytic motifs allowing its interaction with the µ2-subunit of the AP-2 complex that triggers clathrin recruitment (27). Therefore, annexin-A2 constitutes a strong candidate to participate in the coupling of secretory granule exocytosis with the subsequent compensatory endocytosis.

#### **PI(4,5)P2: ORCHESTRATING EXO-ENDOCYTOSIS COUPLING**

Phosphatidylinositol 4,5 bisphosphate [PI(4,5)P2] has been largely described as an important regulator in exocytosis (28, 29) but it is also known to recruit and regulate multiple components involved in clathrin-mediated endocytosis. Therefore, the PI(4,5)P2-enriched membrane microdomains where exocytosis occurs is likely to constitute preferential spots for endocytosis, a perfect way to couple these two processes. For example, interaction of the adaptor protein AP-2 with YXX∅ endocytic motifs is driven by its interaction with PI(4,5)P<sup>2</sup> (30). Local PI(4,5)P<sup>2</sup> concentration regulates the membrane binding and deformation capacity of proteins containing Bin/amphiphysin/Rvs (BAR) domains including endophilin, syndapin, and amphiphysin, three membrane-deforming scaffold proteins that have been implicated in endocytotic processes (31, 32). Interestingly, it has been proposed that BAR domains participate in the fission of budding vesicles by synergistically cooperating with dynamin, a GTPase also sensitive to PI(4,5)P<sup>2</sup> (33). Moreover, in collaboration with the group of Dr. Cardenas, we have recently demonstrated that dynamin-2 controls both exocytosis, by regulating fusion pore expansion, and the subsequent endocytosis of secretory granules in chromaffin cells (34). Finally, PI(4,5)P<sup>2</sup> also regulates the dynamics of actin filaments, which are believed to limit plasma membrane protein diffusion and/or to directly participate in endocytosis (35, 36). Accordingly, we have previously demonstrated that actin filaments are formed at a post-docking step of exocytosis (37) and disruption of actin filament organization inhibits compensatory endocytosis after full fusion. These data suggest that actin remodeling is also implicated in the process of internalization *per se* (12).

#### **PHOSPHOLIPID SCRAMBLING: A SIGNAL TO TRIGGER COMPENSATORY ENDOCYTOSIS?**

One key feature of cell membranes is the asymmetric distribution of phospholipids between the leaflets. In the plasma membrane, phosphatidylserine (PS) and phosphatidylethanolamine (PE) reside in the inner cytoplasmic leaflet while phosphatidylcholine and sphingomyelin are located in the outer leaflet (38). In non-apoptotic cells, several biological functions are accompanied by a disruption of this phospholipid asymmetry resulting in the externalization of PS in the outer leaflet of the plasma membrane (39). This phenomenon has been observed during regulated exocytosis in mast cells (40), nerve terminals (41), and the neuroendocrine PC12 and chromaffin cells (12, 42, 43). More recently, we have shown that PS externalization occurs in specific domains at the frontier between the fused granule membrane and the plasma membrane and is triggered by the calcium-sensitive phospholipid scramblase-1 (PLSCR1). Interestingly, in chromaffin cells cultured from PLSCR1 knock-out mice, surface exposure of PS is not involved in exocytosis, but is required for granule membrane compensatory endocytosis (44).

To date no current evidence is available to explain the mechanism by which PS externalization is linked to compensatory endocytosis. However, two scenarios are possible. Firstly, as loss of phospholipid asymmetry can modify the mechanical stability of membranes (45), this might facilitates local reorganization of lipids surrounding the granule membrane transiently inserted within the plasma membrane and preserve the integrity of the granule membrane. Secondly, as an anionic phospholipid, PS confers negative charges and directly binds various proteins involved in exocytosis like annexin-A2, rabphilin, DOC 2, or synaptotagmin (46). PLSCR1-induced local decrease in PS concentration in the inner leaflet of the plasma membrane could therefore represent a signal to switch from exocytosis to endocytosis, thereby permitting the release of exocytotic components and/or the recruitment of the endocytic machinery.

However, whether externalization of PS simply reflects the loss of PS asymmetry or reveals more profound lipid reorganization is basically unknown and requires further investigation.

#### **HOW TO RECAPTURE A LARGE DENSE CORE VESICLE?**

In chromaffin cells, we have found that clathrin is rapidly recruited to the granule membrane right after merging with the plasma membrane and that knocking-down clathrin expression drastically blocks compensatory endocytosis (12). The clathrin dependency of large dense core granule endocytosis raises the question of the mechanism by which a granule displaying a mean diameter of about 250 nm may be recaptured by a clathrin coat? Energetic constraints indicate that *in vitro* clathrin baskets assemble with a mean diameter of 100 nm or even less (80 nm) in the presence of adaptor proteins (47). In INS-1 insulinoma cells for example, the mean diameter of secretory granule ranges from 110 to 170 nm but the endocytotic events detected by capacitance measurements correspond to vesicles of 70 nm diameter (48). In mouse chromaffin cells, the average size of endocytic vesicles calculated from endocytic capacitance step sizes is 122 nm (49). Accordingly, recent electron microscopy experiments performed on chromaffin cells demonstrated that clathrin-coated vesicles size increases

upon potassium stimulation (mean diameter of 87 nm) compared to resting cells (19). Altogether, these observations tend to demonstrate that large granule membranes may be recaptured as small pieces rather than a whole. Indeed, direct observations of the internalization of fused granule membranes allowed us to reveal early endocytotic events corresponding to 50–80 nm coated vesicles budding from granular membrane-bound proteins clusters (**Figure 2**).

incubated in the presence of antibodies raised against the luminal region of dopamine-beta hydroxylase (DBH), a transmembrane marker of

#### **ENDOCYTIC PATHWAY**

The intracellular route followed by the post-exocytotic internalized granules has not been fully characterized. The main difference with synaptic vesicle recycling is that, to be reused, large dense core granules need to be reloaded with matrix proteins, which most likely implies a re-maturation process involving the Golgi apparatus. The transient accessibility of granule-bound proteins at the cell surface during full fusion exocytosis has been exploited to label and follow post-fusion granules with specific antibodies or biotinylation. This approach was widely used in the 80s in order to demonstrate that granule-bound proteins transit through the Golgi region before being recycled in newly mature granules (10, 15, 50, 51). Alternatively, granule markers have been proposed to be degraded through a lysosomal pathway (52, 53). Both recycling and degradation pathways coexist and their proportion may depend on cell secretory activity. The recycling pathway leading to releasable granules is more predominant upon mild stimulation whereas above a certain threshold of membrane incorporation during intense exocytotic activity, the occurrence of bulk endocytosis will lead to the degradation of the internalized membrane (54, 55). Altogether, these data do not provide any information concerning the immediate fate of internalized granule membrane. Our group has observed that internalized granule-bound markers rapidly co-localize with the early endosomal marker EEA1, suggesting that chromaffin granule components might be retrieved through early endosomes after regulated exocytosis (12). Since clathrin is likely to retrieve the collapsed granule membrane as pieces and not as a whole, the early endosomes might constitute a transient sorting

station that would sort the retrieved pieces to reconstitute a functional granule prior to entering the retrograde transport pathway to the TGN. Accordingly, we have observed budding of the immunogold-labeled DBH clusters present on endosomes (unpublished data).

arrowhead shows a granule fusing with the plasma membrane.

#### **CONCLUSION**

Bar = 100 nm.

New evidence is now emerging to support the idea that, in neuroendocrine cells but also in neurons, vesicle/granule membranes do not intermix with the plasma membrane following full fusion exocytosis (56, 57). Compensation of membrane incorporation by endocytosis is a critical process and selective recapture of secretory organelles is required to maintain cellular homeostasis. Resolving the mechanisms that specifically preserve the granule membrane platform and retain granular components together after its incorporation in the plasma membrane to is the next challenging question to answer. Highly resolutive biophotonic approaches are now required to precisely investigate the dynamic behavior of secretory granules merged with the plasma membrane during and after exocytosis. Lipids clearly play a central role in this process, but attention should also be given to bi-functional proteins regulating both exo- and endocytosis, in particular annexin, synaptotagmin, and BAR domain-containing proteins.

#### **ACKNOWLEDGMENTS**

We wish to thank Dr. Nancy Grant for critical reading of the manuscript. We thank V. Calco and A. M. Haeberlé for setting up preliminary experiments and T. Thahouly for technical assistance. Part of the work from our group discussed in this review was supported by various grant to Stéphane Gasman: ANR grant (ANR-07-JCJC-088-01), "Association pour la Recherche sur le Cancer" grant (ARC #1055), and EMMA service (under the EU contract Grant Agreement Number 227490) of the EC FP7 Capacities Specific Programme. We acknowledge the Plateforme "Imagerie in vitro" (NeuroPôle; Strasbourg). Especially A. M. Haeberlé and V. Demais for their excellent technical assistance in electron microscopy.

#### **REFERENCES**


(1986) **102**:510–5. doi:10.1083/jcb. 102.2.510


Gerke V, Bader MF. Annexin 2 promotes the formation of lipid microdomains required for calcium-regulated exocytosis of dense-core vesicles. *Mol Biol Cell* (2005) **16**:1108–19. doi:10.1091/mbc.E04-07-0627


release of neurotransmitters from cholinergic nerve terminals isolated from electric ray *Narke japonica*. *Neurosci Lett* (2000) **291**: 21–4. doi:10.1016/S0304-3940(00) 01365-3


*Proc Natl Acad Sci U S A* (2002) **99**:1943–8. doi:10.1073/pnas. 042688399


fusion. *Commun Integr Biol* (2010) **3**:427–9. doi:10.4161/cib.3.5.12132

**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: 10 July 2013; accepted: 13 September 2013; published online: 02 October 2013.*

*Citation: Houy S, Croisé P, Gubar O, Chasserot-Golaz S, Tryoen-Tóth P, Bailly Y, Ory S, Bader M-F and Gasman S (2013) Exocytosis and endocytosis in neuroendocrine cells: inseparable membranes! Front. Endocrinol. 4:135. doi: 10.3389/fendo.2013.00135*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Houy, Croisé, Gubar, Chasserot-Golaz, Tryoen-Tóth, Bailly, Ory, Bader and Gasman. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, providedthe 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.*

# Dynamin-2 function and dysfunction along the secretory pathway

#### **Arlek M. González-Jamett <sup>1</sup> , Fanny Momboisse<sup>1</sup> , Valentina Haro-Acuña<sup>1</sup> , Jorge A. Bevilacqua<sup>2</sup> , Pablo Caviedes <sup>3</sup> and Ana María Cárdenas <sup>1</sup>\***

<sup>1</sup> Facultad de Ciencias, Centro Interdisciplinario de Neurociencia de Valparaíso, Universidad de Valparaíso, Valparaíso, Chile

<sup>2</sup> Programa de Anatomía y Biología del Desarrollo, ICBM, Facultad de Medicina, Departamento de Neurología y Neurocirugía, Hospital Clínico Universidad de Chile, Universidad de Chile, Santiago, Chile

<sup>3</sup> Programa de Farmacología Molecular y Clínica, ICBM, Facultad de Medicina, Universidad de Chile, Santiago, Chile

#### **Edited by:**

Stephane Gasman, Centre national de la Recherche Scientifique, France

#### **Reviewed by:**

Jocelyn Laporte, Institut National de la Santé et de la Recherche Médicale, France Fernando Diego Marengo, Universidad de Buenos Aires, Argentina

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

Ana María Cárdenas, Facultad de Ciencias, Centro Interdisciplinario de Neurociencia de Valparaíso, Universidad de Valparaíso, Gran Bretaña 1111, Playa Ancha 2360102, Valparaíso, Chile e-mail: ana.cardenas@uv.cl

Dynamin-2 is a ubiquitously expressed mechano-GTPase involved in different stages of the secretory pathway. Its most well-known function relates to the scission of nascent vesicles from the plasma membrane during endocytosis; however, it also participates in the formation of new vesicles from the Golgi network, vesicle trafficking, fusion processes and in the regulation of microtubule, and actin cytoskeleton dynamics. Over the last 8 years, more than 20 mutations in the dynamin-2 gene have been associated to two hereditary neuromuscular disorders: Charcot–Marie–Tooth neuropathy and centronuclear myopathy. Most of these mutations are grouped in the pleckstrin homology domain; however, there are no common mutations associated with both disorders, suggesting that they differently impact on dynamin-2 function in diverse tissues. In this review, we discuss the impact of these disease-related mutations on dynamin-2 function during vesicle trafficking and endocytotic processes.

**Keywords: dynamin-2, endocytosis, exocytosis, actin, microtubules, mutations, Charcot–Marie–Tooth neuropathy, centronuclear myopathy**

#### **INTRODUCTION**

Dynamin was identified for the first time almost 25 years ago as a 100 kDa microtubule-associated protein that induced microtubule bundles and promoted microtubule sliding *in vivo* (1). As described by the same authors, the motor activity of dynamin required ATP and other co-purified polypeptides (1). A year later, the same group cloned and sequenced dynamin, and found that it contained a consensus GTP-binding site (2), and subsequently characterized its GTPase activity (3). At present, three dynamin isoforms encoded by three distinct genes (*DNM1*, *DNM2*, and *DNM3*) have been described in mammals (4). These exhibit approximately 80% homology in their sequences, yet they differ in their tissue expression pattern; dynamin-1 is mainly expressed in neuronal tissue, dynamin-2 is ubiquitously expressed, and dynamin-3 is expressed in brain, testis, and lungs (5). Of these three dynamin isoforms, only dynamin-2 appears to play a pleiotropic role during embryonic development (6). In fact,studies in knock-out animals show that deletion of dynamin-1 or -3 can be compensated by the other dynamin isoforms (7), while the deletion of dynamin-2 causes early embryonic lethality (8). Moreover, as discussed below, mutations in *DMN2* result in severe hereditary neuropathies and myopathies in humans, strongly suggesting that dynamin-2 has more susceptible functions in the nervous and skeletal muscle tissues.

All dynamin isoforms exhibit at least four alternatively spliced variants, resulting in different dynamin proteins (5) that share a primary structure comprising: a large amino-terminal GTPase domain (G-domain) that binds and hydrolyzes GTP; a middle and a GTPase effector domains (GED) that form a "stalk" structurally essential region; a pleckstrin homology domain (PH) that binds inositol phospholipids and a carboxy-terminal proline and arginine rich-domain (PRD) that allows interaction with SH3-domain-containing-proteins (5) (**Figure 1**).

Dynamin function relies on its ability to form high order oligomers, and its self-assembly is necessary to promote its catalytic activity. Purified dynamin has been shown to spontaneously polymerize in the presence of negatively charged tubular templates such as lipid membranes (11), microtubules (3, 12), or actin bundles (13, 14) as well as after incubation in low ionic strength solutions (15). Over the last years several cryoelectron microscopy (16–18) and X-ray crystallographic studies (19, 20) of dynamin and its domains (21–24) have allowed a better understanding of the mechanisms mediating dynamin oligomerization. It appears that the stable dimers formed by the crossed interaction between the "stalk" regions of monomeric dynamins (20, 25) are the basic unit that allows dynamin polymerization (18), thus promoting the GTPase activation required for membrane remodeling and scission in different cellular processes.

In the present review, we discuss the different roles of dynamin during endocytosis, vesicle trafficking, and exocytosis, specially focusing in dynamin-2, and how disease-linked mutations in dynamin-2 gene alter such cellular processes.

#### **DYNAMIN AS A KEY COMPONENT OF ENDOCYTOSIS AND VESICLE RECYCLING**

Dynamin is a GTPase that plays a crucial role in the recycling of secretory vesicle in neuroendocrine cells (26).

The first evidence suggesting a role for dynamin in endocytosis came from the mapping and characterization of the *Drosophila shibire* gene, which was identified to be the *Drosophila* homolog of mammal dynamin (27, 28). *Drosophila* bearing mutations in the *shibire* gene exhibited a rapid and reversible paralysis at temperatures exceeding 29°C (29). The first ultrastructural analyses of synaptic terminals of *shibire* mutants showed a decreased number of synaptic vesicles and accumulation of "collared pits" suggesting a blocked step in the endocytotic process (30, 31). These ultrastructural changes were also observed in garland cells, a type of cell considered to be very active in endocytosis, where horseradish peroxidase uptake activity was also reduced, thus confirming an alteration in endocytosis in the *shibire* mutants (32). The role of dynamin in endocytosis in mammalian cells was later demonstrated using dynamin mutants with reduced GTPase activity (33, 34). The use of the non-hydrolyzable GTP analog GTP-γ-S allowed the visualization of endocytotic pits with elongated necks, decorated by electron-dense rings positive for dynamin immunoreactivity, showing that dynamin oligomerizes around the neck of endocytotic pits (35). Moreover, the fact that the dynamin mutant K44A, defective in GTP binding and hydrolysis, specifically blocked the coated vesicle formation without affecting the coat assembly and invagination revealed that dynamin is required for the constriction and subsequent budding of the coated vesicles (36). However, later studies suggested that dynamin also plays a role during a pre-collar stage, when the clathrin lattice is growing (37, 38). In this stage, dynamin may function as a scaffolding molecule that interacts with SH3 domain-containing proteins that control coated pit assembly and maturation (37). The recruitment of dynamin at endocytosis sites (39) also depends on its interaction with SH3-domain-containing proteins and phosphoinositides present at the plasma membrane via its PRD (40) and PH domain, respectively (41).

Regarding the mechanism by which dynamin catalyzes membrane fission, it has been proposed that the assembly of dynamin into helical oligomers around the neck of clathrin-coated pit promotes the dimerization of G domains of adjacent helical rungs, leading to the hydrolysis of GTP (18). The GTP hydrolysis triggers

a conformational change in the dynamin polymer, allowing the constriction of the dynamin ring (18, 42, 43). The ring constriction strength then drives the constriction of the membrane neck, increasing membrane curvature (44). Such change in membrane curvature raises the local elastic energy, reducing the energy barrier to fission, and subsequently triggering the spontaneous fission at the boundary between the dynamin ring and the bare membranes (44).

The ability of dynamin to catalyze membrane fission is not only required in CME, but is also needed in other types of endocytotic pathways that are independent of clathrin. For instance, dynamin is required for caveolae-mediated endocytosis (45), which is essential for the uptake of molecules such as the complement C5b-9 complex (46). Dynamin also participates in the internalization of the β-chain of the interleukin-2 receptor through a clathrin independent, but RhoA dependent process, which is inhibited by the overexpression of a dominant negative mutant of dynamin GTPase activity (47). The entry in host cells of many pathogens and toxins such as anthrax toxin (48), Ebola virus (49), HIV (50), or Hepatitis C virus (51) also require the participation of dynamin.

Finally, the role of dynamin in vesicle formation is not only restricted to the plasma membrane and, as discussed below; its function is also needed in intracellular compartments.

#### **ROLE OF DYNAMIN IN VESICLE TRAFFICKING AND GOLGI FUNCTION**

Dynamin-mediated vesicle budding and membrane fission has also been reported in intracellular compartments such as endosomes (52, 53) and Golgi complex (54). Regarding dynamin-2 participation in vesicle trafficking from endosomes, this protein appears to play a role in two different steps: (1) the vesicle transport from late endosomes to the Golgi compartment (52) and (2) the recycling pathways from early endosomes (53). However, the establishment of dynamin-2 participation in the post-Golgi vesicle trafficking has been more controversial. Pioneer reports showed that the transport of vesicles from the Golgi to the cell surface or to lysosomes was independent of dynamin (36), and according to these results, no evidence of endogenous dynamin-2

localization in the Trans-Golgi network (TGN) nor of its participation in vesicle formation from this compartment were observed using different cell lines (55). Nevertheless, contemporary studies showed that ectopically expressed dynamin-2 localizes in the TGN in hepatocytes (54) and that the formation of clathrin-coated pits from Golgi membranes, in a cell-free assay, was inhibited in the presence of anti-dynamin antibodies, thus indicating the importance of dynamin at this level (54). Furthermore, canine kidney cells expressing a GTPase defective dynamin-2 mutant showed a restricted traffic of the protein p75 from the Golgi to the apical membrane (56). Also, the overexpression of a dominant negative mutant of dynamin-2 was shown to lead to the retention of proteins and accumulation of cisternae at the Golgi compartment, suggesting a role of dynamin-2 in keeping both the structure and function of the TGN (57).

In agreement with a role of dynamin-2 in Golgi vesicle formation, it was demonstrated that endogenous dynamin-2 localizes in the TGN in neuroendocrine mouse pituitary corticotrope cells (58), where it interacts with the βγ subunit of G-proteins via its PH domain. Interestingly, the overexpression of the purified PH domain induced the translocation of dynamin-2 from the Golgi complex to the plasma membrane, increasing receptor-mediated endocytosis but inhibiting basal and CRH-induced secretion of β-endorphins, suggesting a key role of dynamin-2 in the secretory pathway (58). Several studies have highlighted the importance of dynamin-2 to the proper traffic of nascent proteins from the TGN to the plasma membrane, a process that seems to be dependent on the actin cytoskeleton. In this regard, a subset of actin filaments anchored to the Golgi membrane via the small GTPase Arf-1 appears to form complexes with dynamin-2 and the actinbinding-protein cortactin, allowing the emergence and post-Golgi trafficking of secretory vesicles (59). Other proteins such as LimK1 and its substrate cofilin (60), syndapin, and the Wiskott–Aldrich-Syndrome-protein (WASP) (61) have been also involved in the regulation of the peri-Golgi actin cytoskeleton and in the dynaminmediated transport of secretory vesicles from Golgi to the plasma membrane. Additionally, dynamin-2 function is necessary for the Golgi fragmentation and vesicle segregation induced by cholesterol in HeLa cells (62) and for the Golgi vesiculation induced by the c-SRC kinase activation (63) further supporting a pivotal role of dynamin in the Golgi dynamics along the secretory pathway.

#### **DYNAMIN AS A FACILITATOR OF MEMBRANE FUSION**

Dynamins have been involved in different types of fusion processes. Among them regulated exocytosis in neuroendocrine cells (64–69), acrosomal reaction (70), cell-to-cell fusion (71, 72), and fusion of virus with host cells (73, 74).

The first evidences showing the involvement of dynamin in exocytosis came from experiments performed at the beginning of 2000, which suggested that dynamin was involved in kiss-and-run, a transient mode of exocytosis, in neuroendocrine cells (64–66). In this type of exocytosis, the vesicle partially releases its content, and then it is recovered intact (75, 76). It was then hypothesized that dynamin would allow the reclosure of the vesicle mouth, acting through a mechanism similar to that described for vesicle formation during endocytosis (77). However, the mechanism by which dynamin controls the quantal release of hormones still

remains unclear. More intriguing is the mechanism by which dynamin facilitates the fusion process during exocytosis. For instance, dynamin-2 reportedly favors granule secretion in both natural killer (78) and insulin-secreting cells (79). More recently, it was reported that, in chromaffin cells, dynamin-1 speeds up the expansion of the fusion pore, an intermediate structure formed during exocytosis, in a GTPase activity-dependent fashion (68). Therefore, both dynamin-1 and -2 appear to be involved in fusion processes. A possible explanation for this function is that dynamin interacts with SNARE proteins or SNARE-interacting proteins. In this regard, dynamin-2 has been shown to associate to secretory granules in chromaffin cells (80) via its interaction with syntaxin (80) and synaptophysin (67). In mammalian sperm, dynamin-2 associates with the SNARE regulatory protein complexin I (81), where it favors membrane fusion events during acrosomal exocytosis (70). In yeast, the dynamin homolog Vps1p interacts with the t-SNARE Vamp3; the disruption of this association with an antibody against Vps1p inhibits the fusion reaction (82). More recently, Peters' lab has demonstrated that Vps1 binds to the Qa SNARE Vamp3 and controls trans-SNARE formation, which is essential for membrane fusion in yeasts (83).

An alternative explanation is that dynamin controls fusion events via actin cytoskeleton dynamics. In this regard, cortical actin is dynamically rearranged during regulated exocytosis in neuroendocrine cells acting as a barrier as well as a carrier for the access of the secretory granules to the plasma membrane (84– 86). Actin filaments also control the fusion pore expansion (87), as dynamin does (68). As we discuss below, given that dynamin regulates actin organization (13, 14, 88), it is plausible to assume that its actions on fusion processes relay on its ability to modulate actin dynamics. According to this idea, we have recently demonstrated that endogenous dynamin-2 directs a Ca2+-dependent polymerization of cortical actin in adrenal chromaffin cells. Interestingly, both cortical actin and dynamin-2 regulate the initial fusion pore expansion and quantal release of transmitters, suggesting a synergistic action during exocytosis (89). Corey Smith and Collaborators found that the fusion pore expansion in chromaffin cells is controlled by the association of dynamin-1 with syndapin (69), a protein that modulates actin polymerization through neural-Wiskott–Aldrich-syndrome-protein (N-WASP) (90).

The role of dynamin-2 in fusion processes has also been extended to the fusion of myoblasts to form multinucleated myotubes (71). Interestingly, the GTPase activity of dynamin-2 is required at a stage that follows hemifusion but precedes the expansion of the fusion pores (71). The underlying mechanism is still unclear, but it could explain the muscular dysfunction in centronuclear myopathies caused by dynamin-2 mutations. Dynamin-2 has also been involved in the cell-to-cell fusion triggered by HIV-1 virus infection (72) and in fusion between HIV-1 virus and endosomes (73, 74). The latter authors proposed that dynamin promotes the expansion of the fusion pore that connects the HIV-1 envelope and the endosomal membrane, but the possible mechanism remains to be clarified.

Taken together, these findings indicate a pleiotropic role of dynamin in membrane fusion. Although dynamin does not promote membrane fusion by itself, it seems to act after a hemifusion state, facilitating the expansion of the fusion pore (71). The underlying mechanism probably relies on dynamin ability to sense membrane curvature and remodel membranes (91).

#### **DYNAMIN AS A REGULATOR OF MICROTUBULE INSTABILITY**

Although dynamin was first identified as a microtubulesassociated protein (1) its specific role in microtubule integrity and dynamics is still unclear. Using papain-digestion experiments *in vitro*, Herskovits and Co-workers observed that the entire PRD constituted a microtubule-binding site (34). Later, it was demonstrated that only the PRD amino-terminal region was necessary for the association of dynamin to microtubules, whereas its C-terminus appeared to negatively regulate this interaction (92). Pioneering studies also described that dynamin polymerized around microtubules, interconnecting them and allowing bundle formation (1), and in turn, microtubules seemed to stimulate dynamin GTPase activity (12). Other findings showed that the middle domain of dynamin-2 binds to γ-tubulin (93) locating dynamin-2 to the centrosome, therefore suggesting its participation during centrosome cohesion (93). More recently, dynamin-2 was found to be enriched in microtubule bundles at the mitotic spindle of mitotic cells, playing a key role during the cell cycle progression (94). In 2009, Tanabe and Takei observed that depletion of dynamin-2 in COS-7 cells, led to an abnormal accumulation of stable microtubules, consequently inducing disturbance of the microtubules-dependent membrane trafficking. These authors did not find alterations in microtubule assembly in cells that expressed a specific interfering RNA against dynamin-2, but they did report increased stability of pre-existing microtubules, suggesting a role of dynamin-2 in microtubule dynamics instability (95). This dynamin role appears to be independent of its GTPase activity. On the other hand, a deletion of three amino acid residues located at the β3/β4 loops of the PH domain induced decoration of microtubules and accumulation of acetylated tubulin, indicating that this region is required for the correct bundling of microtubules (95). The authors also proposed that this action of dynamin-2 on microtubule instability is necessary for a proper formation of Golgi network and vesicle trafficking.

#### **DYNAMIN AS A DIRECTOR OF ACTIN CYTOSKELETON DYNAMICS**

Dynamin-2 is recruited to different actin-rich structures such as phagocytic cups (96), podosomes (97), *Listeria* actin (98), lamellipodia (99), and filopodia (88), supporting a connection between dynamin-2 function and actin cytoskeleton polymerization. Moreover, both actin and dynamin-2 have regulatory functions during T cells activation (100), phagocytosis (101), and clathrin-dependent endocytosis (102). In the latter, actin and dynamin-2 have been shown to exhibit a synergistic action, where one modulates the recruitment of the other, and both participate in membrane remodeling and scission (103, 104).

In all these processes, dynamin appears to promote actin assembly in a manner dependent on its GTPase activity (14). The mechanism is still unclear, yet it is probably a consequence of its association with several nucleation promoting factors (NPF) via a PRD-SH3 interaction; among them Abp1 (105), the N-WASP (106), and cortactin (13, 88), all of which have been described as dynamin-partners during actin polymerization.

In 2010, Gu and Co-workers reported a direct interaction between dynamin and short actin filaments (F-actin). The latter promoted dynamin oligomerization, and in turn dynamin induced F-actin elongation (14). These authors suggested a model wherein dynamin would favor actin polymerization by removing the "caping"-protein gelsolin of the "barbed" ends (14). More recently, Yamada et al. (88) demonstrated that dynamin induces actin bundle stabilization in a way dependent on its association with cortactin in growth cone filopodia, where they seem to form ring-structures along actin filaments that stabilize F-actin bundles (88). Since dynamin oligomerization enhances its catalytic activity (14), its association with cortactin during actin polymerization fits better with its role as a GTPase than that of a mere "uncapping" protein.

#### **IMPACT OF DISEASE-RELATED DYNAMIN-2 MUTATIONS IN VESICLE TRAFFICKING AND ENDOCYTOSIS**

Specific missense mutations and short deletions into the structural domains of dynamin-2 have been associated with two congenital autosomic neuromuscular disorders: Charcot–Marie–Tooth neuropathy (CMT) (9) and centronuclear myopathy (CNM) (107). While CMT is a demyelinating disease affecting peripheral nerves (108), CNM is characterized by a progressive weakness and atrophy of distal muscles, usually involving facial and extraocular musculature (109–111). Although most dynamin-2 mutations linked to CMT and CNM are located in the middle and PH domains, there are no common mutations to both disorders (112) and the molecular mechanisms that lead to the neuropathy or myopathy remain obscure. **Figure 1** shows the localization of the CMT and CNM dynamin-2 mutations described until now, and **Table 1** summarizes their impacts on different cellular processes.

Regarding how these disease-related mutations affect dynamin-2 oligomerization and catalytic properties, *in vitro* studies have revealed that the CNM-linked mutations E368K/Q, R369Q/W, and R465W located into the middle domain of dynamin-2 (107) as well as the mutations A618T, S619L/W, and V625del clustered into the PH domain (118) exhibit an increased stability of the oligomers and an enhanced GTPase activity (119, 120). However, how these and other dynamin-2 mutations impact on dynamin-dependent cellular processes is still under discussion.

Given that the best-known role of dynamin-2 is to catalyze membrane scission during endocytosis, most of the studies regarding the mechanism underlying CNM associated to mutations in dynamin-2 have been focused in this cellular process. However, the findings are contradictory. The overexpression of the CNM-linked mutants R465W, R522H, S619L, P627H V625del, and E650K impaired CME in COS-1 and COS-7 cells (112, 115). The fact that the CNM mutant D614N causes intracellular mislocalization of both dynamin-2 and clathrin (121) suggests that an impairment in CME could be a consequence of the anomalous distribution of dynamin-2 and other endocytotic proteins. In opposition to the aforementioned findings, fibroblasts from patients harboring the mutations R465W or S619L reportedly display normal CME (112). In a similar manner, embryonic fibroblasts from heterozygous knock-in (KI) mice carrying the mutation R465W exhibited no defects in CME (113). Unlike the heterozygous R465W KI mice, the homozygous animals displayed impaired receptor-mediated

#### **Table 1 | Cellular function alterations induced by CNM and CMT mutants.**


endocytosis, indicating that this mutation differentially impacts on CME according to the heterozygous or homozygous carrier state. In this regard, fibroblasts from patients harboring the homozygous mutation F379V, the first one linked to a lethal congenital syndrome, also exhibited dysfunctional endocytosis (6). In order to compare the effects of dynamin-2 mutations associated to CNM in a context that mimics the homozygous and heterozygous states, Schmid and Collaborators (114) established stable cell lines from dynamin-2 conditional null mouse fibroblasts and expressed two CNM-linked dynamin-2 mutants (E368K and R465W). Cells expressing comparable levels of wildtype dynamin-2 and the given mutant were selected to evaluate clathrin-dependent and independent endocytosis, and also vesicle trafficking. The authors showed that none of these mutations affected CME in the condition that mimics the heterozygous state; and they were able to fully rescue transferrin uptake in dynamin-2 KO cells. Nevertheless, all these mutants impaired clathrin-independent endocytosis of epidermal grow factor receptors and raft-dependent endocytosis of cholera toxin. Furthermore, also in the condition that mimics the heterozygous state, all the mutants disrupted trafficking of p75/neurotrophin receptor from Golgi to plasma membrane. It seems therefore plausible that the physiological dysfunctions in CNM patients harboring heterozygous mutations in dynamin-2 are not a consequence of defective CME, but other dynamin-2-dependent processes, such as clathrin-independent endocytosis or vesicle trafficking could be affected in these conditions.

Regarding the mechanism underlying CMT associated to dynamin-2 mutations, Sidiropoulos and Co-workers (116) suggested that an impairment of CME in Schwanm cells plays an important role in the pathogenesis of CMT neuropathy. They showed that the CMT mutants K562E, G358R, G537C, and L570H, but not the CNM mutants R465W and E560K, strongly reduced myelination in dorsal root ganglia explant cultures derived from heterozygous embryos carrying a dynamin-2 null allele, thus

mimicking the heterozygous state. These CMT mutants also reduced transferrin internalization in Schwann cells and motor neurons. Conversely, none of the 11 CNM mutations evaluated in that work had any effect on transferrin internalization, neither in Schwann cells nor in motor neurons, suggesting that CMT and CNM mutations affect different cellular processes.

The aforementioned cellular impairments induced by CNMand CMT-mutations, as well as others not previously mentioned, are summarized in the **Table 1**.

#### **CONCLUDING REMARKS**

The secretory pathway plays a pivotal role in mammalian cell function. Its disruption underlies many known diseases, such as those described herein. Dynamin-2 is critical in such mechanisms, as it regulates Golgi structure and function, cytoskeletal integrity, vesicle trafficking, and most interestingly, membrane fusion.

In the last 10 years, more than 20 disease-linked mutations have been reported in the dynamin-2 gene, affecting different dynamin-mediated cellular processes. Nevertheless, their impact on neuroendocrine tissues remains obscure. Also elusive are the mechanisms by which the said mutations affect actin cytoskeleton dynamics and exocytosis, processes which are dynamin-dependent and critical in the course of the secretory pathway. The latter opens an interesting research field that could prove useful to understand the mechanisms involved in the pathogenesis of dynamin-2 disease-related mutations.

Several diseases caused by genetic modifications remain as the final frontier in medicine, and although pivotal hallmarks such as the sequencing of the human genome have been attained, we still have much to learn about the cellular and molecular consequences of diverse gene mutations, specially, those that are common in many tissues. In this regard, dynamin-2 is a ubiquitous protein with diverse pleiotropic roles, whose mutations cause tissue-specific phenotypes. The existence of pertinent cell and animal models will be extremely useful for better understanding how dynamin-2 mutations impact on its properties and related functions, and consequently in the identification of potential therapeutic targets.

#### **REFERENCES**


#### **ACKNOWLEDGMENTS**

Fondecyt 1110552 (Chile); Proyecto Anillo ACT-1121 (CONI-CYT); Fondos ICM-ECONOMIA Project P09-022-F (Chile).

et al. Crystal structure of nucleotide-free dynamin. *Nature* (2011) **477**:556–60. doi:10.1038/nature10369


(1973) **120**:107–14. doi:10.1007/ BF00267238


Chandran K. The Ebola virus glycoprotein mediates entry via a non-classical dynamindependent macropinocytic pathway. *Virology* (2011) **419**:72–83. doi:10.1016/j.virol.2011.08.009


MJ, Martínez AD, et al. The association of dynamin with synaptophysin regulates quantal size and duration of exocytotic events in chromaffin cells. *J Neurosci* (2010) **30**:10683–91. doi:10.1523/ JNEUROSCI.5210-09.2010


chromaffin cells. *J Physiol* (2004) **560**:413–28. doi:10.1113/jphysiol. 2004.064410


myosin II accelerate catecholamine release from chromaffin granules. *J Neurosci* (2009) **29**:863–70. doi:10. 1523/JNEUROSCI.2818-08.2009


*J Exp Med* (1999) **190**:1849–56. doi:10.1084/jem.190.12.1849


nexin 9 interacts with dynamin 1 and N-WASP and coordinates synaptic vesicle endocytosis. *J Biol Chem* (2007) **282**:28939–50. doi: 10.1074/jbc.M700283200


(2007) **62**:666–70. doi:10.1002/ ana.21235


gene impairs dynamin 2 localization in skeletal muscle of a patient with late onset centronuclear myopathy. *Neuromuscul Disord* (2013) **23**:219–28. doi:10.1016/j.nmd.2012.12.007

**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: 29 May 2013; accepted: 31 August 2013; published online: 18 September 2013.*

*Citation: González-Jamett AM, Momboisse F, Haro-Acuña V, Bevilacqua JA, Caviedes P and Cárdenas AM* *(2013) Dynamin-2 function and dysfunction along the secretory pathway. Front. Endocrinol. 4:126. doi: 10.3389/fendo.2013.00126*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

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

# GnRH-induced Ca2<sup>+</sup> signaling patterns and gonadotropin secretion in pituitary gonadotrophs. Functional adaptations to both ordinary and extraordinary physiological demands

#### **Maria Luisa Durán-Pastén<sup>1</sup>† and Tatiana Fiordelisio<sup>2</sup>\* †**

<sup>1</sup> Departamento de Neurociencia Cognitiva, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México (UNAM), México DF, México <sup>2</sup> Departamento de Ecología y Recursos Naturales, Facultad de Ciencias, Universidad Nacional Autónoma de México (UNAM), México DF, México

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Gustavo M. Somoza, Instituto de Investigaciones Biotecnologicas-Instituto Tecnologico de Chascomus, Argentina Stanko S. Stojilkovic, National Institutes of Health, USA

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

Tatiana Fiordelisio, Departamento de Ecología y Recursos Naturales, Facultad de Ciencias, Universidad Nacional Autónoma de México (UNAM), Circuito exterior s/n. Ciudad Universitaria, C.P. 04510 México DF, México

e-mail: tfiorde@ciencias.unam.mx

†Maria Luisa Durán-Pastén and Tatiana Fiordelisio have contributed equally to this work.

Pituitary gonadotrophs are a small fraction of the anterior pituitary population, yet they synthesize gonadotropins: luteinizing (LH) and follicle-stimulating (FSH), essential for gametogenesis and steroidogenesis. LH is secreted via a regulated pathway while FSH release is mostly constitutive and controlled by synthesis. Although gonadotrophs fire action potentials spontaneously, the intracellular Ca2<sup>+</sup> rises produced do not influence secretion, which is mainly driven by Gonadotropin-Releasing Hormone (GnRH), a decapeptide synthesized in the hypothalamus and released in a pulsatile manner into the hypophyseal portal circulation. GnRH binding to G-protein-coupled receptors triggers Ca2<sup>+</sup> mobilization from InsP3-sensitive intracellular pools, generating the global Ca2<sup>+</sup> elevations necessary for secretion. Ca2<sup>+</sup> signaling responses to increasing (GnRH) vary in stereotyped fashion from subthreshold to baseline spiking (oscillatory), to biphasic (spike-oscillatory or spikeplateau). This progression varies somewhat in gonadotrophs from different species and biological preparations. Both baseline spiking and biphasic GnRH-induced Ca2<sup>+</sup> signals control LH/FSH synthesis and exocytosis. Estradiol and testosterone regulate gonadotropin secretion through feedback mechanisms, while FSH synthesis and release are influenced by activin, inhibin, and follistatin. Adaptation to physiological events like the estrous cycle, involves changes in GnRH sensitivity and LH/FSH synthesis: in proestrus, estradiol feedback regulation abruptly changes from negative to positive, causing the pre-ovulatory LH surge. Similarly, when testosterone levels drop after orquiectomy the lack of negative feedback on pituitary and hypothalamus boosts both GnRH and LH secretion, gonadotrophs GnRH sensitivity increases, and Ca2<sup>+</sup> signaling patterns change. In addition, gonadotrophs proliferate and grow. These plastic changes denote a more vigorous functional adaptation in response to an extraordinary functional demand.

**Keywords: pituitary, gonadotrophs, calcium, gonadotropins, GnRH, secretion**

#### **GONADOTROPHS FUNCTION AND CHARACTERISTICS**

The reproductive function and sexual maturation is under the control of the hypothalamic-pituitary-gonadal axis. Pituitary gonadotrophs, which constitute 7–15% of the anterior pituitary gland secrete two dimeric glycoproteins, gonadotropins, luteinizing (LH) and follicle-stimulating (FSH) hormones that play an essential role in the control of steroidogenesis, gametogenesis, and ovulation (1). The regulation of their synthesis and secretion are under control of hypothalamic stimulation (gonadotropin-releasing hormone; GnRH), gonadal sex steroids (estradiol, progesterone, testosterone) and peptides (inhibins), and paracrine factors (inhibins, activins, and follistatin). The pituitary gland must adapt to different physiological changes from prepubertal to mature sexual life, therefore gonadotrophs plasticity and gonadotropins secretion are essential to produce the changes needed in different situations, for example the rapid daily hormonal variations along the reproductive female cycle. Integration of the different regulatory signals by the gonadotrophs results in the coordinated control of synthesis, packaging, and differential secretion of gonadotropins to accurately

respond and control sexual maturation and normal reproductive function.

Immunocytochemical studies have demonstrated the presence of bihormonal (70%) and monohormonal (15%) gonadotrophs whose percentage shifts under different physiological conditions, such as castration or estrous cycle (2). LH and FSH have a common alpha (α) and distinct beta (β) subunit. After its synthesis in the endoplasmic reticulum (ER) and its passage trough the Golgi apparatus, hormones are delivered to the plasma membrane trough a constitutively or regulated secretory pathway; in the latter, fusion of secretory vesicles to the plasma membrane is arrested waiting for specific signals to be secreted. Gonadotropin synthesis and secretion diverges under a range of physiological and experimental conditions (3), indicating that GnRH and other regulators of gonadotropins selectively activate this pathways.

Exocytosis in excitable cells is a process highly dependent of intracellular calcium concentration ([Ca2+]i) rise, gonadotrophs as other pituitary endocrine cells display spontaneous intracellular Ca2<sup>+</sup> transients in dependence of changes in the membrane electrical activity. However, this membrane potential oscillations are small and do not produce the necessary [Ca2+]<sup>i</sup> increase to generate hormonal secretion (4, 5), as a result, basal secretion is low and not affected by extracellular Ca2<sup>+</sup> (4, 6). In both cases, the principal regulation is done by GnRH, a decapeptide that is synthesized in the hypothalamus, stored in axon terminals in the median eminence, and released in a pulsatile manner into the hypophyseal portal circulation (7). Numerous studies have shown that isolated gonadotrophs in primary culture (and more recently, also gonadotrophs *in situ*) present robust and stereotyped dosedependent intracellular Ca2<sup>+</sup> signals in response to suprathreshold concentrations of GnRH (8–11), the rise produced in cytosolic [Ca2+] triggers gonadotropins exocytosis and synthesis.

Understanding the origin and meaning of these intracellular Ca2<sup>+</sup> signals are essential to the knowledge of the physiology of normal reproduction, as well as reproductive function disorders. This review outlines different regulators of the gonadotrophs biology with special regard in the recent progress on GnRH-induced Ca2<sup>+</sup> signaling and secretion in pituitary gonadotrophs, both at the cellular and tissue level.

#### **Ca<sup>2</sup>**<sup>+</sup> **SIGNALS INDUCED BY GnRH AND OTHER SECRETAGOGUES**

In order to mediate multiple effects such as secretion, synthesis, and phenotype maintenance, the GnRH variants in different species interact with their receptor (GnRHR), which is a member of the rhodopsin-like G-protein-coupled receptors (GPCR) superfamily (12). Upon GnRH binding to the GnRHRs in the gonadotroph membrane, the α subunit of the Gq/<sup>11</sup> protein dissociates and activates phospholipase C (PLC-β), resulting in the rapid hydrolysis of phosphatidylinositol 4, 5-biphosphate (PIP2) and the production of two second messengers: diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (InsP3); long lasting GnRH stimulation (∼5–10 min) could also activate phospholipase D (PLD) and phospholipase A<sup>2</sup> (PLA2) (12). InsP<sup>3</sup> generates Ca2<sup>+</sup> mobilization from intracellular pools, and DAG triggers protein kinase C (PKC) activation which in turn reduces depolarizationmediated Ca2<sup>+</sup> influx, while increasing gonadotropin secretion (13) (**Figure 1**). PKC sensitizes the secretory machinery to Ca2<sup>+</sup> (14), which explain why GnRH application is more effective to induce secretion than membrane depolarization or caged Ca2<sup>+</sup> photolysis (5). PKC activation is also involved in other exocytosisassociated processes, like GnRH self-priming and cytoskeletal rearrangement (3).

In the lumen of the ER, [Ca2+] is maintained higher (between 10 and 250µM free) than in the cytosol (50–250 nM) by the pumping activity of the sarco-ER Ca2<sup>+</sup> ATPase (SERCA) located in the ER membrane (15). This membrane holds intracellular channel that allow Ca2<sup>+</sup> efflux from the ER down its concentration gradient; the InsP<sup>3</sup> receptor (InsP3R), a ligand-gated Ca2<sup>+</sup> channel

**FIGURE 1 | Schematic representation of a gonadotroph illustrating the main control pathways of gonadotropin synthesis and secretion**. GnRH, gonadotropin-releasing hormone; Gq, protein Gq/11; PLCβ, phospholipase C; PIP2, phosphatidylinositol 4,5 bisphosphate; DAG, diacylglycerol; IP3, inositol 1,4,5-trisphosphate; PKC, protein kinase C; VGCC, voltage-gated calcium channels; CaMKII, calcium calmodulin type II

kinase; RE pump, endoplasmic reticulum Ca<sup>2</sup><sup>+</sup> pump; PM pump, plasma membrane Ca<sup>2</sup><sup>+</sup> pump; SK, small conductance calcium-activated potassium channels; P, progesterone; PR, progesterone receptor; E, estradiol; ERα, estrogen receptor α; T, testosterone; AR, androgen receptor; Raf, serine/threonine kinase; MEK, mitogen-activated protein kinase; ERK, extracellular-signal-regulated kinases.

that opens after InsP<sup>3</sup> binding (16). Besides InsP<sup>3</sup> binding, Ca2<sup>+</sup> interaction with high-affinity (activation) sites on the cytoplasmic side of the InsP3R is essential for channel opening. In fact, Ca2<sup>+</sup> and InsP<sup>3</sup> operate as co-agonists. Ca2<sup>+</sup> signal amplification and spreading phenomena, involving assemblies of InsP3Rs originate from this synergistic role of Ca2<sup>+</sup> (17). The large and abrupt [Ca2+]<sup>i</sup> increase, triggered by InsP3Rs activation results from the combination of Ca2<sup>+</sup> released and its amplification by Ca2<sup>+</sup> induced Ca2<sup>+</sup> release [CICR; (18)]. Even if cytosolic InsP<sup>3</sup> remains high,Ca2<sup>+</sup> efflux often ceases because Ca2<sup>+</sup> binds to a lowaffinity (inactivating) site of the receptor, which closes the InsP3R channel. This occurs when cytosolic Ca2<sup>+</sup> close to the InsP3Rs is high, i.e., after an episode of fast release. As in most pituitary cells, agonist stimulation in gonadotrophs produce a [Ca2+]<sup>i</sup> peak which decays to sustained Ca2<sup>+</sup> level (plateau phase). At intermediate GnRH concentration the initial Ca2<sup>+</sup> spike is often followed by large [Ca2+]<sup>i</sup> oscillations resulting from opening and closing cycles of the InsP3R channels as a consequence of [Ca2+]<sup>i</sup> fluctuations near its cytoplasmic side (19). The frequency of these Ca2<sup>+</sup> oscillations is determined by the dose of GnRH applied and the intracellular (InsP3) reached (20) (**Figure 1**).

Gonadotropin-releasing hormone-induced [Ca2+]<sup>i</sup> oscillations can be reproduced with mathematical models that include a Ca2<sup>+</sup> gradient between the ER lumen and the cytosol maintained by a SERCA Ca2<sup>+</sup> pump, Ca2<sup>+</sup> influx trough voltage-gated Ca2<sup>+</sup> channels, and InsP3R channels co-activated by InsP<sup>3</sup> and low [Ca2+]<sup>i</sup> , and inactivated by high [Ca2+]<sup>i</sup> (8, 15, 21, 22). Nonetheless, Ca2<sup>+</sup> oscillations in real cells requires the precise coordination of Ca2<sup>+</sup> mobilization/uptake/extrusion mechanisms, it is for it that immortalized gonadotroph cell lines αT3–1 (21) and LβT2 (23) are not good cell models for studies on GnRH-induced calcium signaling and modulation of voltage-gated calcium influx, as well as goldfish (24, 25) and immature mammalian gonadotrophs, since these cells respond to GnRH with non-oscillatory amplitudemodulated Ca2<sup>+</sup> signals. When SERCA pumps in gonadotrophs are blocked by thapsigargin, the agonist-induced Ca2<sup>+</sup> oscillations become non-oscillatory biphasic responses (8, 26). Therefore different factors, i.e., the amount and speed of InsP<sup>3</sup> production, the total number of InsP3R channels available for activation, the rate of Ca2<sup>+</sup> leakage from the store and the efficiency of the SERCA Ca2<sup>+</sup> pump vary from cell to cell, and they ultimately determine the characteristics of gonadotrophs Ca2<sup>+</sup> signaling patterns. It is important to note that the oscillatory behavior is intrinsic to the Ca2<sup>+</sup> handling properties of gonadotrophs (17).

Gonadotropin-releasing hormone produces Ca2<sup>+</sup> oscillations: i.e., large Ca2<sup>+</sup> spikes, arising from a flat baseline as well as smaller sinusoidal Ca2<sup>+</sup> oscillations superimposed on an elevated plateau. Under sustained GnRH stimulation, the amplitude of these Ca2<sup>+</sup> spikes gradually diminishes, probably due to intracellular Ca2<sup>+</sup> pool depletion, until a "plateau" without oscillations is reached. Ca2<sup>+</sup> influx through voltage-gated Ca2<sup>+</sup> channels is essential to maintain this plateau, and also for the replenishment of intracellular Ca2<sup>+</sup> pools. GnRH induces continuous AP firing periodically interrupted by hyperpolarizations, which occur in phase with each Ca2<sup>+</sup> elevation, and resulting from the opening of Ca2<sup>+</sup> dependent SK-type K<sup>+</sup> channels (6, 27). Immediately after each hyperpolarization, the cell fires a burst of APs, which open Ca2<sup>+</sup> channels allowing Ca2<sup>+</sup> influx, predominantly

high-voltage – activated L-type calcium channels. This Ca2<sup>+</sup> entry does not contribute to Ca2<sup>+</sup> elevation or gonadotropin secretion, but is crucial for refilling the intracellular Ca2<sup>+</sup> pools (20) (**Figure 1**).

Oscillatory Ca2<sup>+</sup> signals in gonadotrophs can also be elicited by endothelin (ET) (28, 29), pituitary adenylate cyclase-activating polypeptide, (PACAP) (30), and substance P (SP) (31). Conversely, neuropeptide Y (NPY) and melatonin, in neonatal gonadotrophs, inhibit GnRH-induced Ca2<sup>+</sup> signals and gonadotropin secretion. Lactotrophs, gonadotrophs, and somatotrophs produce ETs, and gonadotrophs express ET receptors (32) under the control of ovarian steroid hormones, suggesting a paracrine function. ET binding, leads to Gq/<sup>11</sup> activation, intracellular Ca2<sup>+</sup> fluctuations, and gonadotropin secretion (29). SP, which is a weaker agonist than GnRH, produces amplitude-modulated [Ca2+]<sup>i</sup> responses and secretion in gonadotrophs (31), being the first phase of secretion dependent of intracellular Ca2<sup>+</sup> release, and the second phase Ca2<sup>+</sup> influx-dependent. The hypothalamic factor PACAP which stimulates cAMP production and potentiates gonadotropin release (33), also induces Ca2<sup>+</sup> oscillations in rat gonadotrophs through activation of PVR1, a G-protein-coupled receptor and InsP<sup>3</sup> production (30). The activation of coupled Gi/o melatonin receptors MT1 and MT2, expressed in gonadotrophs only at neonatal stage, inhibits both calcium influx through voltagegated calcium channels and calcium mobilization from intracellular stores, decreasing intracellular cAMP production and protein kinase A (PKA) activity, with a consequent diminution on gonadotropin secretion (34–36); tonic melatonin inhibition of immature gonadotrophs prevents premature initiation of puberty. NPY inhibits GnRH-induced Ca2<sup>+</sup> signaling and LH release (37); its receptors Y1 and Y5 expression on gonadotrophs is regulated by estrogens (38).

#### **Ca<sup>2</sup>**<sup>+</sup> **SIGNALING PATTERNS AND SECRETION IN GONADOTROPHS IS DEPENDENT ON GnRH CONCENTRATION**

Dissociated pituitary gonadotrophs respond to increasing doses of GnRH with a stereotyped progression of intracellular Ca2<sup>+</sup> signaling: i.e., subthreshold GnRH concentrations produce either small monophasic Ca2<sup>+</sup> transients or irregular, small Ca2<sup>+</sup> spikes. With higher GnRH concentrations (0.1–10 nM) regular, oscillatory, frequency-modulated, large Ca2<sup>+</sup> transients (baseline Ca2<sup>+</sup> spiking) are produced. Eventually (∼50–100 nM GnRH), these Ca2<sup>+</sup> spikes fuse into an amplitude-modulated biphasic Ca2<sup>+</sup> response (9, 10, 39) which comprises two variants; biphasic oscillatory and biphasic non-oscillatory, also known as spike-plateau (40). It is reasonable to assume that different Ca2<sup>+</sup> release patterns observed with increasing doses of GnRH underlie the dosedependent increase of gonadotropin secretion. Nonetheless, it has also been suggested that these patterns encode other cell functions. For instance, spike-plateau Ca2<sup>+</sup> responses were associated to LH secretion and oscillatory Ca2<sup>+</sup> responses to the synthesis of LH β-subunits (9). Later, it was established that GnRH-induced Ca2<sup>+</sup> oscillations trigger exocytosis (41) and that both oscillatory and spike-plateau Ca2<sup>+</sup> signals can initiate LH release (10, 40). Furthermore, gonadotrophs do not respond in the same way to the secretagogue: i.e., individual cells can respond with different patterns of activity to the same GnRH concentration

(40). Conversely, when the same dose of GnRH is applied repetitively, individual cells respond with similar latency and signaling pattern (11). It remains to be established which cellular aspects determine the Ca2<sup>+</sup> signals displayed by individual gonadotrophs in response to GnRH and how these different patterns affect gonadotropin synthesis and secretion. Moreover, LH and FSH are secreted through parallel pathways (see below) and hormones that alter their synthesis, release, and/or storage can dynamically regulate their output.

#### **GONADOTROPIN EXOCYTOSIS. CONTRIBUTION OF VGCC-MEDIATED Ca<sup>2</sup>**<sup>+</sup> **INFLUX AND INTRACELLULAR Ca<sup>2</sup>**<sup>+</sup> **RELEASE**

A rise in [Ca2+]<sup>i</sup> is the key signal to trigger regulated exocytosis in neuronal and endocrine tissues. Endocrine cell models used to study the role of Ca2<sup>+</sup> in exocytosis include adrenal chromaffin and PC12 cells (42–46), pancreatic β cells (47–49), and pituitary cells (6, 50). Cytosolic Ca2<sup>+</sup> levels regulate several maturation steps that secretory vesicles must undergo prior to fusion, like priming of secretory vesicles (51). An entirely different phenomena occurs when [Ca2+]<sup>i</sup> rises abruptly, promoting the fusion of docked secretory vesicles with the plasma membrane (47, 52). In contrast to nerve synapses, where Ca2<sup>+</sup> influx is primarily responsible for this abrupt [Ca2+]<sup>i</sup> rise, exocytosis in endocrine cells is triggered to a large extent by Ca2<sup>+</sup> released from intracellular stores (17, 53).

Ca2<sup>+</sup> controls the fusion of secretory vesicles with the plasma membrane to release neurotransmitters and hormones when is needed [regulated exocytosis, (51)]. The first phase of GnRHinduced exocytosis in gonadotrophs is mediated by InsP3-sensitive Ca2<sup>+</sup> pools, while the second "plateau" phase of secretion involves voltage-gated Ca2<sup>+</sup> influx (54). GnRH-InsP<sup>3</sup> induced Ca2<sup>+</sup> oscillations produce much greater exocytosis than the simple general rise in [Ca2+]<sup>i</sup> induced by micropipette injection or uncaging [Ca2+]<sup>i</sup> (5, 41). This suggests that in contrast with other pituitary cell types, the formation of sub-plasmalemmal microdomains of high Ca2<sup>+</sup> in gonadotrophs is insufficient to induce vesicular fusion. Instead large Ca2<sup>+</sup> signals that propagate across the entire cell are needed to accomplish this task (6). Exocytosis can be directly monitored electrically as changes in membrane capacitance due to the addition of new plasma membrane. Using capacitance measurements, exocytosis is detected in gonadotrophs whenever [Ca2+] rises above 300 nM (55), but for strong exocytosis high [Ca2+]<sup>i</sup> with half maximal concentration of 16µM are required (6). When the responses induced by GnRH are oscillatory, step increases in membrane capacitance can be seen in each Ca2<sup>+</sup> spike (40, 41, 55). The first Ca2<sup>+</sup> oscillations elicit the largest exocytosis events, returning to full capacity within about 2 min (5). GnRH-induced secretion continues in the absence of external Ca2+, but ceases when [Ca2+] rises are blocked by the introduction of a strong intracellular Ca2<sup>+</sup> buffer (41).

Secretory granules must undergo a well-defined series of events: (1) recruitment, (2) tethering at the plasma membrane, (3) priming, and (4) vesicle fusion with the plasma membrane. Regulated hormone secretion is a Ca2+-dependent exocytosis that uses the secretory vesicle synaptotagmin as the Ca2<sup>+</sup> sensor and is mediated by SNARE (soluble *N*-ethylmaleimide-sensitive factor attachment protein receptors) proteins as effectors. Syntaxin, SNAP25

(synaptosome-associated protein of 25 kDa in molecular weight), and synaptobrevin (vesicle-associated membrane protein, VAMP, also termed vSNARE) constitute SNARE proteins. Syntaxin and SNAP25 (also known as "target" tSNAREs) are the plasma membrane proteins to whichVAMP couples (**Figure 1**). Then, vSNAREs and tSNAREs form trans-SNARE complexes, which join secretory vesicles and plasma membrane (56–58). Vesicle priming, another Ca2+-dependent step in exocytosis probably involves early SNARE complex formation (particularly tSNARE), before its association to the trans-SNAREs. Finally, synaptotagmin detects the [Ca2+] elevation and provides the extra drive needed to overcome the energy barrier of lipid-to-lipid interaction, allowing membrane fusion (58). The use of high-resolution microscopy techniques have allowed to demonstrate in PC12 cells that tSNARE molecules are distributed on the plasma membrane in areas of low and high density, and in contrast to current models of SNARE-driven membrane fusion (59), this data suggest that secretory vesicles are targeted over areas of low tSNARE density as sites of docking, hence a relatively low number of tSNAREs close to the secretory vesicle (less than seven) are sufficient to drive membrane fusion. Moreover, using atomic forces microscopy and scanning electron microscopy it has been described that gonadotrophs mainly present "single and simple fusion pore" with diameter ranging from 100 to 500 nm, which appear more frequently after stimulation with GnRH; this pore configuration supports the idea of a "kiss and stay" mechanism for the exocytosis process (60), in addition pores of 20–40 nm diameter have also been found, probably representing the constitutive pathway of gonadotropins (60).

#### **FSH AND LH DIFFERENTIAL SECRETION UNDER PHYSIOLOGICAL CONDITIONS**

Along the follicular phase of the estrus cycle, LH secretion is maximal while FSH secretion is reduced; even though gonadotrophs secrete both hormones, the mechanisms underlying this differential release are unclear. FSH appears to be released mostly through the constitutive pathway in accordance to its rate of synthesis. Conversely, LH-containing granules are released through the regulated pathway in response to GnRH,with no effect on LHβ mRNA production (61). Moreover, LH and FSH appear to be packaged into different secretory granules (62). Large, moderately electrondense granules show antigenicity for FSH, LH, and chromogranin A (CgA), while smaller, electron-dense storage granules released by GnRH contain LH and secretogranin II (SgII) (3); thereby protein sorting domains in the β subunit of gonadotropins and the association with certain proteins may be responsible for differential sorting and packaging of LH and FSH into different secretory granules (3). The movement of these granules toward the membrane defining a secretory pathway and differential exocytosis could explain the disparity on the gonadotropins secretion (63). Accordingly, in LβT2 mouse cells, FSH released in response to activin/GnRH is constitutively secreted via a granin-independent pathway; while LH is released in response to GnRH is co-released with SgII via a regulated, granin-dependent pathway (64).

Gonadotropin subunits (α-GSU, FSHβ, and LHβ) mRNAs levels, which reflect changes in gene transcription in pituitary gonadotrophs, are GnRH pulse frequency modulated (65– 67). GnRH pulses (30–60 min interval), preferentially increases synthesis and secretion of LH by the mediation of the transcription factor Egr-1 (68–71); whereas slower GnRH pulsing (120–240 min interval) favors FSH secretion (65–67, 72) by the activation of PKA (73–75). There is no a definitive explanation to how GnRH pulses can activate in a different manner gonadotropin subunit gene transcription; nevertheless several routes have been proposed which may contribute to this regulation; one is through the increase on Ca2<sup>+</sup> levels and PKC activation, which as a consequence activated mitogen-activated protein kinase (MAPK) cascade, culminating in an activation of extracellular-signal-regulated kinase (ERK) 1/2, cJun NH2-terminale kinase (JNK), p38 MAPK, and ERK 5 (76–80), it is also believe that the rise in [Ca2+]<sup>i</sup> , activates a calcium/calmodulin-dependent kinase II (CAMK2), whose autophosphorylation could be important in transmitting Ca2<sup>+</sup> pulse frequency and amplitude signals, as fast and highamplitude Ca2<sup>+</sup> influxes, which results in greater and/or sustained Ca2+/CALM1 levels (79, 81) (**Figure 1**). GnRH pulses at lower frequency selectively increase the expression of PACAP and its receptor (PAC1-R) in gonadotrophs (82), where they subsequently stimulate the synthesis of gonadotropin subunits (83).

Gonadotropin-releasing hormone-induced LH and FSH synthesis and secretion are modulated by steroid hormones, such as estrogen, progesterone, and testosterone, in addition to peptide hormones, such as activin, inhibin, and follistatin (**Figures 1** and **2**). This modulation occurs principally through gonadal feedback at the pituitary and hypothalamus level (84–86). During most part of the female reproductive cycle and in males, pulsatile GnRH release drives tonic gonadotropin secretion (84, 87) while steroids and inhibins provide negative feedback to limit further gonadotropin stimulation and maintaining low circulating levels of gonadotropins; in females, this happens until the pre-ovulatory surge when, in response to low levels of progesterone (88) and an increase in estrogen, feedback switches to positive (89), producing changes on GnRHergic neurons (90) and gonadotrophs (91), which results in increased LH and FSH secretion (**Figure 2**). In some female species a secondary FSH surge occurs after ovulation when LH levels are already low, this rise produces the recruitment of the next cohort of follicles and it is GnRH independent (92) and more likely depends on the reduction of circulating inhibin (93).

Estradiol (E) exerts a direct action at the pituitary level through its α-receptor (94–97), increasing gonadotroph responsiveness to GnRH (98–100) raising synthesis and insertion of GnRH receptor into gonadotroph membrane (86, 91, 101–104) and decreasing the concentration of GnRH needed to produce the threshold response and frequency of Ca2<sup>+</sup> spiking (101, 105, 106). Nevertheless, these actions seems to be and indirect action that depends of the increased expression produced by GnRH on its own receptor (101, 103, 107–109). Besides these changes, during the gonadotropin surge, the pituitary gland shows cellular modifications, implying an augmentation on the number of secreting gonadotrophs (98, 104) and hypertrophy and re-organization of its intracellular organelles (110–112). However, it has been documented that

E can act to suppress the transcriptional rate of LH subunit genes. Controversial results have been reported for FSHβ synthesis (113–117), although serum levels of both hormones increased markedly.

Progesterone (P) exerts some of its effects at hypothalamic level, decreasing GnRH secretion and pulse frequency (91, 103) contributing to the abrupt decline in gonadotropin levels. P does not inhibit LH secretion induced by GnRH (100, 118) but it can stimulate murine FSHβ promoter activity alone or in synergy with activins (103). In dependence with the time of exposition, P can either inhibit or facilitate the estrogen-induced LH surge during the rat estrous cycle (100, 103, 119). P modulates the E effect on GnRH production of LH surge by the modulation of Ca2<sup>+</sup> mobilization and Ca2<sup>+</sup> entry to gonadotrophs. In E-primed cells P alters the intracellular Ca2<sup>+</sup> signaling patterns produced by GnRH. In the short-term P treatment shifts subthreshold [Ca2+]<sup>i</sup> responses to oscillatory, and oscillatory to biphasic responses; in contrast, long-term P exposure led to decreased GnRH sensitivity, changing oscillatory response into subthreshold [Ca2+]<sup>i</sup> response profiles (105, 106).

Androgens [testosterone (T) and 5α-dihydrotestosterone (DHT)] are important component of the male gonadal feedback and they act either at the hypothalamic level by regulating the secretion of GnRH into the hypophyseal portal circulation (120– 122), directly at the pituitary level (99) or by the combination of both sites (123) (**Figure 2**).

At hypothalamic level, T reduce GnRH synthesis (122,124–127) and pulsatile patterns of GnRH release (128–131). At pituitary level, it is known that testosterone and more dramatically DHT inhibits LH synthesis and GnRH-induced LH secretion in a concentration and time dependent manner (132–137), but increase basal FSH secretion and synthesis (138, 139). In castrated rats it has been shown that LH secretion increase (140, 141) as well as gonadotrophs size and number (140, 142, 143). These hypertrophied cells are called castration cells (144–146), and they present a dilated rough ER and an extended Golgi complex (147, 148). On these cells, the secretion granules content are progressively diminished (149) and their cisternae fused to form large vacuoles that originated the typical "signet ring cell" (148, 150, 151).

It is widely accepted that in gonadotrophs an increase in [Ca2+]<sup>i</sup> is essential for the transduction of GnRH signal; T but specially DHT regulate GnRH-induced [Ca2+]<sup>i</sup> variations (152) changing the type of calcium patterns (153), these effects are not seen in all species (145) and it could be related with the influence of this hormone on the regulation of the GnRH receptor density (154–156) and the change in their sensitivity to the GnRH stimulus (134).

Tobin and collaborators (153) demonstrated that in cultured gonadotrophs of gonadectomized male rats, the relationship between GnRH concentration and the type of intracellular Ca2<sup>+</sup> response is altered, most gonadotrophs (∼70%) show oscillatory responses regardless of the GnRH concentration. Correlated with this results it has been demonstrated that in T or DHT treated cells, there is an inhibition of the GnRH increase in [Ca2+]<sup>i</sup> ; at low GnRH doses (0.1 nM) 30% of gonadotrophs were unable to initiated threshold spiking and in the residual cells the frequency of oscillations decreased, as in controls, androgen treated cells, respond with a spike-plateau type of signal to 1 nM GnRH, but

the frequency of spiking was also reduced (134, 152). Finally at high dose GnRH (100 nM) induce biphasic elevations of [Ca2+]<sup>i</sup> with a minor reduction in the amplitude (134). Testosterone inhibits both phases of GnRH-stimulated LH secretory responses, the early extracellular Ca2+-independent spike phase and the sustained Ca2<sup>+</sup> and extracellular Ca2+-dependent plateau phase (134). These results suggest that androgens act on the efficacy of the agonist to release Ca2+, leading to a decrease in the secretory output.

As it has been previously established, secretion of FSH and LH are not co-ordinately regulated, their discordant regulation must be related to differential intracellular responses to several stimuli, factors as activins, inhibins, and follistatin, may play a key role on establishing such differences. In this regard, activins which are produced in a variety of tissues, including gonadotrophs, stimulates FSHβ transcription (132, 157–159) and enhance its sensitivity to GnRH by up-regulation of the GnRH receptor expression (92, 160). Contrary, inhibins which are produced in Sertoli and granulosa cells as well as in gonadotrophs (161), have been shown to rapidly reduce FSHβ synthesis and secretion independently of GnRH (162), by binding to activin receptors on gonadotrophs preventing the assembly of active signaling complexes (92).

Follistatins, which are glycoprotein ubiquitously expressed (including gonadotrophs and follicle stellated cells) bind to activins with high-affinity modulating its actions (92, 132, 160, 163). Activin and follistatin function in a reciprocal feedback loop altering their secretion, internalization, and degradation (92, 114, 160, 163), modifying the rise and fall of biosynthesis and secretion across the reproductive cycle (160, 163).

One mechanism that contributes to differential FSH and LH production may be related to the observation that different patterns of GnRH pulses produce differential effects on inhibin/activin and follistatin mRNA levels (160). Estrogen, progesterone, testosterone, inhibin, activin, follistatin, and hypothalamic GnRH, may combine to distinct regulate LH and FSH during the reproductive cycle (97).

#### **GONADOTROPHS ACTIVITY AT THE TISSUE LEVEL**

Endocrine cells are organized in three-dimensional networks, which facilitate the coordination of the activity of thousands of individual cells to respond to different regulation factors and achieve hormone output (164, 165). The magnitude of the hormone pulses into the systemic circulation is apparently not just the simple addition of the individual endocrine activity, instead, biophysical and biochemical interactions in the whole tissue must be essential for*in vivo* organization. However, as it has been described in this and other works, most of the studies have been done in individual cell activity where this networks and relations are disrupted, due to methodological difficulties, just few recently approaches has been done in the understanding of the endocrine activity in a tissue context.

In this regard, the distribution of gonadotrophs in fixed and live slices at different female reproductive stages has been analyzed (166). Across different physiological stages, pituitary gonadotrophs shows changes in their distribution within the gland and in response to GnRH stimulation (166), this might represent and adaptation to better respond at different conditions.

The possibility of changes in gonadotrophs activity within its tissue context and physiological conditions was recently addressed using Ca2<sup>+</sup> imaging in male mouse acute pituitary slices (11, 145). Cells in this preparation are amenable to functional studies in their native environment. We showed that rather than a constant number of gonadotrophs responding to GnRH stimulus, the number of responding cells grew with increasing GnRH concentration (GnRH), and in general, gonadotrophs Ca2<sup>+</sup> signaling resembled that recorded in primary cultures (11, 145). However, Ca2<sup>+</sup> imaging in acute mouse pituitary slices revealed Ca2<sup>+</sup> signaling patterns unique to *in situ* conditions, gonadotrophs (58%) under increasing doses of GnRH stimulation exhibited a progression of Ca2<sup>+</sup> signaling patterns termed "non-canonical" [i.e., oscillatory responses at a given (GnRH) and transient responses at both lower and higher concentrations as described before in this review; **Figure 3**], and some of them (3.6%) even showed atypical (non-oscillatory) responses, regardless of the (GnRH) used (145). Furthermore, responses to a given dose of GnRH varied considerably from one cell to another, reflecting a range of dose-response properties in the *in situ* gonadotroph population.

As it has been described in this review, following the removal of the gonads, the population of pituitary gonadotrophs undergoes drastic functional and morphological modifications concomitantly with the large (five to sixfold) increase in gonadotropin secretion that characterizes this condition (123, 154, 167, 168) some changes as amplitude and frequency of GnRH-induced Ca2<sup>+</sup> signaling has been reported in dissociated cells (10) and there is no difference with respect of what it has been reported in acute pituitary slices from 15 and 45 days castrated male mice (GnX) (145). Nevertheless, other characteristics on the intracellular Ca2<sup>+</sup> signaling appear to be different; gonadotrophs of pituitary slices from GnX responding with "non-canonical" sequences of Ca2<sup>+</sup> signaling (described earlier in this review) to increasing GnRH were significantly augmented (80% of GnRH responding gonadotrophs) and "canonical" sequences were significantly reduced (145) (**Figure 3**), indicating that probably this sequences of Ca2<sup>+</sup> signaling in response to GnRH are modulated by paracrine and systemic factors as testosterone, allowing gonadotrophs to adapt to different physiological requirements. Additionally, median effective dose (ED50) for GnRH decreased from 0.17 nM (control) to 0.07 nM after GnX, suggesting an increased GnRH responsiveness of the gonadotroph population (145). Different sizes of gonadotrophs are present in intact mice pituitary gland, most gonadotrophs (97%) were smaller than 60µm<sup>2</sup> with a mean of 31.3 ± 0.6µm<sup>2</sup> in area and even if large interindividual variation on the peak amplitude of Ca2<sup>+</sup> transients (Max DF) was seen, no matter the size of the cell, they generated intracellular Ca2<sup>+</sup> signals smaller than 40 fluorescence arbitrary units (a.u.) poorly correlated with the cell size (**Figure 4**). By contrast it is reported that 15-day castrated male mouse pituitary gonadotrophs, whose size increase to a mean of 54.4 ± 1.24µm<sup>2</sup> and 26% of cells larger than 60µm<sup>2</sup> present less variation on the Ca2<sup>+</sup> peak amplitude and significantly higher correlation of this with the cell size (i.e., hypertrophied gonadotrophs tended to generate Ca2<sup>+</sup> signals

of greater amplitude) (145) (**Figure 4**), suggesting that in this condition, Ca2<sup>+</sup> peak amplitude correlated with cell size, and that hypertrophied gonadotrophs tended to produce stronger GnRH-induced Ca2<sup>+</sup> signals.

Functional adaptation of the gonadotrophs in the pituitary gland to different external and internal conditions may involucrate not just alterations in cell number, size, and morphology, as it has been considered for many years, recent methodological techniques allowed us to understand that it is a more complicated process that involucrates different aspects at the cellular physiology level but in coordination with the whole tissue environment.

#### **ACKNOWLEDGMENTS**

The authors are indebted to Dr. Arturo Hernández who provided invaluable insights throughout this work. We are also grateful with

#### **REFERENCES**


5. Tse FW, Tse A, Hille B, Horstmann H, Almers W. Local Ca2+ release from internal stores controls exocytosis in pituitary gonadotrophs. *Neuron* (1997) **18**:121–32. doi:10.1016/S0896- 6273(01)80051-9


Daniel Diaz for his support on data analysis and figure design. Grant support: PAPIIT IN222613 and IN222413, CONACYT 166430.

agonist-induced calcium oscillations in pituitary gonadotrophs. *Mol Endocrinol* (1991) **5**:949–58. doi:10.1210/mend-5-7-949


gonadotrophs recorded from acute pituitary slices in vitro. *Neuroendocrinology* (2010) **91**:239–55. doi:10.1159/000274493


oscillations in pituitary cells of newborn rats. *Neuroendocrinology* (1997) **65**:276–83. doi:10.1159/000127185


(2005) **38**:87–99. doi:10.1016/j. ceca.2005.06.006


*Rev Mol Cell Biol* (2006) **7**:631–43. doi:10.1038/nrm2002


requirement for pulsatile administration for luteinizing hormone-beta gene stimulation. *Mol Endocrinol* (1990) **4**:1444–50. doi:10.1210/mend-4-10-1444


of follicle-stimulating hormone beta by pulsatile gonadotropinreleasing hormone is mediated by functional antagonism of bZIP transcription factors. *Mol Cell Biol* (2010) **30**:1028–40. doi:10.1128/MCB.00848-09


gene expression by pulsatile GnRH stimulation. *Biol Reprod* (2013) **88**:35. doi:10.1095/biolreprod.112. 105601


doi:10.1016/j.fertnstert.2010.03. 034


to changes in pituitary and plasma gonadotropins in ovariectomized hypothalamo/pituitarydisconnected ewes. II. A marked rise in receptor number during the acute feedback effects of estradiol. *Biol Reprod* (1988) **39**:349–54.


*Endocrinology* (1984) **115**:114–20. doi:10.1210/endo-115-1-114


ovine follicle-stimulating hormone-beta gene by 17 beta-estradiol. *Endocrinology* (1996) **137**:3437–46. doi:10.1210/en.137.8.3437


encoding luteinizing hormonereleasing hormone in intact, castrated, and hyperprolactinemic male rats. *Endocrinology* (1991) **128**:459–66. doi:10.1210/endo-128-1-459


hormone beta-subunit mRNA levels in rat pituitary cells. *Mol Endocrinol* (1990) **4**:1620–6. doi:10.1210/mend-4-11-1620


method with a photobiotinylated complementary RNA probe. *Mol Endocrinol* (1987) **1**:926–32. doi:10.1210/mend-1- 12-926


*Virchows Arch B Cell Pathol* (1973) **13**:145–52.


(1995) **136**:1885–91. doi:10.1210/ en.136.5.1885


Fiordelisio T, et al. Coordination of calcium signals by pituitary endocrine cells in situ. *Cell Calcium* (2012) **51**:222–30. doi:10. 1016/j.ceca.2011.11.007


**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: 06 July 2013; accepted: 31 August 2013; published online: 30 September 2013.*

*Citation: Durán-Pastén ML and Fiordelisio T (2013) GnRH-induced Ca2*<sup>+</sup> *signaling patterns and gonadotropin secretion in pituitary gonadotrophs. Functional adaptations to both ordinary and extraordinary physiological demands. Front. Endocrinol. 4:127. doi: 10.3389/fendo.2013.00127*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

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

### The regulated secretory pathway and human disease: insights from gene variants and single nucleotide polymorphisms

### **Wei-Jye Lin<sup>1</sup> and Stephen R. Salton1,2,3\***

<sup>1</sup> Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, NY, USA

<sup>2</sup> Friedman Brain Institute, Icahn School of Medicine at Mount Sinai, New York, NY, USA

<sup>3</sup> Department of Geriatrics, Icahn School of Medicine at Mount Sinai, New York, NY, USA

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Paolo Magni, Università degli Studi di Milano, Italy T. Rajendra Kumar, University of Kansas Medical Center, USA

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

Stephen R. Salton, Department of Neuroscience, Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, Box 1065, New York, NY 10029, USA e-mail: stephen.salton@mssm.edu The regulated secretory pathway provides critical control of peptide, growth factor, and hormone release from neuroendocrine and endocrine cells, and neurons, maintaining physiological homeostasis. Propeptides and prohormones are packaged into dense core granules (DCGs), where they frequently undergo tissue-specific processing as the DCG matures. Proteins of the granin family are DCG components, and although their function is not fully understood, data suggest they are involved in DCG formation and regulated protein/peptide secretion, in addition to their role as precursors of bioactive peptides. Association of gene variation, including single nucleotide polymorphisms (SNPs), with neuropsychiatric, endocrine, and metabolic diseases, has implicated specific secreted proteins and peptides in disease pathogenesis. For example, a SNP at position 196 (G/A) of the human brain-derived neurotrophic factor gene dysregulates protein processing and secretion and leads to cognitive impairment.This suggests more generally that variants identified in genes encoding secreted growth factors, peptides, hormones, and proteins involved in DCG biogenesis, protein processing, and the secretory apparatus, could provide insight into the process of regulated secretion as well as disorders that result when it is impaired.

**Keywords: single nucleotide polymorphism (SNP), dense core granule (DCG), insulin, brain-derived neurotrophic factor (BDNF), chromogranin, neurotrophin, proopiomelanocortin (POMC), prohormone convertase**

#### **GENERAL OVERVIEW OF REGULATED SECRETION**

Neuronal and endocrine peptides, growth factors, and hormones maintain physiological homeostasis. Their release is therefore tightly controlled by regulated secretion (1). Messenger RNAs (mRNAs) are translated and secreted proteins enter the cisternae of the rough endoplasmic reticulum (RER), are transported to the trans-Golgi network (TGN), are targeted into immature dense core granules (DCGs), and are retained in mature DCGs (also called large dense core vesicles or LDCVs), in endocrine and neuronal cells. Proteins including members of the granin family, such as chromogranin A (CgA), chromogranin B (CgB), secretogranin II (SgII), and secretogranin III (SgIII), the exopeptidase carboxypeptidase E (CPE), and prohormone convertases 1/3 and 2 (PC1/3 and PC2), are also sorted into DCGs, which in addition, increase the diversity of biologically active peptides through processing of granin precursor proteins (2).

Although the sorting of proteins into DCGs is not fully understood, interaction of DCG cargo proteins with TGN lipid raftanchoring or membrane-spanning proteins (SgIII, CPE, sortilin) plays an essential role in docking and concentrating of DCG components and is critical for the correct sorting of cargo proteins and for vesicle biogenesis (3, 4). DCG sorting domains have been identified in prohormones, propeptides, and granins, and although not highly conserved, these polypeptide motifs are required for regulated secretion. After budding from the TGN, immature DCGs undergo acidification, mediated by the DCG membrane-spanning proton pump, leading to activation of resident prohormone convertases and carboxypeptidases. All granin members and most other propeptides,prohormones,and prohormone convertases are known to undergo proteolytic processing during granule maturation, which is essential for enzymatic activation and the generation of biologically active peptides.

Granin family members, including CgA, CgB, SgII, and SgIII, are the most abundant DCG components, with CgA, for example, constituting almost half of the soluble protein content of the adrenal chromaffin cell secretory granule (5). The importance of granin proteins in vesicle biogenesis has been demonstrated by overexpression of CgA or CgB in cells that lack a classical regulated secretory pathway, such as fibroblasts, which results in the production of granule-like structures and the regulated release of vesicle contents (6, 7). Moreover, gene silencing studies demonstrate reciprocal effects on DCG biogenesis, with decreased DCG number reported in PC12 cells that contain reduced levels of CgA or SgII, and in the adrenal medulla of CgA knockout mouse (8–10). Other CgA and CgB knockout mouse models, although showing no change in DCG number in adrenal medulla and other endocrine tissues, were found to have increased levels of other granin proteins, which is likely due to compensatory mechanisms (11, 12).

#### **MECHANISMS BY WHICH GENETIC VARIANTS CAN IMPACT SECRETED PROTEINS**

Single nucleotide polymorphisms (SNPs) are the most common genetic variations in genomic DNA, occurring 1/1200 bp on average (13). Individual SNPs submitted to the Single Nucleotide Polymorphism Database (dbSNP), a service provided by the National Center for Biotechnology Information (NCBI), are assigned a unique reference SNP ID number (rs#) to map the SNP to other external databases. Many SNPs are silent variations and their occurrence does not affect gene expression or protein function. However, a single nucleotide change has the potential to modulate protein level or function if the SNP is located: (1) in the promoter region, which may alter transcriptional activity, (2) in intronic regions, which may interfere with splicing efficiency if the SNP is found at exon-intron boundaries in splice donor sites, regulatory sequences, or splice acceptor sites, (3) in the 3<sup>0</sup> untranslated region (30UTR), which may affect the expression levels of either mRNA or protein due to altered *cis*-elements including mRNA destabilization sequences, AU-rich elements, and translational repressor binding sites, including miRNA targeting sites, and (4) in the protein coding region, which can cause missense or nonsense mutations. Nonsense mutations can result in premature termination of translation and the production of truncated proteins, while missense mutations can lead to loss-of-function or gainof-function. With respect to proteins in the regulated secretory pathway, missense SNPs could impact enzymatic catalytic activity, alter peptide affinity for receptors, impair protein sorting when found in targeting motif(s) of secretory proteins, or block peptide processing when located in the cleavage sites of prohormones or propeptides (**Figure 1**).

#### **REPORTED SNPs IN GENES ENCODING SECRETED PROTEINS THAT ARE ASSOCIATED WITH NEUROPSYCHIATRIC OR ENDOCRINE/METABOLIC DISEASE**

Neurotrophic growth factors, hormones, peptide precursors, prohormone convertases, granins, and other DCG proteins are secreted through the regulated or constitutive secretory pathways. Even relatively subtle changes in protein processing, levels, or regulated secretion, due to specific SNPs that impact a variety of proteins in the secretory pathway, have been reported to be associated with neuropsychiatric or metabolic disease, and these studies are reviewed below.

#### **NEUROTROPHINS**

#### **Brain-derived neurotrophic factor**

Brain-derived neurotrophic factor (BDNF) is well known for its role in regulating neuroplasticity and neurogenesis in the brain. A functional variant in the propeptide domain of BDNF, SNP rs6265 (G- to A-allele, Val66Met), was found to affect the secretion of BDNF (14). Previous biochemical studies have shown that BDNF (Met66, encoded by minor A-allele) fails to bind to sortilin through its pro-domain region, and this lowers its activity-dependent secretion due to failed DCG sorting and localization (14, 15). BDNF (Met66) is associated with reduced brain volume, impaired episodic memory (16), and high trait anxiety (17). Two other SNPs located at the protease cleavage site of the proBDNF protein, SNPs rs1048220 (G- to T-allele, Arg125Met) and rs1048221 (G- to T-allele, Arg127Leu), partially impair proBDNF cleavage but not DCG sorting and secretion (18). The minor alleles of SNP rs1048220 (T-allele) and SNP rs1048218 (T-allele) have also been associated with familial and sporadic Alzheimer's disease (19).

#### **Nerve growth factor**

Mature nerve growth factor (NGF) binds to tropomyosin-related kinase A (TrkA) receptor tyrosine kinase and activates signaling pathways that regulate neuronal differentiation and survival. Unprocessed proNGF, however, has higher binding affinity for the p75NTR neurotrophin receptor, and can stimulate either cell survival or programed cell death (20). SNP rs6330 (C- to T-allele, Ala35Val) was previously identified in the pro-domain region of NGF, which has the potential to alter the efficiency of proNGF sorting or processing. The nucleotide variant of SNP rs6330 was found to be associated with increased anxiety in women (C-allele) (21), was over-transmitted in Attention Deficit Hyperactivity Disorder (ADHD)-affected children (C-allele) (22), and was associated with Alzheimer's disease onset (both C- and minor T-alleles) (19, 23).

#### **NEUROPEPTIDES**

#### **Proopiomelanocortin**

Expression of proopiomelanocortin (POMC), the precursor of several processed neuropeptides [α-MSH, adrenocorticotrophic hormone (ACTH), β-MSH, and β-endorphin], is regulated by the adipocyte-derived hormone leptin, which signals by binding receptors on neurons in the hypothalamic arcuate nucleus. The POMC-derived peptide α-MSH activates the melanocortin 4-receptor (MC4R) and suppresses food intake. Lack of POMC in humans and mouse models leads to the development of severe obesity, ACTH deficiency, and hypopigmentation (24, 25). A missense amino acid substitution by SNP rs28932472 (G- to C-allele, Arg236Gly) in the *POMC* gene was reported to disrupt the dibasic processing site between β-MSH and β-endorphin. The minor Callele that encodes POMC (Gly236) results in an aberrant fusion peptide of β-MSH and β-endorphin, which still binds to the MC4R but antagonizes its activation. As a consequence, the Gly236 variation is associated with early onset obesity in several ethnic groups (26).

#### **Agouti-related peptide**

Expressed in the arcuate nucleus of the hypothalamus, agoutirelated peptide (AgRP) is an endogenous antagonist of the MC4R that increases feeding behavior. The amino acid substitution of Ala67 to Thr67 caused by SNP rs5030980 (G- to A-allele) in the *AgRP* gene is associated with Anorexia Nervosa and leanness (27). In healthy subjects, homozygosity for the Thr67 allele is associated with low body fat mass and low lean body mass, while the Ala67 allele is associated with late-onset obesity. Although the Thr67 variation results in no detectable change in binding to the MC4R or sorting efficiency into and secretion from DCGs, the polarity of the amino acid substitution may cause a conformational change in AgRP, affecting other peptide functions that need further characterization (28).

#### **Neuropeptide Y**

Neuropeptide Y (NPY) is a potent orexigenic peptide that is also expressed in the hypothalamus. It regulates energy balance

Single nucleotide polymorphisms (SNPs) can affect granule biogenesis and release through these mechanisms: (1) SNPs in the promoter or 3<sup>0</sup>UTR of granule-associated genes modulate gene expression levels, resulting in altered granule biogenesis, (2) SNPs in the pro-domain of peptides or hormones block their processing and/or sorting into the regulated pathway, (3) enzymatic activity and lead to aberrant propeptide processing, (4) SNPs in the signal peptide and/or sorting domain(s) disrupt regulated secretion, and (5) SNPs that change the processing or sequence of neuropeptides modulate membrane receptor binding affinity. TrkB, tropomyosin-related kinase B receptor; INSR, insulin receptor.

through effects on energy intake, expenditure, and partitioning. A non-synonymous SNP rs16139 that leads to an amino acid change (T- to C-allele, Leu7Pro) in the signal peptide domain of preproNPY has been reported to cause a tertiary structural change in its sorting domain, and this Pro7 substitution alters intracellular proNPY packaging, processing, and secretion (29–31). NPY (Pro7) is associated with elevated food intake, altered free fatty acid (FFA) metabolism and high serum cholesterol and LDL cholesterol levels, but doesn't affect insulin sensitivity, insulin secretion, or glucose metabolism (30, 32, 33). Lower plasma NPY and norepinephrine concentrations, and lower insulin but higher glucose concentrations in plasma, were also reported in the population with the Pro7 substitution (34).

#### **HORMONES Insulin**

A nucleotide variation in the proinsulin gene is located at the Cpeptide-A-chain junction (C65, causes Arg to His), and the His65 substitution prevents processing of the dibasic cleavage site, resulting in hyperproinsulinemia that is caused by the accumulation of a circulating, biologically defective form of the proinsulin intermediate peptide, which fails to be metabolized via receptor-mediated endocytosis (35). The other identified proinsulin nucleotide variation results in an amino acid substitution of B10 (His to Asp), resulting in aberrant proinsulin sorting into the constitutive secretory pathway and a subsequent failure in peptide processing, which is also associated with hyperproinsulinemia in affected individuals (36, 37).

#### **GRANINS**

Proteins of the granin family, including the chromogranins and secretogranins, have been demonstrated to play an important role in DCG biogenesis, in neural, neuroendocrine, and endocrine cells (2). It is not too surprising, then, that a number of SNPs which alter granin expression levels have been associated with metabolic diseases or neurological disorders, because physiological homeostasis is tightly regulated by neuropeptides, growth factors, and hormones, all of which are processed and stored in DCGs.

#### **Chromogranin A (CHGA)**

The combination of SNPs that are inherited together is called a haplotype, which can be used for studying genetic linkage of diseases. Two haplotype polymorphism carriers, haplotype (A-T-C) of SNPs rs9658634–rs9658635–rs7159323 in the *CHGA* promoter region, and haplotype (T-C) of SNPs rs7610–rs875395 in the *CHGA* 3 <sup>0</sup>UTR and downstream regions, are linked to hypertensive renal disease (38). SNP rs9658634 in the *CHGA* promoter was found to be located in a predicted PPARγ/RXRα binding motif, and the nucleotide variant A-allele disrupted reporter expression that was co-stimulated by PPARγ/RXRα and their cognate ligands. Physiologically, the minor A-allele is associated with lower leptin secretion, as well as lower BMI, especially in women (39). SNP rs7610 (minor T-allele) has been identified in *CHGA*, showed decreased reporter expression in PC12 pheochromocytoma cells, and was associated with lower plasma chromogranin A levels and blood pressure (BP) in a sex-dependent manner (40). Other polymorphisms identified in *CHGA* coding sequence (rs9658667, G- to A-allele, Gly364Ser, and rs9658668, C- to T-allele, Pro370Leu) result in altered catestatin activity, changing its potency to inhibit nicotinic acetylcholine receptor (nAChR)-stimulated catecholamine release from chromaffin cells, and likely linking these SNPs to an increased risk of developing hypertension (41, 42). SNP rs9658635 and haplotype (C-T) of SNPs rs9658635–rs729940 are both linked to the onset of schizophrenia in the Japanese population, but their effect on *CHGA* gene expression remains to be determined (43).

#### **Chromogranin B (CHGB)**

Reduced chromogranin B levels have been observed in the thalamic subregion, the mediodorsal nucleus, parvocellular division (MDNp; *CHGB* mRNA levels), and cerebrospinal fluid (both CgA and CgB protein levels) of schizophrenic patients (44, 45). Two polymorphisms identified in the *CHGB* coding sequence, rs236152 (C- to G-allele, Arg353Gly) and rs236153 (A- to Gallele, Glu368Glu), are associated with schizophrenia in a Japanese population study, although the functional consequences of these two SNP variants are unknown (46). A haplotype (A-T) of SNPs rs236140–rs236141 identified in the *CHGB* gene promoter region shows the highest transcriptional strength in reporter assays carried out in PC12 cells, and interestingly, the (A-T) haplotype is strongly associated with hypertension (47). The SNP rs2821 (minor A-allele) found in an RNA-destabilizing A/U-rich motif of the *CHGB* 3 <sup>0</sup>UTR was reportedly associated with lower plasma chromogranin B levels in population studies, likely due to shortened *CHGB* mRNA half-life (48).

#### **Secretogranin II and secretogranin III (SCG2 and SCG3)**

Single nucleotide polymorphism rs1017448 (minor A-allele) was found in the first intron of the *SCG2* gene. It is significantly associated with elevated BP, and likely increases *SCG2* gene expression by facilitating the recruitment of paired-like homeobox transcriptional factor 2a (Phox2a) (49). SNP rs3764220 (A-allele), which is found in the promoter region of the *SCG3* gene, is associated with metabolic syndrome-related physiological changes, including dyslipidemia, hypertension, and impaired glucose tolerance, with increased risk of cardiovascular diseaserelated morbidity and mortality (50). SNP rs16964465 (minor C-allele) in the *SCG3* promoter region and minor G-allele of rs16964476 in the first intron, both enhanced transcriptional activity of a reporter in the neuroblastoma cell line SH-SY5Y (51). Interestingly, in hypothalamus, SgIII protein is detected in both POMC and NPY neurons in the arcuate nucleus, and in orexin-expressing and melanin-concentrating hormone (MCH) expressing neurons in the lateral hypothalamus; all these neuropeptides regulate food intake (51). Notably, these two minor alleles, both associated with increased SgIII expression, seemed to confer resistance to the onset of obesity, suggesting the possibility that decreased SgIII levels could reciprocally increase the risk of obesity.

#### **PEPTIDASES AND CONVERTASES Prohormone convertase 1/3**

Prohormone convertases are a family of endopeptidases that cleave proteins at internal sites, generally paired or clustered basic lysine, and/or arginine residues. PC1 is encoded by the proprotein convertase, subtilisin/kexin-type 1 (*PCSK1*) gene. Two functionally relevant SNPs, originally identified in different *PCSK1* alleles of a female patient with childhood early onset obesity, are associated with abnormal glucose homeostasis, and elevated plasma proinsulin and POMC levels (52, 53). SNP rs137852821 (G- to A-allele) changes an amino acid from Gly483 to Arg483, which prevents processing of proPC1 and favors its retention in the RER. Another C-allele variant found in the intron-5 donor splice site of *PCSK1* gene also causes exon 5 skipping and results in premature termination of PC1 translation in the catalytic region. Yet another identified non-synonymous allele substitution that is associated with early onset obesity is SNP rs6232 (A- to G-allele), which changes Asn221 to Asp221 and causes impaired catalytic function of PC1 (54).

#### **Carboxypeptidase E**

Carboxypeptidase E trims C-terminal lysine and arginine residues from peptides that are generated by prohormone convertase cleavage of precursor proteins at paired dibasic residues (55). Notably, CPE is also a sorting/retention receptor for proinsulin and proBDNF trafficking to the regulated secretory pathway (56, 57). The SNP rs144727363 (C- to T-allele, Arg189Trp) in the *CPE* coding region changes arginine to tryptophan, which reduces enzymatic activity of CPE, and is associated with hyperproinsulinemia and type II diabetes mellitus in affected Ashkenazi Jewish families (58). Similarly, a missense mutation in the *CPE* gene (Ser202Pro), identified from the mouse *fat/fat* model, which generates an enzymatically inactive and unprocessed protein product, results in impaired processing of proinsulin and other propeptides, chronic hyperproinsulinaemia, and obesity (59–61).

#### **OTHER PROTEINS THAT REGULATE SECRETORY PATHWAY FUNCTION ATP6V0A1**

*ATP6V0A1* encodes the α1 subunit of the vacuolar H+ translocating ATPase complex, which functions in acidification of intracellular organelles. Bafilomycin A1,a chemical inhibitor of the vacuolar H<sup>+</sup> ATPase, impairs DCG formation, and the sorting of secretory proteins into the regulated pathway (62). A reported SNP rs938671 (minor C-allele, 3246 T/C) in the 30UTR of *ATP6V0A1* gene creates a binding motif for the micro-RNA hsa-miR-637, which decreases overall gene expression of *ATP6V0A1*. Its clinical association with hypertension may be due to altered DCG acidification by the risk C-allele, and as a consequence, decreased sorting, retention, and/or processing of DCG components (63).

#### **Sortilin (SORT1)**

Sortilin, a Vps10p domain protein, binds to proBDNF, and other polypeptides and facilitates their trafficking into the regulated secretory pathway (15). Plasma membrane sortilin also functions as an internalization receptor for apolipoprotein A-V and progranulin uptake (64, 65). Several SNPs are located in a non-coding region downstream of the *SORT1* allele and strongly correlate with increased sortilin expression, including SNP rs12740374 (minor Tallele), which creates a C/EBPα binding site, SNP rs646776 (minor C-allele), and SNP rs599839 (minor G-allele). Increased sortilin expression is reported to associate with reduced ApoB secretion,

#### **REFERENCES**


similar role of chromogranin A. *J Biol Chem* (2003) **278**:40581–9. doi: 10.1074/jbc.M304942200


enhanced LDL uptake in the liver, and increased uptake of plasma progranulin (66–69). How these existing SNPs affect other known function of sortilin, including propeptide sorting, is still unclear.

#### **FUTURE PERSPECTIVES**

Protein secretion from neuroendocrine, neural, and endocrine cells is a complex, highly regulated process. Due to the nature of the cargo, which includes critical growth factors, hormones, peptide precursors, the enzymes that process them, and proteins that function in secretory vesicle biogenesis, subtle alterations in the regulated secretory pathway, and/or the proteins transiting through it, can have a significant physiological impact. Many studies reviewed here describe the association of specific, discrete SNPs in the genes of secreted proteins with the risk or onset of human disease. Neuropsychiatric disorders such as schizophrenia, metabolic disorders including obesity and diabetes, and hypertension are strongly associated with a number of well-characterized SNPs. These identified SNPs together with genome-wide association studies (GWAS) and molecular and cellular analyses of SNP function will advance our understanding of the process of regulated secretion and the important roles that secreted and secretory proteins play in maintaining physiological homeostasis.

#### **ACKNOWLEDGMENTS**

Research in the authors' laboratory is supported by NIH grants MH086499 and DE021996, and by the Diabetes Action Research and Education Foundation.

al. Targeted ablation of the chromogranin a (Chga) gene: normal neuroendocrine dense-core secretory granules and increased expression of other granins. *Mol Endocrinol* (2006) **20**:1935–47. doi:10.1210/ me.2005-0398


to the regulated secretory pathway. *J Neurosci* (2005) **25**:6156–66. doi:10. 1523/JNEUROSCI.1017-05.2005


**70**:308–16. doi:10.1016/j.bcp.2005. 04.033


*U S A* (1987) **84**:2194–7. doi:10. 1073/pnas.84.8.2194


variants in PCSK1 confer risk of obesity. *Nat Genet* (2008) **40**:943–5. doi:10.1038/ng.177


with a carboxypeptidase E mutation which reduces enzyme activity. *Nat Genet* (1995) **10**:135–42. doi:10.1038/ng0695-135


apolipoprotein A-V by members of the low density lipoprotein receptor and the VPS10p domain receptor families. *J Biol Chem* (2008) **283**:25920–7. doi:10.1074/jbc.M802721200


69. Strong A, Ding Q, Edmondson AC, Millar JS, Sachs KV, Li X, et al. Hepatic sortilin regulates both apolipoprotein B secretion and LDL catabolism. *J Clin Invest* (2012) **122**:2807–16. doi:10.1172/JCI63563

**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: 29 May 2013; accepted: 23 July 2013; published online: 06 August 2013. Citation: Lin W-J and Salton SR (2013) The regulated secretory pathway and human disease: insights from gene variants and single nucleotide polymorphisms. Front. Endocrinol. 4:96. doi: 10.3389/fendo.2013.00096*

*This article was submitted to Frontiers in Neuroendocrine Science, a specialty of Frontiers in Endocrinology.*

*Copyright © 2013 Lin and Salton. 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.*

# Platelet granule exocytosis: a comparison with chromaffin cells

#### **Jennifer L. Fitch-Tewfik and Robert Flaumenhaft\***

Division of Hemostasis and Thrombosis, Department of Medicine, BIDMC, Harvard Medical School, Boston, MA, USA

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Ricardo Borges, University of La Laguna, Spain Joshua J. Park, University of Toledo College of Medicine, USA

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

Robert Flaumenhaft, Center for Life Science, Beth Israel Deaconess Medical Center, Room 939, 3 Blackfan Circle, Boston, MA 02215, USA e-mail: rflaumen@bidmc.harvard.edu

The rapid secretion of bioactive amines from chromaffin cells constitutes an important component of the fight or flight response of mammals to stress. Platelets respond to stresses within the vasculature by rapidly secreting cargo at sites of injury, inflammation, or infection. Although chromaffin cells derive from the neural crest and platelets from bone marrow megakaryocytes, both have evolved a heterogeneous assemblage of granule types and a mechanism for efficient release. This article will provide an overview of granule formation and exocytosis in platelets with an emphasis on areas in which the study of chromaffin cells has influenced that of platelets and on similarities between the two secretory systems. Commonalities include the use of transporters to concentrate bioactive amines and other cargos into granules, the role of cytoskeletal remodeling in granule exocytosis, and the use of granules to provide membrane for cytoplasmic projections. The SNAREs and SNARE accessory proteins used by each cell type will also be considered. Finally, we will discuss the newly appreciated role of dynamin family proteins in regulated fusion pore formation. This evaluation of the comparative cell biology of regulated exocytosis in platelets and chromaffin cells demonstrates a convergence of mechanisms between two disparate cell types both tasked with responding rapidly to physiological stimuli.

**Keywords: exocytosis, granules, platelets, chromaffin system, cytoskeleton, dynamins, SNAREs**

#### **INTRODUCTION**

Platelets are small, anucleate blood cells derived from bone marrow megakaryocytes. They are best known for their central role in maintaining the integrity of the vasculature (hemostasis) and for their pathological role in clotting arteries and veins (thrombosis) during myocardial infarction, stroke, peripheral vascular disease, and deep vein thrombosis. In addition to their role in hemostasis, platelets have also been proposed to function in many other aspects of host defense. Stimulus-induced release of platelet granules contributes to nearly all platelet functions including hemostasis and thrombosis, inflammation, angiogenesis, and anti-microbial activities (Blair and Flaumenhaft, 2009). Platelets contain three granule types: α-granules, dense granules, and lysosomes (**Figure 1**; **Table 1**). Absence of dense granules, as observed in inherited syndromes such as Hermansky–Pudlak syndrome or Chediak–Higashi syndrome, results in a bleeding diathesis (Hermansky and Pudlak, 1959). Absence of α-granules, as observed in gray platelet syndrome, also increases bleeding (Buchanan and Handin, 1976; Costa et al., 1976). The bleeding phenotype associated with these disorders underscores the importance of platelet granules in hemostasis.

Despite the functional importance of platelet granule secretion in maintaining vascular integrity and promoting host defense, the molecular basis of platelet granule secretion remained poorly studied until the late 1990s, despite transformative advances in secretion biology that had occurred over the preceding decade (Rothman and Orci, 1992; Sollner et al., 1993). This knowledge deficit was due in part to the fact that platelets are anucleate, complicating the use of standard molecular biological approaches that

have been widely used to study regulated secretion in nucleated cells. In addition, the small size (2–3µm in diameter) and unusual membrane system of the platelet prevented application of classic electrophysical approaches such as patch-clamp studies. Earlier studies evaluating the molecular mechanisms of platelet granule secretion relied on applying knowledge derived from other systems to the study of platelets. The chromaffin cell has been influential in this regard. Although these two cell types have different embryonic derivations and functions, both cells store bioactive amines and peptides at high concentrations and release their cargos rapidly in response to stress signals (**Table 1**). The study of platelet granule secretion has matured considerably over the past decade, making relevant a comparison of the mechanisms by which platelets and chromaffin cells store and release their granule contents in response to environmental signals.

#### **PLATELET GRANULE TYPES**

#### α**-GRANULES**

α-Granules are by far the most abundant platelet granule type (**Figure 1**). There are ∼50–80 α-granules/platelet, ranging in size from 200 to 500 nm. They comprise roughly 10% of the platelet volume, 10-fold more than dense granules. α-Granules contain a variety of membrane proteins and soluble cargo that give them a distinct appearance when stained with osmium and viewed by transmission electron microscopy (TEM). Proteomic analyses indicate that these granules contain hundreds of different types of proteins (Coppinger et al., 2004; Piersma et al., 2009). Protein cargos found in α-granules include neuroactive peptides that are more typically associated with chromaffin cells, including tachykinins and enkephalins (Graham et al., 2004). Conversely, proteomic studies suggest that chromaffin large dense-core vesicles (LDCVs) contain several major constituents of α-granules that can act in the vasculature, including platelet basic protein precursor, TGF-β, collagen isoforms, and metalloproteases (**Table 2**) (Wegrzyn et al., 2010). As with chromaffin cells, the mechanisms by which proteins are packaged in platelet storage granules are incompletely understood.

Platelet α-granule cargos can include coagulants and anticoagulants, angiogenic and antiangiogenic factors, proteases and proteases inhibitors, and proinflammatory and anti-inflammatory

mediators. This observation has raised the question of how αgranules are able to efficiently mediate their biological functions when they contain so many proteins with opposing functions (Italiano et al., 2008; Blair and Flaumenhaft, 2009). One possibility is that there are different α-granule subpopulations that store distinct cargo. However, the number of discrete types of α-granule is not known. Evidence that α-granules are heterogeneous comes from several sources. Immunofluorescence microscopy demonstrated that the two α-granule cargos von Willebrand factor and fibrinogen do not localize to the same granule (Sehgal and Storrie, 2007). Subsequent studies showed that angiogenic factors localize to distinct compartments and were differentially released by different agonists (Italiano et al., 2008). The molecular mechanisms that mediate differential release are unclear. Differential distribution of SNAREs among subpopulations of α-granules may account for differential release. For example, Peters et al. (2012) showed that a population of granules containing vesicle-associated membrane protein-7 (VAMP-7) physically separated from VAMP-3 and VAMP-8-containing granules during spreading. However, the idea of α-granule heterogeneity remains controversial and some investigators in the field believe that granule cargos are stochastically distributed and that differential release either does not occur or is controlled at the level of pore expansion.

Granule heterogeneity and differential release have also been evaluated in chromaffin cells. Morphologic studies demonstrate heterogeneity among both LDCVs and synaptic-like microvesicles (SLMVs) (Koval et al., 2001). Studies using carbon-fiber amperometry to measure catecholamine release from individual granules indicated distinct granule populations on the basis of release kinetics (Tang et al., 2005). Different SNAREs and SNARE chaperones may associate with different granule populations and facilitate differential release. For example, different synaptotagmin isoforms associated with LDCVs and SLMVs and this observation could account for their differential secretion in response to calcium (Matsuoka et al., 2011). Other factors influencing chromaffin granule release include pore expansion kinetics and degree. Basal levels of catecholamine release may occur through a restricted fusion pore, while in response to excitation dynamin and myosinmediated mechanisms may elicit fusion pore expansion (Chan et al., 2010). In addition, large aggregates of chromogranin A require complete fusion to facilitate release (Perrais et al., 2004; Felmy, 2007).


the plasma membrane or the open canalicular system.



#### **DENSE GRANULES**

Dense granules are a subtype of lysosome-related organelle (LRO). There are ∼3–6 dense granules/platelet (Flaumenhaft, 2013). These granules are so electron dense that they can be detected by whole mount electron microscopy in the absence of staining. They are highly osmophilic when viewed by TEM. Dense granules play a critical role in hemostasis and thrombosis, releasing factors such as ADP and epinephrine that act in an autocrine and paracrine manner to stimulate platelets at sites of vascular injury. Dense granules also contain factors that are vasoconstrictive such as serotonin (Flaumenhaft, 2013).

Dense granules and LDCVs have been compared based on their unusually high concentrations of cations,polyphosphates, adenine nucleotides, and bioactive amines such as serotonin and histamine (Sigel and Corfu, 1996) (**Figure 2**; **Table 2**). In platelets, adenine nucleotides are concentrated at ∼653 mM ADP and ∼436 mM ATP (Holmsen and Weiss, 1979). Calcium is at 2.2 M. Chromaffin granules and platelet dense granules are among the few mammalian granule types to contain polyphosphates (Aikawa et al., 1971; Ruiz et al., 2004). Active transport mechanisms are thought to contribute to efficient concentration of these constituents in platelets (**Figure 2**). A vesicular H+-ATPase proton pump maintains the dense-granule lumen at pH ∼5.4 (Dean et al., 1984), similar to the pH of LDCVs. The multidrug transporter MRP4, a multidrug resistance protein, is found on platelet dense granules and is proposed to transport adenine nucleotides into these granules (Jedlitschky et al., 2004). Uptake of serotonin from platelet cytosol into dense granules is mediated by vesicular monoamine transporter 2 (VMAT2). Transport is driven by an electrochemical proton gradient across the granule membrane. VMAT2 also appears to mediate histamine transport into dense granules (Fukami et al., 1984). The primary nucleotide transporter in chromaffin cells is Slc17A/VNUT (Sawada et al., 2008). Whether or not platelets use Slc17 family transporters to

concentrate dense-granule cargo has yet to be evaluated. Like platelets, chromaffin cells use VMAT2, in addition to VMAT1, to pump monoamines from the cytosol into their granules.

#### **LYSOSOMES**

Platelets contain few primary and secondary lysosomes. These lysosomes contain many acid hydrolases and cathepsins as cargo and express CD63 and LAMP-2 in their membrane. Platelet lysosome function is not well-studied. They may serve a role in endosomal digestion, as observed in nucleated cells (Flaumenhaft, 2013).

#### **AN OVERVIEW OF PLATELET GRANULE RELEASE**

Platelets are uniform discoid cells that circulate in a quiescent state and undergo a dramatic morphological change when activated. Their plasma membrane surface area is ∼19µm<sup>2</sup> and the total surface area of their granules is ∼14µm<sup>2</sup> . They have an unusual membrane system, including an open canalicular system (OCS), which is a system of tunneling invaginations of the plasma membrane that is unique to platelets and is estimated to have a surface area of ∼14µm<sup>2</sup> (Flaumenhaft, 2013). The OCS tracks through the platelet, but is topologically similar to the plasma membrane in that it possesses both an extracellular and a cytosolic face. Platelets also have a dense tubular system (DTS), which is a membrane system thought to be derived from the megakaryocytic endoplasmic reticulum. The DTS serves as an intracellular calcium storage site, but is not directly connected to either the plasma membrane or the OCS (van Nispen tot Pannerden et al., 2010).

Ultrastructural studies have demonstrated several atypical features of the platelet release reaction. In the resting state, platelet α-granules and dense granules are distributed throughout the platelet. With activation-induced shape change, granules become localized in a central granulomere. As with chromaffin granules, platelet granules may fuse with one another in a process termed

**FIGURE 2 | A comparison of platelet dense granules and chromaffin LDCVs**. **(A)** Several membrane pumps concentration granule contents in the maturing granule. VMAT2 concentrates serotonin (green). An H<sup>+</sup> -ATPase proton pump maintains the granule at pH ∼5.4 (yellow). MRP4 (blue) is thought to concentrate adenine nucleotides into dense granules. Dense granules also express the tetraspanin CD63 (red) and the lysosomal marker LAMP-2 (purple). Dense granules contain a core of calcium chelated by polyphosphate. **(B)** The chromaffin large dense-core vesicle (LDCV) express a

homotypic fusion (Ginsberg et al., 1980). However, during exocytosis platelet granules then fuse with the OCS (Stenberg et al., 1984; Escolar and White, 1991). Granule contents are released into the OCS and diffuse out into the extracellular environment (Escolar and White,1991). Exocytosis via fusion directly with plasmalemma has also been described (Morgenstern et al., 1987). SNAREs are localized on platelet membranes in a manner to support fusion of granules with OCS, plasma membrane, or other granules (Feng et al., 2002). Despite the morphological differences between exocytosis in platelets and chromaffin cells, similarities in the release mechanism have enabled platelet biologists to use chromaffin cells as a model in studying platelet granule release. For example, both platelets and chromaffin cells require Ca2<sup>+</sup> influx as a mediator of exocytosis via different mechanisms. Upon platelet activation by agonists, the concentration of cytosolic Ca2<sup>+</sup> increases activating protein kinase c (PKC), which is important for granule secretion (Knight et al., 1988; Flaumenhaft, 2013). Formation of an action potential in chromaffin cells triggers Ca2<sup>+</sup> influx via Ca2<sup>+</sup> channels thereby triggering exocytosis (Knight and Scrutton, 1980; Knight et al., 1982; Knight and Baker, 1985; Penner and Nicher, 1988; Cheek and Barry, 1993; Livett, 1993; Aunis, 1998; Garcia et al., 2006).

#### **THE CYTOSKELETAL AS BOTH BARRIER AND FACILITATOR IN EXOCYTOSIS**

The observation that platelet granule secretion occurs concurrently with a dramatic change in the shape of the platelet has prompted investigators to evaluate the role of the cytoskeleton in granule release. Platelets are rich in actin, which is the most abundant platelet protein. The resting platelet contains 40% filamentous actin (F-actin). Upon platelet activation, the percentage of F-actin increases to 80%. Studies using cytochalasins (Cox,

variety of membrane proteins including VMAT1 amine transporter (red), H + -ATPase (yellow), Cytochrome b561 (orange), p65 (pink), peptidyl α-amidation monooxygenase (PAM) (blue), LAMP-1 (dark purple), and VNUT/Slc17a nucleotide carrier (green). In addition, the following peripheral proteins are associated with the LDCV membrane: endopeptidases PC1/PC2 (brown), GPIII/SGP2/clusterin (black), carboxypeptidase H (lavender), and Dopamine β-hydroxylase (DβH) (turquoise). The LDCV core contains a large number of different proteins and bioactive compounds.

1988), latrunculin A (Flaumenhaft et al., 2005), Ca2+-mediated stimulation of the F-actin severing protein scinderin (Marcu et al., 1996), and PKC-mediated stimulation of MARCKS (Trifaro et al., 2002) demonstrate increased dense-granule release with inhibition of actin polymerization or with cleavage of F-actin. Inhibition of actin polymerization also augments the kinetics and degree of α-granule release (Flaumenhaft et al., 2005). These results suggest that F-actin disassembly might actually be required for normal granule secretion and that activation-mediated granule release is related to actin.

In contrast to the barrier function that the cytoskeleton serves in the resting state, *de novo* actin polymerization during platelet activation contributes to granule release as evidenced by the observation that high concentrations of inhibitors of actin polymerization block α-granule release (Woronowicz et al., 2010). These studies led to speculation that an actin barrier helps prevent inappropriate α-granule exocytosis,but that some *de novo* actin polymerization is required for α-granule release. Woronowicz et al. (2010) demonstrated that the target membrane SNARE (t-SNARE) SNAP-23 associates with the actin cytoskeleton of resting and activated platelets. In a cell-free platelet granule secretory system, inhibition of F-actin formation blocks release of SNARE-dependent α-granule contents, whereas actin polymerization stimulates αgranule release (Woronowicz et al., 2010). Yet the molecular mechanism by which the binding of SNAREs to the platelet cytoskeleton facilitates granule release is unknown. Overall, actin polymerization appears to serve a bipartite role in platelet granule secretion, both as a barrier to prevent inadvertent loss of thrombogenic cargo and as a facilitator of secretion.

Actin has been shown to serve a barrier function in chromaffin cells. The most well-studied pathways for disrupting the cortical Factin barrier during chromaffin exocytosis include Ca2+-mediated stimulation of scinderin and PKC-mediated stimulation of MAR-CKS (Trifaro et al., 2002). Scinderin also potentiates Ca2+-induced granule secretion in a permeabilized platelet system and inhibitory peptides directed at scinderin inhibited granule release in this same assay (Marcu et al., 1996). MARCKS-derived inhibitory peptides blocks phorbol ester-induced platelet granule release, invoking MARCKS phosphorylation and deactivation in facilitating the disruption of F-actin required for granule release (Elzagallaai et al., 2000, 2001).

#### **SNARE FUNCTION IN PLATELET AND CHROMAFFIN GRANULE EXOCYTOSIS**

Soluble NSF attachment protein receptors, or SNAREs, assemble into complexes to form a universal membrane fusion apparatus (Jahn and Scheller, 2006). Although all cells use SNAREs for membrane fusion, different cells possess different SNARE isoforms. Neurons and neuroendocrine cells use a set of SNAREs that is distinct from those used in non-neuronal cells. In contrast, platelets and chromaffin cells use many of the same chaperone proteins to regulate SNARE-mediated secretion (**Table 3**).

VAMP-8 (endobrevin) is the primary and most abundant vesicular SNARE (v-SNARE) in platelets (Ren et al., 2007; Graham et al., 2009). It is required for activation-induced release of αgranules, dense granules, and lysosomes (Ren et al., 2007) as evidenced by studies using permeabilized human platelets exposed to anti-VAMP-8 antibodies and by evaluation of secretion from VAMP-8−/<sup>−</sup> platelets (Ren et al., 2007). Platelet-mediated thrombus formation relies on ADP and other factors released from platelet granules. VAMP-8−/<sup>−</sup> mice demonstrate decreased thrombus formation upon vascular injury (Graham et al.,2009). Electron microscopy indicates that platelet VAMPs localize primarily to granule membranes (Feng et al., 2002). VAMP-2, -3, -5, and - 7 are also present in platelets. VAMPs 2 and 3 mediate granule release in VAMP-8 deficiency (Ren et al., 2007). VAMP-7 contains a profilin-like longin domain, has been shown to function in neurite extension, and associates with F-actin during cell spreading (Alberts et al., 2006). Granules expressing VAMP-7 move to the periphery of the platelet during spreading and may represent a distinct granule type that functions to provide membrane to cover growing cytoskeletal structures following activation (Peters et al., 2012). Future studies will evaluate the respective roles of VAMP-8 and VAMP-7 in mediating granule release during spreading and identify the participating membrane compartments.

Synaptosomal-associated protein 23 (SNAP-23), a t-SNARE, is required for release from all three types of granules in platelets (Chen et al., 2000; Lemons et al., 2000). Nearly 2/3rds of SNAP-23 associates with the platelet plasma membrane, with the remaining SNAP-23 distributed between the granule membrane and membranes of the OCS (Feng et al., 2002). SNAP-23 contains five palmitoylation sites in its membrane-binding domain. Cleavage of palmitate by acyl-protein thioesterase 1 releases SNAP-23 from platelet membranes demonstrating that SNAP-23 associates with membranes via these palmitoylation sites (Sim et al., 2007). In addition, SNAP-23 associates with the actin cytoskeleton in both resting and activated platelets (Woronowicz et al., 2010). Antibodies to SNAP-23 or addition of an inhibitory C-terminal peptide against SNAP-23 both block dense-granule release in human **Table 3 | SNAREs and SM proteins in platelets and chromaffin cells.**


Essential components of the secretory machinery are highlighted.

Criteria: platelets: Vamp-8, murine knockout; SNAP-23, inhibitory antibodies, inhibitory peptides, overexpression of dominant negative construct; syntaxin-11, FLH4; Munc13-4, murine knockout, FLH3; Munc18-2, FLH5.

Chromaffin: VAMP-2, neurotoxin cleavage; SNAP-25, deletion of C terminus; Syntaxin 1, botulinum neurotoxin C1 and inhibitory antibodies; Munc18-1, murine knockout.

platelets (Chen et al., 2000). In addition, overexpression of dominant negative SNAP-23 inhibits dense-granule release in murine platelets (Gillitzer et al., 2008).

Our understanding of the role of syntaxins, another family of t-SNAREs, in platelet granule release has recently evolved. Platelets express syntaxin-2, -4, -7, -8, -11, and -12. Whiteheart's group identified a patient with Familial Hemophagocytic Lymphohistiocytosis type four (FHL-4) who was deficient in syntaxin-11 and exhibited a significant granule secretion defect. An inhibitory antibody that this group had previously used to demonstrate a role for syntaxin-2 in granule release was found to cross-react with syntaxin-11, further suggesting a role for syntaxin-11 in platelet exocytosis (Ye et al., 2012). They also demonstrated that syntaxin-2 <sup>−</sup>/<sup>−</sup> mice, syntaxin-4−/<sup>−</sup> mice, and double knockout mice all demonstrated normal granule release. On the basis of these results, syntaxin-11 appears to be the primary syntaxin involved in platelet granule release.

In chromaffin cells, VAMP-2 is the primary v-SNARE and is required for efficient, rapid release of granule constituents in response to agonists (**Table 3**). Proteolytic cleavage of VAMP-2 by botulinum neurotoxins A through G or tetanus neurotoxin results in decreased DCV secretion in chromaffin cells (Knight et al., 1985; Schiavo et al., 1992; Xu et al., 1998). VAMP-3 is less efficient than and plays a subordinate role to VAMP-2, only functioning in its absence (Borisovska et al., 2005). VAMP-7 serves a central role in neurite outgrowth in chromaffin-like cells (Coco et al., 1999; Martinez-Arca et al., 2000, 2001), analogous to its putative role in providing membrane for platelet spreading (Peters et al., 2012). Studies in PC12 cells indicate that the NH2-terminal domain of VAMP-7 negatively regulates neurite outgrowth since neurite outgrowth is blocked by overexpression of this domain and enhanced by its deletion (Martinez-Arca et al., 2000, 2001). SNAP-25a has established roles in both docking and priming of DCVs and participates in agonist-dependent fusion of secretory vesicles in complex with both VAMP-2 and syntaxin-1A. Deletion of the C-terminal synaptotagmin-interacting residues of SNAP-25 in PC12 cells results in decreased DCV secretion (Zhang et al., 2002). Adrenal chromaffin cells express syntaxins-1A, -1B, -2, -3, and -4. Viral infection with botulinum neurotoxin C1 cleaves syntaxin-1A, 1B, 2, and 3 resulting in reduced DCV docking at the plasma membrane (de Wit et al., 2006) and an inhibitory antibody to syntaxin-1 decreases catecholamine release in bovine chromaffin cells (Gutierrez et al., 1995).

#### **SNARE CHAPERONE FUNCTION IN PLATELET AND CHROMAFFIN GRANULE EXOCYTOSIS**

Although not members of the exocytic core complex, another important group of proteins involved in degranulation in secretory cells are the Sec1/Munc18-like (SM) proteins which function as SNARE chaperones (Carr and Rizo, 2010). Platelets and chromaffin cells possess a similar repertoire of SM proteins, including members of the Munc13 and Munc18 families. These proteins are SNARE regulators that have no apparent membrane-binding domain, but bind syntaxin upon phosphorylation by PKC (Houng et al., 2003; Schraw et al., 2003) and interact with the regulatory N-terminal sequence of syntaxins (Ashery et al., 2000;Rosenmund et al., 2002). Munc13 family members include Muncs13-1, -2, -3, and 4. These proteins have two C2 (Ca2+-binding) and one C1 (DAG/phorbol ester-binding) domains (**Figure 3**). They interact with SNARE proteins via two Munc13 (mammalian) homology domains, MHDs 1 and 2 (Guan et al., 2008), which are involved

in dissociating Munc18 protein/syntaxin interactions (Sassa et al., 1999), thereby promoting *trans* SNARE complex assembly.

Munc13-4 is the only Munc13 family member found in platelets. It lacks the N-terminal C1 domain present in Munc13-1, -2, and -3 and the MHD2 domain present in the other Munc13 family members, but has a central MHD1 domain and binds directly to syntaxins in platelets via interaction with the syntaxin H3 domain (Boswell et al., 2012). It is ubiquitously expressed, but enriched in cells of the hematopoietic lineage (Song et al., 1998; Feldmann et al., 2003). In platelets, the Munc13-4 interaction with activated Rab27a/b is important for SNARE binding (via MHD1 interaction), granule formation and plasma membrane interaction (**Figure 3**) (Song et al., 1998; Shirakawa et al., 2004; Ishii et al., 2005; Boswell et al., 2012). Boswell et al. (2012) determined that the C2A domain of Munc13-4 is required for Ca2+-dependent SNARE interaction, whereas the C2B domain mediates Ca2+ dependent membrane association. Mutation of Munc13-4 results in another form of familial hemophagocytic lymphohistiocytosis (FHL3) (Feldmann et al., 2003) and Munc13-4 deletion from murine platelets results in complete ablation of dense-granule release and impaired release from α-granules *in vitro* indicating its importance in Ca2<sup>+</sup> regulation of SNARE interactions with the plasma membrane (Ren et al., 2010).

Munc13-4 is a rate-limiting protein for granule exocytosis in both platelets and chromaffin cells. As with platelet granule release, Munc13-4 triggers rapid and efficient release of catecholamines from chromaffin-like PC12 cells (Boswell et al., 2012). Munc13-4 promotes trans-exocytic core complex formation in a Ca2+-dependent manner in both chromaffin cells and platelets. In addition to Munc13-4, Munc13-1 serves a role in DCV secretion in chromaffin cells. Overexpression of Munc13-1 results in increased DCV secretion (Ashery et al., 2000; Stevens et al., 2005) and its interaction with syntaxin-1 is important for DCV priming (Stevens et al., 2005).

Platelets express three Munc18 isoforms: Munc18-1, 18-2 and 18-3. All three isoforms are associated with granule and OCS membranes in resting platelets (Schraw et al., 2003). Al Hawas et al. (2012) recently demonstrated that defects in the Munc18-2 gene result in familial hemophagocytic lymphohistiocytosis type

**FIGURE 3 | Assemblage of SNAREs and SM proteins during platelet granule exocytosis**. Munc18b sequesters syntaxin in an inactive state. Munc13-4 docks opposing membranes via interactions with Rab27a, which also binds Slp1. Activation promotes a conformational change in Munc18b that enables the coiled-coil

domain of syntaxin to form a four-helical bundle with SNAP-23 and VAMP. Mutations in Munc13-4, as in familial hemophagocytic lymphohistiocytosis (FHL)-3, syntaxin-11 (FHL-4), Munc18b (FHL-5), or Rab27a (Griscelli syndrome) result in defective secretion (figure adapted from Flaumenhaft, 2013).

5 (FHL5). These patients demonstrate decreased α- and densegranule secretion and levels of both Munc18-2 and syntaxin-11 were diminished, indicating that Munc18-2 plays a key role in platelet exocytosis and, potentially, a regulatory role toward syntaxin-11.

In chromaffin cells, Munc18-1 participates in granule docking/priming and SNARE engagement via its interaction with syntaxin 1 (Hata et al., 1993; Pevsner et al., 1994). Munc18-1 knock out in embryonic cells results in decreased DCV docking at the plasma membrane (Voets et al., 2001; Gulyas-Kovacs et al., 2007). In addition, syntaxin 1 expression is decreased by 50% in Munc18-1 deficient neurons and chromaffin cells (Voets et al., 2001; Gulyas-Kovacs et al., 2007). Munc18-1 interaction with the "closed" conformation of syntaxin 1 appears to be important for docking of the secretory vesicle at the plasma membrane (Dulubova et al., 1999; Yang et al., 2000; Schutz et al., 2005; Gulyas-Kovacs et al., 2007). However, Munc18-1 interactions with the N-terminal peptide of syntaxin-1 (in the "open" conformation) is required for membrane fusion to occur (Khvotchev et al., 2007; Gerber et al., 2008; Rathore et al., 2010), indicating that Munc18-1 is important in both early and late stages of exocytosis. Munc18-2 also shows affinity for syntaxins-1, -2, and -3 in chromaffin cells. While Munc18-2 rescued the reduced docking phenotype in Munc18-1−/<sup>−</sup> animals, they continued to exhibit impaired vesicle priming (Gulyas-Kovacs et al., 2007). Munc18-3 is ubiquitously expressed and has been implicated in secretion in chromaffin cells. However, Munc18-3 only partially rescued the Munc18-1−/<sup>−</sup> secretion defect in chromaffin cells and deletion of Munc18-3 from chromaffin cells did not cause defects in granule secretion (Gulyas-Kovacs et al., 2007).

#### **THE PLATELET FUSION PORE**

Although there are many methods to evaluate platelet granule release, platelet secretion assays are largely restricted to bulk assays of cargo release (e.g., ADP, serotonin, platelet factor 4) or granule membrane receptor surface expression (e.g., P-selectin, CD63). These assays are inadequate for evaluation of the release of single granules and unable to detect membrane fusion events that occur in the millisecond time frame. Standard electrophysiology using patch-clamp techniques are difficult to apply to the platelet because of their small size and atypical membrane system. More recently, however, platelet investigators are applying some of the same approaches used to evaluate fusion pore dynamics in chromaffin cells. In particular, investigators are using single-cell amperometry to evaluate the release kinetics of single granules from platelets. Carbon-fiber microelectrode amperometry is being used to detect serotonin release from platelets stimulated with thrombin (Ge et al., 2008, 2009, 2010). Tracings indicate previously unrecognized fusion events such as "kiss and run" fusion and foot process formation (Wightman and Haynes, 2004) (**Figure 4**). This approach has enabled an appreciation of nuances of membrane fusion in platelets that have previously gone unrecognized and, more importantly, have enabled investigators to begin to evaluate the molecular mechanisms of pore formation in platelets. Amperometry has recently been used to evaluate the role of dynamin family proteins in platelet and chromaffin cell granule release.

Dynamins are a family of large GTPases that act as mechanoenzymes, demonstrating both oligomerization-dependent GTPase and membrane modeling activities (Piersma et al., 2009). Although originally described as mediators of membrane scission during vesicle endocytosis (Graham et al., 2004; Wegrzyn et al., 2010), dynamin GTPases are now recognized to function in exocytosis (Graham et al., 2002; Tsuboi et al., 2004; Fulop et al., 2008; Anantharam et al., 2010, 2011; Gonzalez-Jamett et al., 2010). In particular, dynamins act immediately upon membrane fusion to regulate the release of granule content. Dynamin and dynamin-related proteins are found in platelets. Dynamin 3 is upregulated during megakaryopoiesis (Reems et al., 2008; Gieger et al., 2011; Wang et al., 2011). Dynamin 2 and dynamin-related protein 1 (Drp1) are present in platelets, but dynamin 1 is not.

Drp1 is phosphorylated upon platelet activation (Koseoglu et al., 2013). Inhibition of platelets using dynasore or MiTMAB, which inhibit the activity of dynamin family proteins, block agonistinduced platelet granule secretion (Koseoglu et al., 2013). The Drp1 inhibitor, mdivi-1, also blocks platelet granule exocytosis. Studies using single-cell amperometry demonstrate that mdivi-1 exposure results in fusion pore instability as evidenced by decreased foot process formation and inefficient pore expansion as evidenced by an increased T1/2 (Koseoglu et al., 2013). These observations implicate dynamin and dynamin-related proteins in platelet fusion pore dynamics. However, the mechanism by which Drp1, which is typically associated with mitochondrial fission, impacts platelet granule release remains to be determined.

Dynamin-mediated pore expansion in chromaffin has been evaluated using total internal reflection fluorescence microscopy and amperometry. In chromaffin cells overexpressing a dynamin I mutant with low GTPase activity, deformations in the membrane associated with fusion are long-lived, indicating defective pore expansion (Anantharam et al., 2011). Chromaffin cells overexpressing a dynamin I mutant with enhanced GTPase activity demonstrate increased pore expansion. These observations have led to a model in which dynamin restricts fusion pore expansion until GTPase activity is stimulated. The higher the GTPase activity, the faster the expansion of the fusion pore (Gerber et al., 2008). Dynamins appear to associate with actin, SNAREs, and synaptotagmin family proteins to participate in fusion pore expansion (Chan et al., 2010; Gu et al., 2010; Anantharam et al., 2012). However, the importance of these associations is poorly understood.

#### **REFERENCES**


Schmid, S. L., Axelrod, D., et al. (2011). A new role for the dynamin GTPase in the regulation of fusion pore expansion. *Mol. Biol. Cell* 22, 1907–1918. doi:10.1091/ mbc.E11-02-0101


#### **CONCLUSION**

Some characteristics of regulated secretion shared between platelets and chromaffin cells are common to all secretory systems. However, these secretory systems also share some unusual features that, if not unique to these cells, are not universally observed among secretory systems. These special commonalities may provide avenues for researchers investigating these cells types to further define these secretory systems. For example, the unusual density of LDCVs and platelet dense granules and their ability to concentrate nucleotides, bioactive amines, and polyphosphates raises the possibility that may use similar transporters. Some similarities in transporters such as VMAT2 have already been described. Further probing could reveal further overlap (e.g., Scl17A transporters in platelet dense granules, multidrug resistance transporters in chromaffin cells, or yet undiscovered transporters that are common to both cells). The ability of chromaffinlike PC12 cells to use VAMP-7 for neurite outgrowth and platelets to use VAMP-7 during spreading speaks to potential underlying similarities between the molecular mechanisms of membrane utilization during shape change. The role of dynamin family proteins in exocytosis is an emerging area of interest in secretion biology and further studies of these two cell types may reveal how they use these mechanoenzymes to regulate fusion pore formation during exocytosis. Historically, the study of the chromaffin cell has advanced more quickly than that of the platelet and has helped direct how platelet biologists have approached the study of granule exocytosis. As the study of platelet exocytosis progresses, understanding this secretory system may help chromaffin cell biologists better understand elements of granule formation and exocytosis in neuroendocrine cells.


for calcium. *J. Exp. Biol.* 184, 183–196.


D. Michelson (London: Academic Press), 343–366.


Fitch-Tewfik and Flaumenhaft Platelet versus chromaffin cell exocytosis

secretion. *Nature* 296, 256–257. doi:10.1038/296256a0


and dendritic outgrowth.*J. Neurosci.* 21, 3830–3838.


Fukuda, M., et al. (2004). Munc13-4 is a GTP-Rab27-binding protein regulating dense core granule secretion in platelets. *J. Biol. Chem.* 279, 10730–10737. doi:10.1074/jbc.M309426200


exocytosis in insulin-secreting cells. *J. Biol. Chem.* 279, 47115–47124. doi:10.1074/jbc.M408179200


neurotoxin sensitivity. *Nat. Neurosci.* 1, 192–200. doi:10.1038/642


**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: 17 April 2013; accepted: 11 June 2013; published online: 26 June 2013. Citation: Fitch-Tewfik JL and Flaumenhaft R (2013) Platelet granule exocytosis: a comparison with chromaffin cells. Front. Endocrinol. 4:77. doi: 10.3389/fendo.2013.00077*

*This article was submitted to Frontiers in Neuroendocrine Science, a specialty of Frontiers in Endocrinology.*

*Copyright © 2013 Fitch-Tewfik and Flaumenhaft . This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.*

# Regulated mucin secretion from airway epithelial cells

#### **Kenneth B. Adler <sup>1</sup> , Michael J. Tuvim<sup>2</sup> and Burton F. Dickey <sup>2</sup>\***

<sup>1</sup> Department of Molecular Biomedical Sciences, North Carolina State University College of Veterinary Medicine, Raleigh, NC, USA <sup>2</sup> Department of Pulmonary Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, USA

#### **Edited by:**

Rafael Vazquez-Martinez, University of Cordoba, Spain

#### **Reviewed by:**

Ricardo Borges, University of La Laguna, Spain Gunnar C. Hansson, University of Gothenburg, Sweden

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

Burton F. Dickey, Department of Pulmonary Medicine, University of Texas MD Anderson Cancer Center, Unit 1462, 1515 Holcombe Boulevard, Houston, TX 77030-4009, USA e-mail: bdickey@mdanderson.org

Secretory epithelial cells of the proximal airways synthesize and secrete gel-forming polymeric mucins. The secreted mucins adsorb water to form mucus that is propelled by neighboring ciliated cells, providing a mobile barrier which removes inhaled particles and pathogens from the lungs. Several features of the intracellular trafficking of mucins make the airway secretory cell an interesting comparator for the cell biology of regulated exocytosis. Polymeric mucins are exceedingly large molecules (up to 3 × 10<sup>6</sup> Da per monomer) whose folding and initial polymerization in the ER requires the protein disulfide isomerase Agr2. In the Golgi, mucins further polymerize to form chains and possibly branched networks comprising more than 20 monomers. The large size of mucin polymers imposes constraints on their packaging into transport vesicles along the secretory pathway. Sugar side chains account for >70% of the mass of mucins, and their attachment to the protein core by O-glycosylation occurs in the Golgi. Mature polymeric mucins are stored in large secretory granules ∼1µm in diameter. These are translocated to the apical membrane to be positioned for exocytosis by cooperative interactions among myristoylated alanine-rich C kinase substrate, cysteine string protein, heat shock protein 70, and the cytoskeleton. Mucin granules undergo exocytic fusion with the plasma membrane at a low basal rate and a high stimulated rate. Both rates are mediated by a regulated exocytic mechanism as indicated by phenotypes in both basal and stimulated secretion in mice lacking Munc13- 2, a sensor of the second messengers calcium and diacylglycerol (DAG). Basal secretion is induced by low levels of activation of P2Y<sup>2</sup> purinergic and A3 adenosine receptors by extracellular ATP released in paracrine fashion and its metabolite adenosine. Stimulated secretion is induced by high levels of the same ligands, and possibly by inflammatory mediators as well. Activated receptors are coupled to phospholipase C by Gq, resulting in the generation of DAG and of IP<sup>3</sup> that releases calcium from apical ER. Stimulated secretion requires activation of the low affinity calcium sensor Synaptotagmin-2, while a corresponding high affinity calcium sensor in basal secretion is not known. The core exocytic machinery is comprised of the SNARE proteins VAMP8, SNAP23, and an unknown Syntaxin protein, together with the scaffolding protein Munc18b. Common and distinct features of this exocytic system in comparison to neuroendocrine cells and neurons are highlighted.

**Keywords: secretion, exocytosis, mucin, mucus, MARCKS, Munc18, Munc13, synaptotagmin**

#### **BIOLOGY AND PATHOPHYSIOLOGY OF AIRWAY MUCUS**

Mucus has physical characteristics on the border between a viscous fluid and a soft and elastic solid (1). These characteristics are conferred by a semi-dilute network of polymerized mucins in water. The secreted, polymeric mucins expressed in the airways are Muc5ac and Muc5b (2, 3). (Note, lower case letters are used to designate non-human mammalian mucins, while MUC5AC and MUC5B designate the human orthologs. In this review, we use lower case letters to designate all mammalian mucins, and only use upper case letters when referring specifically to human data.) In healthy airway mucus, water accounts for about 98% of the mass, mucins for about 0.7%, and salts and small amounts of other macromolecules for the rest. The mucus layer lies atop a denser periciliary layer containing membrane-tethered glycoconjugates, including glycosaminoglycans and membrane-spanning mucins (Muc 1, 4, and 16) (4–6). The mucus layer is continually swept from distal to proximal airways by beating cilia, and is eventually propelled out of the lungs into the pharynx and swallowed, removing entrapped particles, pathogens, and dissolved chemicals. The critical importance of the mucus layer in airway defense is shown by the spontaneous inflammatory lung and nasal disease that develops in mice in which the constitutively produced mucin, Muc5b, has been deleted (1).

In order to replenish the mucus layer, mucins are continuously synthesized and released by secretory cells that form a mosaic with ciliated cells, with similar numbers of both cell types (**Figure 1**, left). In allergic lung inflammation, which appears to be a parasitic defense gone awry (4, 7), the second polymeric airway mucin, Muc5ac, is produced in large quantities (**Figure 1**, center and right). Whereas increased mucin production alone does not appear

(AB-PAS) staining. Center – Numerous large mucin granules are visible in

intraepithelial mucin stored in inflamed airway epithelium induced by brief exposure to an ATP aerosol as described (51). Scale bar is 10µm.

to lead to pathology (8), the production of large amounts of mucin together with its rapid secretion ("mucus hypersecretion") can overwhelm available liquid resulting in formation of excessively viscoelastic mucus that is poorly cleared by ciliary action or cough. When coupled with airway narrowing due to bronchoconstriction in asthma, this can lead to widespread airway closure with serious consequences. Mucus hypersecretion is also an important feature of chronic obstructive pulmonary disease (COPD), cystic fibrosis (CF), and idiopathic bronchiectasis (1).

A key event in mucus secretion is its hydration immediately after exocytosis (4, 9, 10). Mucins are packaged dehydrated in secretory granules, and must adsorb more than 100-fold their mass of water soon after secretion in order to attain the appropriate viscoelasticity for ciliary clearance. Water in the airway lumen is controlled both by the release of chloride through CFTR, CaCC, and Slc26A9, and by the absorption of sodium by ENaC, with water passively following the flux of ions (11–13). Coupling of the secretion of chloride and mucins is accomplished by paracrine signaling through ATP, adenosine, and other extracellular signaling molecules. In CF, mutation of the principal chloride channel, CFTR, results in insufficient luminal water that causes the formation of underhydrated mucus which is excessively viscoelastic and difficult to clear. The underhydration is exacerbated because CFTR is also a channel for bicarbonate, which is needed to chelate calcium (9, 14). In acidic secretory granules of the intestinal epithelium, calcium binds to the N-terminus of MUC2, which is the secreted mucin most similar to MUC5AC in structure, and organizes the mucin in such a way that it can be secreted without entanglement (15). Too little bicarbonate at mucin release prevents normal mucin unfolding and leads to formation of abnormally dense mucus (16); similar mechanisms likely operate in the airway.

In summary, the mucus layer forms a mobile, essential barrier that protects the lungs when it is functioning properly, but dysfunction of the mucus layer plays a prominent role in all of the common diseases of the airways.

#### **MUCIN SYNTHESIS, PROCESSING, AND PACKAGING**

Muc5b/MUC5B is transcribed constitutively throughout the conducting airways from the trachea down to but not including terminal bronchioles (1, 17, 18). Muc5ac is produced in low amounts or not at all in healthy mice, although in humans MUC5AC is produced constitutively in proximal airways (trachea and bronchi) (1). In allergic inflammation the production of Muc5ac increases dramatically (40- to 200-fold) in the airways of mice and in cultured human airway epithelial cells (19–21). Both mucins are produced in the same secretory cells, but even in conditions of severe inflammation they are not produced in small airways, which makes teleologic sense in that their small luminal diameters (<200µm) make them highly susceptible to occlusion.

After translation at the ER, Muc5ac, and Muc5b undergo initial polymerization as homodimers (3). These are among the largest macromolecules encoded in the mammalian genome, and their processing induces a stress response in the ER (22). Proper folding and polymerization require the protein disulfide isomerase Agr2, whose deletion in mice results in absent intestinal mucin (23) and in reduced airway mucin in the setting of allergic lung inflammation (24). The transport of polymeric mucins from the ER to the Golgi and through the Golgi has been little studied, but likely involves modulation of COP-II transport vesicle size to accommodate large cargoes as has been described for collagen (25). In the Golgi, Muc5b undergoes further polymerization in linear chains up to 20 monomers in length (3, 26). The structure of Muc5ac is less well studied, but appears to be more similar to Muc2 that forms branched polymers resulting in formation of a covalent net (9). Both mucins undergo O-glycosylation in the Golgi that result in mature glycoproteins that are more than 70% carbohydrate with a general negative charge due to sulfation or sialylation of many terminal sugars (27).

Export of polymeric mucins from the trans-Golgi and lateral fusion of post-Golgi vesicles to form secretory granules are additional transport steps that have been poorly studied. Similar to the case of COP-II vesicles, post-Golgi clathrin-coated vesicles have recently been shown to be capable of size variation to accommodate large cargo proteins (28). It is probable this mechanism is also utilized for mammalian mucins because, in fruit flies, assembly of large salivary mucin granules requires clathrin and the adaptor AP-1 (29). Mature mucin secretory granules are very large, with a mean diameter of 1µm. Their exocytic fusion is highly regulated by extracellular secretagogues (**Table 1**), as described below. It should be noted that even though additional secretory pathways

**Table 1 | Ligands shown to induce mucin secretion.**


In the left column are ligands reported to induce secretion of mucin from airway surface epithelial cells. In the middle column are receptors for these ligands and whether they act directly on epithelial cells. In the right column are selected references that offer evidence for the activity of the ligands, their receptors, and their cellular localization. The first three rows (white background) show ligands that appear to act directly on epithelial cell surface receptors based upon in vitro and/or in vivo studies; the next four rows (gray background) show ligands that generally act on cell surface receptors, but may activate cells in the airway other than epithelial cells that in turn activate epithelial cells; the last two rows show ligands that act on intracellular targets. PMA, phorbol 12-myristate 13-acetate.

have been described in other cell types, such as a minor regulated pathway for secretion of immature granules and the compound exocytosis of mature granules (30), these are not well-described in airway secretory cells and will not be addressed in this review.

#### **MUCIN GRANULE POSITIONING FOR SECRETION**

The movement of mature mucin granules to the plasma membrane for exocytosis has been the subject of work in numerous laboratories for many years. In the 1990s, work in the Adler laboratory turned to the Myristoylated Alanine-Rich C Kinase Substrate (MARCKS) protein. MARCKS was a known actin-binding protein and protein kinase C (PKC) substrate (31), and it was known that PKC activation enhanced mucin secretion (32), so MARCKS was a logical candidate regulator of mucin granule movement.

MARCKS is a rod-shaped 87 kDa protein that is ubiquitously expressed. Three domains of MARCKS are conserved evolutionarily. First is the Phosphorylation Site Domain (PSD), also known as the "effector domain," a highly basic 25 amino acid stretch containing a number of serine residues that are phosphorylated by PKC. This domain also binds calcium/calmodulin and crosslinks actin filaments. Second is the Multiple Homology (MH2) domain, whose function is unknown. Third is the Nterminal region containing 24 amino acids and a myristic acid moiety involved in binding to membranes. MARCKS knockout mice die at birth or soon afterward, so peptides that might compete with native MARCKS to inhibit its function were generated by the Adler laboratory in collaboration with the Blackshear laboratory. These were tested using normal human bronchial epithelial (NHBE) cells grown in air-liquid interface culture to maintain their well-differentiated state. Peptides identical to the PSD site tended to induce a toxic response, but a peptide identical to the N-terminus had a strong inhibitory effect on mucin secretion induced by a combination of phorbol ester, a PKC activator, and 8-bromo-cyclic GMP, a protein kinase G (PKG) activator (33), or by the more physiologically relevant stimulus UTP. This peptide was named Myristoylated N-terminal Sequence (MANS), and a control missense peptide was named Random N-terminal Sequence (RNS). In contrast to MANS, RNS was without effect on mucin secretion. Additional studies showed that MARCKS phosphorylation in response to protein kinase activation, followed by dephosphorylation catalyzed by a protein phosphatase type 2A (PP2A), were critical to MARCKS function. This was the first publication to show a specific biological function for MARCKS, and suggested a mechanism whereby MARCKS came off the inner face of the plasma membrane when phosphorylated by PKC, then bound to mucin granules at the N-terminus and the cytoskeleton at the PSD site, serving as a bridge for granule transport to the plasma membrane by the cytoskeleton (33).

To examine the function of MARCKS *in vivo*, mice with mucous metaplasia induced by allergic inflammation (see Mucin Exocytosis, below) were then exposed to aerosolized methacholine to induce mucin secretion. Intratracheal pretreatment with MANS dose-dependently inhibited mucin secretion (34), and it attenuated airflow obstruction about 40% (35). Gold-labeling of stimulated cells revealed MARCKS to be morphologically associated with mucin granules, and treatment with MANS but not RNS blocked the association (34). Additional studies performed in mice with human neutrophil elastase instilled in the airways to induce mucous metaplasia showed similar results, with MANS but not RNS attenuating both mucin secretion and airway hyperreactivity in response to serotonin (36).

Subsequent studies have revealed that PKC δ and ε isoforms are involved in stimulated mucin secretion (37, 38), and that PKCδ-provoked secretion depends on phosphorylation of MAR-CKS (38, 39). Another question was the mechanism of translocation of MARCKS from the plasma membrane to mucin granules. Co-immunoprecipitation studies revealed an association between MARCKS and two previously described chaperones – Heat Shock Protein 70 (HSP70) and Cysteine String Protein (CSP) (40). Of interest, there was previously known to be direct and specific interaction of HSP70 with CSP (41). Western blotting and proteomic analysis of mucin granule membranes, ultrastructural immunohistochemistry, and immunoprecipitation experiments showed that MARCKS, HSP70, and CSP form a trimeric complex associated with the granule membrane (42, 43). Functional studies in a bronchial epithelial cell line using siRNA to knock down expression of MARCKS, HSP70, or CSP resulted in the attenuation of stimulated mucin secretion (42).

Additional studies have examined interactions among MAR-CKS, the chaperones, and cytoskeletal proteins. Treatment of NHBE cells with the pyrimidinone MAL3-101, an HSP70 inhibitor, or siRNA against HSP70, attenuated phorbol esterstimulated mucin secretion, and blocked trafficking of fluorescenttagged MARCKS (42, 44). In preliminary studies, cell-permeant peptides that target different domains of CSP were utilized to show that the C-terminus of CSP, rather than the more frequently studied "J" domain, appears to be involved in attachment of MARCKS to mucin granule membranes and resultant secretion. MARCKS has been found to bind both actin and myosin (33), and recent experiments showed that the myosin family involved is Myosin V (45). A possible contributing mechanism of MARCKS action besides granule transport could be the remodeling of apical actin (30). Exocytic Rab GTPases of the 3 and 27 subfamilies interact with the cytoskeleton and catalyze loose tethering of secretory granules to the plasma membrane in other cell types; Rab3D and Rab27A are expressed in airway secretory cells (30, 46), though they have not yet been functionally implicated in mucin secretion or MARCKS interaction. Another possibly important interaction is with VAMP8 that has been identified as the principal t-SNARE in mucin secretion (47) (see Core Exocytic Machinery, below). Preliminary studies from the Adler laboratory show that MARCKS and CSP bind VAMP8 on mucin granules. In summary, MARCKS engages in multiple protein interactions that together help position mucin secretory granules for exocytotic release. For a listing of proteins known to localize to the mucin granule membrane, see **Table 2**.

#### **MUCIN EXOCYTOSIS**

Mucins are secreted into the airway lumen at a low basal rate and a high stimulated rate (1, 30). It is difficult to precisely define the difference in these rates because their measurement depends upon intracellular mucin content, the time interval of observation, and the post-exocytic release of mucins and their maturation to mucus for most assays. Despite these limitations, the rate of stimulated secretion has been generally found to exceed the rate of basal secretion by ∼5-fold over durations of 1 h or less by a variety of techniques (18, 37, 48, 49). The basal rate of secretion matches the basal rate of mucin synthesis in the distal airways of humans and all the airways of mice so that there is little intracellular mucin accumulation in the healthy state (**Figure 1**, left). Small amounts of intracellular mucin in this setting can be detected by sensitive immunohistochemical techniques that involve signal amplification (18, 50), but generally are not detectable by histochemical stains (51). The proximal airways of humans do contain histochemically apparent mucin associated with the constitutive

#### **Table 2 | Proteins associated with airway epithelial mucin granules.**


Proteins that have been found to be associated with mucin granules of airway surface epithelial cells are listed alphabetically in the first column, and references for the association are reported in the second column. See Table 1 in Ref. (43) for a full listing of all proteins found by LC/MS to associate with mucin granules, though not all of these have been validated. ClCa, calcium-activated chloride channel; CFTR, cystic fibrosis transmembrane conductance regulator; CSP, cysteine string protein; HSP, heat shock protein; VNUT, vesicular nucleotide transporter.

expression of MUC5AC (1),but most functional studies of the exocytic machinery have been performed in mouse models so these will be the focus of further discussion. For comparison with *in vitro* systems studied by electrophysiologic (49) and videomicroscopic (52) techniques, the reader is referred to the referenced articles.

A regulated exocytic mechanism mediates both basal and stimulated mucin secretion as indicated by abnormal phenotypes in both basal and stimulated secretion when Munc13-2, a sensor of second messengers (see Extracellular Signaling and the Exocytic Regulatory Machinery), is deleted in mice (18). A defect in basal mucin secretion can be detected as the spontaneous accumulation of intracellular mucin in the absence of increased mucin synthesis (53). To measure stimulated secretion, it is useful to first induce increased mucin production and accumulation (mucous metaplasia) with allergic inflammation (**Figure 1**, center), such as by IL-13 instillation or ovalbumin immunization and challenge (51, 53). A defect in stimulated mucin secretion can then be detected as the failure to release intracellular mucin in response to a strong agonist such as ATP (**Figure 1**, right). Differential effects of the deletion of genes encoding various exocytic proteins on basal and stimulated mucin secretion indicate which proteins participate in which secretory state. In general, deletion of components of the core exocytic machinery give phenotypes in both basal and stimulated secretion, indicating that there is a single core exocytic machine, whereas deletion of components of the regulatory machinery give variable phenotypes, as described below.

#### **CORE EXOCYTIC MACHINERY**

Every step of vesicular transport on the exocytic and endocytic pathways involves the interactions of a four helix SNARE bundle with an SM protein (30, 54, 55). The SNARE proteins impart specificity to the pairing of transport vesicles with target membranes, mediate tight docking of vesicles to target membranes, and induce fusion of vesicle and target membranes when they fully coil. SM proteins provide an essential platform for sequential interactions of SNARE proteins, and also mediate interactions of the SNARE complex with tethering proteins (**Figure 2**). Three of the SNARE helices localize to the target membrane (called t-SNAREs for target SNAREs, or Q-SNAREs since the ionic amino acid of their SNARE domains is generally glutamine), and one SNARE helix is localized on the vesicle membrane (v-SNARE for vesicle SNARE, or R-SNARE since the ionic amino acid is generally arginine).

Syntaxins are Qa SNAREs that can be considered the central component of the core machinery since they initiate formation of the SNARE complex and their structure is ordered even in the absence of interaction with the other SNARE components. The Syntaxin that mediates mucin granule exocytosis remains unknown. Candidates are Stx 2, 3, and 11, all of which have been shown to functionally pair with Munc18b in other cell types since Munc18b has been definitively implicated in airway mucin exocytosis (56) (see below). Efforts are underway in the Dickey laboratory using genetically modified mice to test the roles of these Syntaxins in mucin secretion.

In both yeast and neurons, the Qb and Qc SNAREs involved in exocytosis are contributed by a single protein with two SNARE domains connected by a linker region. In yeast this protein is Sec9, which is essential for cell viability. In neurons the cognate protein mediating axonal synaptic vesicle release is SNAP25 (57). While SNAP25 is essential for post-natal life, brain development to the time of birth is nearly normal. In unpublished work, the Dickey laboratory has obtained evidence that SNAP23 mediates both basal and stimulated airway mucin secretion. SNAP23 is expressed ubiquitously, and knockout mice experience early embryonic lethality (58, 59). However heterozygous knockout mice show spontaneous airway epithelial cell mucin accumulation, indicating a defect in basal mucin secretion, as well as epithelial mucin retention after stimulation with aerosolized ATP, indicating a defect in stimulated secretion. Thus, SNAP23 appears to mediate most or all Qbc function in both basal and stimulated mucin secretion.

Recently, the R-SNARE (v-SNARE) in airway mucin secretion was identified as VAMP8 by immunolocalization to mucin secretory granules, *in vitro* functional analysis by RNA interference, and *in vivo* analysis of knockout mice (47). Both basal and stimulated mucin secretion were reduced by loss of VAMP8, though the defects were not as severe as from the loss of some other exocytic proteins, consistent with the viability of knockout mice, and suggesting that other v-SNAREs also participate in mucin secretion.

The scaffolding function of SM proteins in exocytosis in different cell types is mediated by three Munc18 proteins (54, 56). Munc18a (Stxbp1) and Munc18b (Stxbp2) appear to be paralogs functioning in axonal/apical secretion, whereas Munc18c (Stxbp3) is a ubiquitous isoform functioning in dendritic/basolateral secretion. Munc18a is expressed in neurons and neuroendocrine

**FIGURE 2 | Regulated airway mucin secretion**. Left – In the basal state, mucin granules are thought to become tethered to the plasma membrane by Rab proteins and effectors that have not yet been identified, in the vicinity of components of the exocytic machinery. Center – Activation of heptahelical receptors such as those for ATP (P2Y2) and adenosine (A3R) leads to activation of the trimeric G-protein, Gq, and phospholipase C (PLC), resulting in generation of the second messengers diacylglycerol (DAG) and inositol trisphosphate (IP3). Diacylglycerol activates the priming protein Munc13-2, and IP<sup>3</sup> induces the release of calcium from apical ER to activate Synaptotagmin-2 (Syt2). Munc13-4 also participates in granule priming, and an unknown high affinity calcium sensor likely functions in basal secretion rather than Syt2. Right – Activation of the regulatory Munc13 and Syt proteins leads to full coiling of the SNARE proteins (SNAP23, VAMP8, and an unknown Syntaxin, all shown in black) to induce fusion of the granule and plasma membranes. The interactions of the SNARE proteins take place on a scaffold provided by Munc18b. In other secretory cells that form the basis for this model, exocytic Syntaxins contain four hydrophobic coiled-coil domains that must be opened to initiate secretion (left panel), and during fusion the associated Munc18 protein remains associated only by an interaction at the Syntaxin N-terminus (right panel).

cells, whereas Munc18b is expressed in polarized epithelia. Together, these data suggested that Munc18b mediates airway mucin secretion, and localization and functional data support this. Munc18b is highly expressed in airway secretory cells where it localizes to the apical plasma membrane (56). Munc18b knockout mice are not viable postnatally, but heterozygous knockout mice show an ∼50% reduction in stimulated mucin secretion, indicating that Munc18b is a limiting component of the exocytic machinery (56). These heterozygous mutant mice do not show spontaneous mucin accumulation, unlike heterozygous SNAP23 mutant mice, suggesting that another SM protein besides Munc18b also plays a scaffolding role in basal mucin secretion whereas no other protein besides SNAP23 appears to function as a Qbc SNARE in mucin exocytosis. Ruling out the possibility that Munc18b functions only in stimulated and not basal mucin secretion, conditional mutant mice with Munc18b deleted only in airway secretory cells are viable and show spontaneous mucin accumulation, although preliminary results suggest that the accumulation is less than in Munc13-2 mice.

#### **EXTRACELLULAR SIGNALING AND THE EXOCYTIC REGULATORY MACHINERY**

The extracellular ligands and signal transduction pathways controlling mucin secretion have been studied for longer and in more depth than the exocytic machinery itself (30). The best-studied ligand is ATP that acts on the P2Y<sup>2</sup> receptor to activate Gq and PLC-β1, resulting in generation of the second messengers IP<sup>3</sup> and diacylglycerol (DAG). ATP is released in a paracrine fashion from ciliated cells in response to mechanical shear stress and in an autocrine fashion along with uridine nucleotides from secretory granules (60, 61). The ATP metabolite adenosine acting on the A3 adenosine receptor appears to activate the same Gq-PLC pathway (62). It is possible that other G-protein coupled receptors, such as those sensing serotonin or acetylcholine, also function on airway secretory cells since those ligands induce mucin secretion *in vivo* (34, 36), however they may be acting in a paracrine fashion by inducing contraction of smooth muscle cells leading to the release of ATP that in turn induces mucin release (**Table 1**).

In airway secretory cells, the second messenger IP<sup>3</sup> activates receptors on apical ER to induce the release of calcium. In contrast to excitable cells in which calcium enters the cytoplasm from outside through voltage-gated channels, or secretory hematopoietic cells such as mast cells in which an initial release of calcium from intracellular stores triggers further calcium entry from outside through ICRAC, all of the cytoplasmic calcium involved in exocytic signaling in airway secretory cells appears to come from intracellular stores (30). This may be an adaptation to the fact that the calcium concentration in the thin layer of airway surface liquid is not stable due to the variable release of mucins that carry calcium as a counterion and the variable secretion via CFTR of bicarbonate that chelates calcium. Calcium does enter airway secretory cells from the basolateral surface to maintain intracellular stores, presumably by communication between the basolateral and apical ER since mitochondrial barriers segregate cytoplasmic calcium signals (63). Nonetheless, the chelation of extracellular calcium *in vitro* does not acutely affect mucin secretion. Rough ER at the apical pole of airway secretory cells lies in close apposition to mucin granules (46, 64), which should allow localized calcium signaling to the exocytic machinery through proteins such as Synaptotagmins and Munc13s.

Synaptotagmins are a family of proteins containing two C2 domains capable of calcium-dependent phospholipid binding, of which several members mediate calcium-dependent exocytosis. Using Syt2 knockout mice, we have found that Syt2 serves as a critical sensor of stimulated but not of basal mucin secretion (46). There was no spontaneous mucin accumulation in these mice, consistent with the fact that Syt2 and its close homolog Syt1 inhibit rather than promote synaptic vesicle release at baseline levels of cytoplasmic calcium (65). In contrast, there was a complete failure of ATP-stimulated mucin release in homozygous knockout mice and a dose-dependent failure in heterozygous knockout mice (46). This was a surprising result for several reasons. First, there is no impairment of synaptic vesicle release in heterozygous mutant Syt1 or Syt2 mice, indicating that some structural or functional feature of stimulated exocytosis in airway secretory cells differs from that in neurons to make Syt2 levels limiting, such as the difference in size of the secretory vesicles (50 nm in neurons versus 1000 nm in airway secretory cells) or the concentration of exocytic proteins at the active zone. Second, Syt2 is the fastest among the low affinity, fast calcium exocytic sensors Syt 1, 2, and 9, yet mucin secretion is a slow exocytic process (measured in hundreds of milliseconds) compared to synaptic vesicle release (measured in milliseconds). This suggests that some other feature of Syt2 besides its kinetics makes it a suitable regulator of mucin release. The calcium-sensing protein in basal mucin secretion that performs a role comparable to that of Syt2 in stimulated secretion is not yet known. Nonetheless such a protein likely exists since a second, high affinity calcium sensor functions in neurons and neuroendocrine cells, and basal mucin secretion has been shown to be calcium dependent (66).

Munc13 comprises a family of four calcium and lipid sensing proteins with variable numbers of C1 and C2 domains that function in the priming of secretory vesicles. As mentioned above, Munc13-2 knockout mice show defects in both basal and stimulated mucin secretion, with the basal defect being more dramatic than the stimulated defect (18). Munc13-2 contains a C1 domain that binds the second messenger DAG. Another member of this family, Munc13-4, is also expressed in airway secretory cells (67). Munc13-4 does not contain a C1 domain, though it does contain two C2 domains that may bind phospholipids in a calciumdependent manner. In unpublished results, the Dickey laboratory has found that deletion of Munc13-4 causes a mild defect in stimulated secretion, and that deletion of both Munc13-2 and Munc13-4 together causes a severe (though incomplete) defect in stimulated secretion. Whether a third protein also functions in mucin granule priming to account for the residual secretion or whether mucin granule exocytosis depends only partially on priming function is not yet known.

There are additional targets of the regulation of mucin secretion besides Synaptotagmin and Munc13 proteins, such as PKC that also binds DAG and calcium (37, 38). Here we have focused on components of the exocytic machinery. A full accounting of the regulation of mucin secretion will require further knowledge of second messengers and their targets, together with analysis of their integrated function.

### **SUMMARY**

Airway secretory cells continuously synthesize and secrete polymeric mucins that form a protective mucus layer. Both the synthesis and secretion of mucins are highly regulated, with low basal rates and high stimulated rates for each. Mature mucin granules are positioned for secretion by interactions of MARCKS, CSP, HSP70, Rab proteins, and the cytoskeleton. A core exocytic machine consisting of the SNARE proteins VAMP8, SNAP23, and an unknown Syntaxin, along with the scaffolding protein Munc18b, mediates both basal and stimulated mucin secretion. Regulatory proteins including Munc13-2, Munc13-4, and Syt2 respond to second messengers to control the rate of mucin secretion in response to extracellular signals. These regulatory proteins show differential

#### **REFERENCES**


*Med* (2012) **2**(11):a014159.


activities in basal and stimulated secretion, suggesting that they variably associate with the core machinery depending on the levels of second messengers. Close coordination of mucin production and secretion with physiologic need are essential to lung health since either a deficiency or excess of airway mucus causes disease. The medical importance of airway mucin secretion and its scientific value as a model of large-granule exocytosis in polarized epithelia insure its continued study.

#### **ACKNOWLEDGMENTS**

This work was supported by grants from the U.S. National Institutes of Health HL36982 (Kenneth B. Adler) and HL097000 and HL094848 (Michael J. Tuvim and Burton F. Dickey).

human salivary gland, respiratory tract and endocervix: identification of glycoforms and Cterminal cleavage. *Biochem J* (1998) **334**(Pt 3):685–93.


promotes allergen-induced mucin overproduction. *Am J Respir Cell Mol Biol* (2012) **47**:178–85. doi:10. 1165/rcmb.2011-0421OC


a murine asthma model. *Am J Respir Cell Mol Biol* (2008) **38**:256–62. doi: 10.1165/rcmb.2007-0279RC


fibrosis airway surface liquid volume by phasic shear stress. *Annu Rev Physiol* (2006) **68**:543–61. doi:10.1146/annurev.physiol.68. 072304.112754


*Cell Biol* (2013) **139**:717–26. doi:10.1007/s00418-012-1060-y


Ferraro P, et al. CFTR is expressed in mucin granules from Calu-3 and primary human airway epithelial cells. *Am J Respir Cell Mol Biol* (2013) **49**. doi:10.1165/rcmb.2012- 0419RC

78. Sesma JI, Kreda SM, Okada SF, van Heusden C, Moussa L, Jones LC, et al. Vesicular nucleotide transporter regulates the nucleotide content in airway epithelial mucin granules. *Am J Physiol Cell Physiol* (2013) **304**:C976–84. doi:10.1152/ ajpcell.00371.2012

**Conflict of Interest Statement:** Kenneth B. Adler has an ownership interest in BioMarck Pharmaceuticals that is developing the MANS peptide for commercial use. Michael J. Tuvim and Burton F. Dickey have no potential conflicts of interest.

*Received: 04 June 2013; accepted: 03 September 2013; published online: 18 September 2013.*

*Citation: Adler KB, Tuvim MJ and Dickey BF (2013) Regulated mucin secretion from airway epithelial cells. Front. Endocrinol. 4:129. doi: 10.3389/fendo.2013.00129*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Adler, Tuvim and Dickey. 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.*

### Super-resolution microscopy in studying neuroendocrine cell function

#### *Anneka Bost 1, Mathias Pasche2, Claudia Schirra1 and Ute Becherer <sup>1</sup> \**

*<sup>1</sup> Physiologisches Institut, Universität des Saarlandes, Homburg/Saar, Germany*

*<sup>2</sup> Division of Neurobiology, MRC Laboratory of Molecular Biology, Cambridge, UK*

#### *Edited by:*

*Stephane Gasman, CNRS, France*

#### *Reviewed by:*

*Valentin Nägerl, Université Bordeaux Segalen, France Corey Smith, Case Western Reserve University, USA*

#### *\*Correspondence:*

*Ute Becherer, Physiologisches Institut, Universität des Saarlandes, Kirrberger Strasse, Geb. 59, D-66421 Homburg/Saar, Germany e-mail: ute.becherer@uks.eu*

The last two decades have seen a tremendous development in high resolution microscopy techniques giving rise to acronyms such as TIRFM, SIM, PALM, STORM, and STED. The goal of all these techniques is to overcome the physical resolution barrier of light microscopy in order to resolve precise protein localization and possibly their interaction in cells. Neuroendocrine cell function is to secrete hormones and peptides on demand. This fine-tuned multi-step process is mediated by a large array of proteins. Here, we review the new microscopy techniques used to obtain high resolution and how they have been applied to increase our knowledge of the molecular mechanisms involved in neuroendocrine cell secretion. Further the limitations of these methods are discussed and insights in possible new applications are provided.

#### **Keywords: chromaffin cell, membrane capacitance, amperometry, TIRFM, SIM, PALM, STORM, STED**

#### **INTRODUCTION**

The main function of neuroendocrine cells, e.g., chromaffin cells, is the regulated release of hormones or peptides into the blood stream. This function is well documented in several reviews (Becherer and Rettig, 2006; Stevens et al., 2011; Jahn and Fasshauer, 2012; Kasai et al., 2012). Briefly, regulated exocytosis is a multi-step process controlled by calcium (**Figure 1A**). In order to fuse, large dense core vesicles (LDCVs) containing catecholamines approach the plasma membrane (PM) along actin filaments (Villanueva et al., 2012). They dock to the PM to the target-SNARE (t-SNARE) acceptor complex composed of syntaxin1 and SNAP-25, via synaptotagmin (de Wit, 2010). This process is controlled by Munc18 which acts at several steps during exocytosis (Rizo and Sudhof, 2012). After docking LDCVs undergo maturation reactions in which the vesicular SNAREs, vesicle associated membrane protein 2 and 3, also called synaptobrevin and cellubrevin, associate with the t-SNAREs to form the SNARE core complex. This reaction, that stably binds LDCVs to the PM, is regulated by a variety of proteins such as Ca2+ dependent activator protein for secretion (CAPS), complexin, snapin or tomosyn (Becherer and Rettig, 2006). Upon increase of Ca2<sup>+</sup> above a concentration of 0.5–0.9 μM, fusion is initiated by the interaction of synaptotagmin with the SNARE complex and the PM (Sudhof, 2012). Proteins such as complexin control this reaction. After exocytosis, the LDCV membrane and protein components are taken up via clathrin dependent endocytosis and processed through poorly understood recycling (Becherer et al., 2012). The development of this rather complex model of the exocytosic pathway was enabled by an array of innovative measurement methods that were applied to neuroendocrine cells.

#### **METHODS TO STUDY CHROMAFFIN CELL FUNCTION**

In neuroendocrine cells, regulated exocytosis has been studied since the late 1960s using biochemical methods (Schneider et al., 1967). Among other important findings, researchers were able to describe the Ca2+-dependency of exocytosis (Pollard et al., 1982). The pace of research in this field greatly picked up with the launch of high time resolution measurement methods. In 1982 using patch-clamp electrophysiology, Neher and Marty (1982) were able to measure changes in the membrane capacitance of chromaffin cells that corresponded to either the addition (exocytosis) or subtraction (endocytosis) of vesicular membrane to the PM. Using complex depolarization protocols (Gillis et al., 1996; Voets et al., 1999) or flash photolysis of caged Ca2<sup>+</sup> (Kaplan and Ellis-Davies, 1988; Neher and Zucker, 1993) it was possible to dissect four main functional pools of LDCVs termed reserve pool, unprimed pool (UPP), slowly releasable pool (SRP) and readily releasable pool (RRP) (**Figure 1B**1). It was hypothesized that LDCVs dock to join the UPP and that while priming occurs, LDCVs proceed in a sequential manner through SRP and RRP (Sorensen, 2004). The molecular machinery mediating these reactions was dissected using gain or loss of function assays (Becherer and Rettig, 2006). Ten years after establishing membrane capacitance recording, carbon fiber amperometry was developed (Wightman et al., 1991). With this technique the release of oxidizable neurotransmitters or hormones such as catecholamines from individual LDCVs can be measured with very high temporal resolution (Bruns and Jahn, 2002). The derived kinetics of release were used to identify proteins such as synaptobrevin, that play a role in fusion pore opening during exocytosis (Borisovska et al., 2005) (**Figure 1B**2). By combining carbon fiber amperometry with membrane capacitance recording, it was then possible to discriminate between exo- and endocytosis when they occur simultaneously (Borges et al., 2008). Furthermore, this combination was instrumental in uncovering the role of for example CAPS in loading LDCVs with catecholamines (Speidel et al., 2005).

The main limitation of both methods is that they measure directly the very last step of exocytosis and they provide only little

information about docking. Electron microscopy (EM) is therefore often used to complement this information (Ashery et al., 2000; Yizhar et al., 2004; de Wit et al., 2009). Vesicles are considered morphologically docked if they are touching the PM or located within 30 nm distance to the PM (Verhage and Sorensen, 2008; de Wit, 2010). These vesicles can belong to all three functional pools UPP, SRP or RRP. The development of a fixation method involving high-pressure freezing produced nearly artifact free, highly preserved cell morphology (Studer et al., 2001; Vanhecke et al., 2008). Electron micrographs of chromaffin cells fixed with this method, revealed fine structures at the contact regions between LDCVs and the PM, which might correspond to assembled SNARE proteins (**Figure 1B**4). Due to this increased resolution a new morphological classification was introduced of docked and tethered vesicles (Verhage and Sorensen, 2008). Until now it was not possible to clearly associate a functional pool to those two morphologically different pools of LDCVs, thereby generating quite some confusion in the field.

In the 1990s total internal reflection fluorescence microscopy (TIRFM) was first introduced in the field of exocytosis (Axelrod, 1981; Steyer et al., 1997; Oheim et al., 1998; Oheim and Stuhmer, 2000). It is used to visualize individual LDCVs approaching the PM and their fusion. The principal feature of TIRFM is that a thin evanescent field of light with decaying exponential excitation energy is generated at the interface of the glass coverslip and the cell. Thus, excitation of fluorophores is restricted to a shallow layer close to the PM (**Figure 1B**3). This technique provides an axial (z) resolution well below 100 nm, while the lateral (x, y) resolution is about 250 nm. The highly contrasted pictures of individual fluorescently labeled LDCVs and the possibility to follow them over time raised very high expectations that TIRFM would provide profound insights in the molecular machinery of docking and priming (Steyer et al., 1997; Oheim and Stuhmer, 2000; Johns et al., 2001). This turned out to be much more complex than originally anticipated (Oheim and Stuhmer, 2000; Nofal et al., 2007). TIRFM helped to understand the role of Munc18-1 (Toonen et al., 2006) and Ca2<sup>+</sup> (Pasche et al., 2012) during docking, and the function of tomosyn in priming (Yizhar and Ashery, 2008).

Taken together, using a combination of membrane capacitance recording, carbon fiber amperometry and EM very effectively uncovered the function of several proteins, such as the SNARE proteins or Munc13, which play a role at only one step of exocytosis in chromaffin cells. TIRFM helped to examine the function of certain proteins or substances, such as Munc18 or Ca2+, that mediate several steps of exocytosis. However, the results of these studies become more and more complex to interpret as we investigate the role of proteins that appear to be involved throughout exocytosis, e.g., Synaptotagmin. The question is whether superresolution microscopy can help in this quest by providing a link between morphological (EM) and functional data (membrane capacitance recording, carbon fiber amperometry or TIRFM).

#### **SUPER-RESOLUTION MICROSCOPY**

The aim of super-resolution microscopy is to provide similar resolution as EM but with light microscopy. In light microscopy the wave-like nature of light limits spatial resolution to half the wavelength of the observed light. This so called diffraction barrier was established by Ernst Abbé 140 years ago and is expressed by the formula:

$$d = \frac{\lambda}{2(n\sin\theta)}$$

(with d the diameter of the spot generated by light of the wavelength λ that travels in a medium with refractive index *n* and converges with an angle θ).

All microscopy technologies developed at the end of the twentieth century, such as confocal microscopy or TIRFM, improve the signal-to-noise ratio and thus produced highly contrasted and crisp images that showed a wealth of unprecedented details, but the resolution was still diffraction limited. An early approach was to use deconvolution algorithms to subtract the predicted point spread function (PSF) of individual fluorophores from the image and thus to reduce the contribution of diffracted light. This technique produces images with even better signal-to-noise ratio, but the resolution improvement is modest. Furthermore, due to the use of flawed PSFs, this technique is prone to generate artifacts like non existing signal patterns. The new super-resolution microscopy methods have addressed most of these problems and can achieve a resolution down to 10 nm (Hell, 2009; Dani and Huang, 2010; Tonnesen and Nagerl, 2013a).

#### **STRUCTURED ILLUMINATION MICROSCOPY (SIM)**

SIM achieves super-resolution by extracting fine structural details from the interference of a structure with predetermined illumination patterns. When a periodic illumination pattern, such as stripes, is applied to a fluorescent sample, an interference pattern is generated. The diffraction-limited fringes of this interference pattern, called moiré fringes, contain information about underlying structural pattern of the sample that cannot be observed with conventional light microscopy. By shifting and rotating the illumination pattern, sub-diffraction-limited structural information of the sample can be extracted from Fourier transformations of the resulting interference pattern (Gustafsson, 2000; Heintzmann et al., 2002) (**Figure 2A**). This produces a doubling of both lateral and axial resolution reaching 100 and 300 nm, respectively. The resolution of the calculated image depends on the number of unique raw images acquired with different diffraction patterns/orientations. In order to generate a single plane highly resolved image a minimum of 15 different illumination patterns have to be applied. Currently, the minimum time required for this acquisition is about 300 ms for a single plane and about 8 s for 7μm thick chromaffin cells (Shao et al., 2011). This time frame is incompatible for life imaging of fast moving structures like LDCVs. However, it can easily be used to visualize 3D distribution of LDCVs in fixed cells without the need of lengthy procedures used in EM. Furthermore, it allows performing precise colocalization studies (Fiolka et al., 2012) in immune cells (Brown et al., 2011; Pattu et al., 2011; Matti et al., 2013) and neurons (Pielage et al., 2008; Sheets et al., 2012; Khuong et al., 2013) that might help to uncover interactions between proteins.

#### **PHOTOACTIVATED LOCALIZATION MICROSCOPY (PALM) AND STOCHASTIC OPTICAL RECONSTRUCTION MICROSCOPY (STORM)**

In PALM or STORM the resolution has been improved to much greater extent since they provide a lateral resolution of 10–30 nm and about 50 nm axial resolution (Betzig et al., 2006; Hess et al., 2006; Rust et al., 2006; Huang et al., 2008; Tatavarty et al., 2009; Shim et al., 2012). Both PALM and STORM rely on the property of certain fluorophores that can be photoactivated and switched on and off alternately (Folling et al., 2008; Shim et al., 2012). PALM relies on photoactivated fluorescent proteins (Betzig et al., 2006; Hess et al., 2006) while STORM, also called direct STORM (dSTORM), was developed using cyanide dyes (Bates et al., 2005; Rust et al., 2006; Heilemann et al., 2008). To generate superresolved images, the fluorescence of the sample is first entirely quenched with the normal excitation wavelength of the fluorophore, converting it from a ground state to a metastable dark state. Then fluorescence is slowly reactivated, either by itself or by a mild illumination with light at 405 nm (**Figure 2B**).

#### **FIGURE 2 | Super-resolution microscopy methods to investigate**

**chromaffin cell function.** Basic principles of three super-resolution microscopy methods are depicted on the left, while some exemplary images using the respective technique are shown on the right. **(A)** In SIM microscopy a stripe pattern of light, which is shifted and rotated, is applied to the cell, so that the entire cell is illuminated within several images. These images contain sub-diffraction-limited structural information, which is extracted via computer processing using Fourier transformations. Resulting images have a lateral and axial resolution of 100 and 300 nm, respectively. This can be appreciated on images of a bovine chromaffin cell, expressing NPY-mCherry and Lifeact-GFP labeling LDCVs and the F-actin, respectively. MIP: maximum intensity projection. **(B)** The principle of PALM and STORM exploit the properties of certain fluorophores that can be switched on and off. At first, all the fluorophores are pushed in a metastable dark state by illuminating them with their specific excitation light. Then, few molecules are brought back to the ground state using a mild illumination at usually 405 nm and visualized using their excitation light, which switches them off again. These cycles are repeated up to 10,000 times generating a movie of blinking fluorophores. The labeled structure is then reconstituted by plotting their exact calculated position. This method generates images with lateral resolution of 10 to 30 nm. This is shown on pictures of a bovine chromaffin cell. The LDCVs were marked by NPY-mCherry overexpression via Semliki Forest virus (shown in red) and the native syntaxin1 (shown in green) was labeled via monoclonal antibody (Synaptic Systems GmbH) and Alexa 647 anti-mouse secondary antibody (Invitrogen). The gain of resolution can clearly be appreciated by comparing the pictures of syntaxin1 acquired with TIRFM and the picture of the same cell acquired with dSTORM. Due to dSTORM resolution, we observed that LDCVs were usually not located on syntaxin1 clusters. **(C)** As can be seen on the Jablonski diagram, the energy of an excited fluorescent molecule can be completely depleted by a photon that matches the energy difference between its excited (S1) and the ground electronic state (S0) before spontaneous fluorescence emission occurs. This process effectively depletes the S1 state of a fluorescent molecule by using a depleting laser that has high photon density, and a higher wavelength than the emission wavelength of the fluorophore. In this example a fluorescent protein such as YFP is excited at 514 nm and releases its fluorescent light at around 520 nm. The stimulated emitted photons are not visible for the light detector (photomultiplier tube, avalanche photodiode) as they travel in the same direction as the stimulating laser beam but good emission filters are needed to block scattered light. Using this technique a STED beam consisting of a beam at the excitation wavelength surrounded by a donut shaped red light beam is applied to the probe. Normal emission occurs only from the central spot thus defining the size of the measured voxel. The STED beam scans the entire probe and the super resolved image is generated online. The lateral resolution is generally 20–50 nm. This very high resolution can be appreciated on the images taken from Sieber et al. (2006) in which syntaxin1 was visualized using antibody labeling on membrane sheets generated from syntaxin1 overexpressing PC12 cells. Images in the lower row correspond to a magnification of the small yellow square drawn on the images in the upper row.

Importantly, not all the fluorescence is reactivated simultaneously, instead individual fluorophores are activated stochastically over time. They are visualized by the normal excitation light, which also brings them back in the dark state. A movie of blinking fluorescent molecules is then acquired over several minutes. Their position can be precisely determined using a simple Gaussian fit algorithm, as long as their fluorescent signals do not overlap. The quality of the position determination and thus of the final image resolution depends on the signal-to-noise ratio of the original image, on the internal jittering of the microscope and on the labeling density of the marked structure. The super-resolution image is then reconstructed by plotting the peak position of all blinkingmolecules. Depending on the sample size and the staining quality, 100 to 10,000 images have to be acquired to reconstruct one single image. Due to the large number of images that need to be recorded and their lengthy processing, this technique is not applicable to life cell imaging. However, PALM was recently used to follow single fluorescent proteins over time in living cells (Tatavarty et al., 2009; Izeddin et al., 2011; Sochacki et al., 2012). When comparing PALM to STORM, both produce images with similar resolution although STORM relies on antibody labeling which increases the length between the protein of interest and the fluorescent label (but see Ries et al., 2012). Furthermore, in STORM a toxic reducing buffer is used to reactivate the cyanide fluorophore (but see Klein et al., 2011). On the other hand, PALM relies on the overexpression of proteins tagged to individual fluorescent proteins, thus overexpression artifacts can occur. In brief, despite their respective weakness, PALM and STORM clearly close the gap between EM and light microscopy. They allow very precise localization of proteins in the cell without the need of complex immunogold techniques (Mennella et al., 2012; Macgillavry et al., 2013).

#### **STIMULATED EMISSION DEPLETION (STED) MICROSCOPY**

In contrast to the aforementioned super-resolution microscopy techniques, STED microscopy does not rely on post processing of blurry raw images but rather uses a specific illumination method and photo-physics to generate directly highly resolved images. In STED microscopy the excited fluorescent dye molecules return to the ground state (S0, **Figure 2C**) via the process of stimulated emission, which is induced by a STED laser beam. The wavelength of the STED beam is at the tail end of the emission spectrum of the dye, where it does not excite the dye and where it can be spectrally separated from the spontaneous fluorescence. The fluorescence quenching scales with the intensity of the STED laser beam. The trick is to focus a depletion laser into a donut shape and superimpose this onto the focused laser excitation spot (**Figure 2C** middle). The spot in the middle of the donut, from which the normal emission occurs, can be made as small as 5.8 nm in diameter (Rittweger et al., 2009) but is usually 20 to 50 nm large. The emission depletion light beam can also be formed in an elongated spherical shape to limit the emission to a volume of 45 nm lateral and 108 nm axial resolution (Wildanger et al., 2009). The generation of an image is similar to classical confocal microscopy but due to the very small voxel size, the scanning speed is relatively slow. Improved hardware allowed the reduction of the excitation beam power and the implementation of live imaging. In 2008, Westphal et al. (2008) visualized moving recycling synaptic vesicles labeled with antibody and in the same year the technique was further adapted to visualize fluorescent proteins in living cells (Hein et al., 2008; Nagerl et al., 2008). Other developments made dual color imaging possible enabling colocalization studies (Donnert et al., 2007; Pellett et al., 2011; Tonnesen et al., 2011). STED provides a resolution that is comparable to EM to identify fine structures, such as Bruchpilot, at the active zone of the drosophila neuromuscular junction (Kittel et al., 2006) or dynamic changes of dendritic spines (Nagerl and Bonhoeffer, 2010; Blom et al., 2011; Urban et al., 2011; Tonnesen and Nagerl, 2013b).

#### **SUPER-RESOLUTION MICROSCOPY TO STUDY CHROMAFFIN CELLS: PRESENT AND FUTURE**

The impact of super-resolution microscopy in the field of neuroendocrinology is just about to spark. STED microscopy has been used to verify that the size of LDCVs was unaltered upon overexpression of a mutated Munc18-1 in neuroendocrine cells (Jorgacevski et al., 2011). SIM was applied to show that a mutated form of synaptobrevin was correctly sorted to LDCVs upon overexpression in mouse chromaffin cells (Borisovska et al., 2012). Furthermore, SIM was used to reveal that the cellular distribution of NPY-mCherry labeled LDCVs was normal and that it was not affected by t-SNARE and Munc18-2 overexpression (Hugo et al., 2013). PALM was used to uncover the size and the shape of clathrin coated pits in PC12 cells during reuptake of vesicular acetylcholine transporters (Sochacki et al., 2012). However, super-resolution microscopy has primarily been used to examine clustering of syntaxin1 and SNAP 25 in the PM. The morphology and the dynamics of syntaxin1 clusters were studied in cracked open PC12 cells using STED (Sieber et al., 2007). In contrast to what was shown using confocal microscopy, Lopez et al. (2009) and Bar-On et al. (2012) used PALM to demonstrate that syntaxin1 and SNAP-25 clusters have a different morphology and that their colocalization is weak in PC12 cells. Additionally, PALM helped to establish that clustered SNARE proteins are not involved in LDCV docking or fusion (Yang et al., 2012). Using a very elegant combination of STED microscopy and Förster resonance energy transfer (FRET) Rickman et al. (2010) showed that overlapping t-SNARE clusters can contain fully assembled t-SNARE acceptor complexes. Finally, PIP2 and PIP3 clustering, that are believed to play a role in the t-SNARE organization, have been investigated using STORM (Wang and Richards, 2012).

Super-resolution microscopy will help us to understand the detailed molecular interactions in chromaffin cells function. As discussed in the introduction, one crucial issue is to clearly demonstrate the correlation between the functional and the morphological data on docking. Another important aspect in chromaffin cell research is to uncover at which time point during the exocytotic process, docking and priming factors bind to the release machinery of LDCVs. FRET is the tool of choice to demonstrate protein interactions in living cells. However, with the remarkable exceptions of Lam et al. (2010) and Zhao et al. (2013), several relatively inconclusive trials using FRET were made to study the interaction of the SNAREs during exocytosis. Thus, it is unlikely that FRET can be used to study the interaction of priming or docking factors with either the LDCVs or the SNARE core complex. One solution to address these problems might be to use a combination of super-resolution microscopy and EM, doing correlative light-electron microscopy (Sjollema et al., 2012). This technique will help us to uncover a relationship between the distance of LDCVs to the plasma membrane and their association with any of these proteins. Finally, aspects of LDCV biogenesis or protein recycling might be better understood using methods such as SIM in combination with molecular manipulations such as gene deletion or protein overexpression.

#### **ACKNOWLEDGMENTS**

Research in our lab has been funded by grants from the Deutsche Forschungsgemeinschaft (SFB 530 and SFB 894) and by competitive, local funding of the Saarland University.

#### **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: 10 June 2013; accepted: 04 November 2013; published online: November 2013. 25*

*Citation: Bost A, Pasche M, Schirra C and Becherer U (2013) Super-resolution microscopy in studying neuroendocrine cell function. Front. Neurosci. 7:222. doi: 10.3389/fnins.2013.00222*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience.*

*Copyright © 2013 Bost, Pasche, Schirra and Becherer. 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.*

### Exocytosis through the lens

### **Alicja Graczyk and Colin Rickman\***

Institute of Biological Chemistry, Biophysics and Bioengineering, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Luke Chamberlain, University of Strathclyde, UK Derek Toomre, Yale University, USA

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

Colin Rickman, Institute of Biological Chemistry, Biophysics and Bioengineering, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, UK e-mail: c.rickman@hw.ac.uk

Exocytosis, the process in which material is transported from the cell interior to the extracellular space, proceeds through a complex mechanism. Defects in this process are linked to a number of serious illnesses including diabetes, cancer, and a range of neuropathologies. In neuroendocrine cells, exocytosis involves the fusion of secretory vesicles, carrying signaling molecules, with the plasma membrane through the coordinated interplay of proteins, lipids, and small molecules. This process is highly regulated and occurs in a complex three-dimensional environment within the cell precisely coupled to the stimulus. The study of exocytosis poses significant challenges, involving rapidly changing, nano-scale, protein–protein, and protein–lipid interactions, at specialized sites in the cell. Over the last decade our understanding of neuroendocrine exocytosis has been greatly enhanced by developments in fluorescence microscopy. Modern microscopy encompasses a toolbox of advanced techniques, pushing the limits of sensitivity and resolution, to probe different properties of exocytosis. In more recent years, the development of super-resolution microscopy techniques, side-stepping the limits of optical resolution imposed by the physical properties of light, have started to provide an unparalleled view of exocytosis. In this review we will discuss how advances in fluorescence microscopy are shedding light on the spatial and temporal organization of the exocytotic machinery.

**Keywords: exocytosis, super-resolution microscopy, palmitic acids, STED, storm, SNARE proteins,membrane fusion**

#### **INTRODUCTION**

Regulated exocytosis is a fundamental process of multicellular life, permitting cells with diverse functions, and properties to act in a coordinated manner. In vertebrates, this cell–cell communication occurs over short distances and between small numbers of cells, for example in neurons, or over longer ranges encompassing many cell types and organs in the case of neuroendocrine signaling (1). In all examples of regulated exocytosis, the final stages of membrane fusion, lipid mixing, and content release are catalyzed by the coordinated action of the SNARE (soluble NSF attachment protein receptor) proteins (2–4). In neurons and neuroendocrine cells three SNARE proteins act synergistically to drive membrane fusion; syntaxin and SNAP-25 on the plasma membrane, and synaptobrevin in secretory vesicles (4–6). Together, these three proteins form a four helical complex, which drives the merger of the plasma and vesicular membranes (7–9). *In vitro* biochemical studies have described many of the individual stages of SNARE protein action; through the formation of a stable binary intermediate, the zippering of the SNARE complex, the formation of cis and trans complexes, and established the SNAREs as the minimal machinery required for exocytosis (6, 8–13). In parallel, *in vivo* studies have demonstrated the importance of the SNARE proteins in the physiological development andfunctioning of secretory cells and organs (14–16).

This complex cascade of protein–protein interactions provides challenges and opportunities to investigation through the use of optical microscopy (**Figure 1**). Optical microscopy inherently provides spatial information, and with improvements in detection technologies, enhanced sensitivity, and speed. By sampling specific wavelengths of visible light, in fluorescence microscopy, it is possible to discriminate multiple labels within a sample (17). This has been used extensively to study exocytotic events in a large number of secretory cell types from dissociated single cells up to intra-vital imaging in whole organisms. The use of fluorescence microscopy provides a huge improvement over electron microscopy, which is incompatible with live cell imaging and often suffers from an inability to discriminate between objects in the resulting image. However, fluorescence microscopy achieves this at a cost; significantly lower resolution. While electron microscopy is ultimately able to resolve objects to the sub-nanometer range, permitting the generation of protein structures with atomic resolution, fluorescence microscopy is typically limited in resolving power to hundreds of nanometers. The resolution on an optical microscope is limited by two factors, aberration and diffraction, which both serve to blur the image and limit the ability to distinguish two adjacent points (17). Aberration can be corrected through the expensive optics found on modern microscopes, but diffraction is ultimately governed by the wavelength of light and the aperture of the objective according to the Abbe formula and Rayleigh criterion (17, 18). Most fluorophores used in cellular imaging emit light in the wavelength range of 400–700 nm which results in a lateral limit of resolution for standard diffraction-limited microscopes of between 170 and 300 nm respectively (**Figure 2**). Now, new technological breakthroughs are pushing the limits in resolution in the spatial and temporal domains revising our understanding of the process of neuroendocrine exocytosis. This short review will cover some of the highlights of how microscopy has advanced our understanding of neuroendocrine secretion.

#### **TEMPORAL INVESTIGATION OF THE PROCESS OF EXOCYTOSIS**

A key characteristic of all regulated exocytosis, regardless of the type of specialized cell studied is that it is a rapid process. Globally, for all secretory vesicles in a cell the rate of cargo release may be slow and multiphasic, but at the single vesicle level the process of bilayer merger and cargo release occurs over a very fast timescale (19, 20). To examine a dynamic process using fluorescence microscopy requires a high temporal resolution. This is difficult to achieve using confocal laser scanning microscopy, due to the requirement for scanning (**Figure 2**). Widefield techniques, which provide spatial information through sensitive cameras can

**FIGURE 1 | A schematic representation of the process of neuroendocrine exocytosis as observed using diffraction-limited fluorescence microscopy**. The synaptic vesicles (red) with the cargo (pink) undergo fusion with the plasma membrane (blue). The color gradients show the areas of the sample acquisition: CLSM (green), WF (red), and TIRFM (blue). The principle of Förster resonance energy transfer (FRET) microscopy is presented with the use of yellow ovals [located on synaptobrevin (green)] and red ovals [located on SNAP-25 (navy blue)]. The Förster resonance energy transfer occurs only when the distance between the donor (yellow) and the acceptor (red) fluorophores are in the range of 1–10 nm. The third SNARE protein, syntaxin, is denoted in orange.

achieve high temporal sampling frequencies at the expense of axial resolution (17). A major advance was achieved through the invention of the total internal reflection fluorescence microscope (TIRFM), which uses total internal reflection to establish an evanescent wave at the coverslip (21). This permits the high temporal resolution of widefield camera-based microscopy but in a thin axial section at the plasma membrane. One of the first applications of this technique was to observe the fusion of secretory vesicles with the plasma membrane (19, 22, 23). Initial experiments examining the fusion process utilized acidic dyes, such as Acridine orange, but these were soon superseded by fluorescently labeled cargo molecules (24, 25). In these early studies it was clear that, at rest, the secretory vesicles are highly dynamic in the cell interior while vesicles at the plasma membrane were largely immobile (22, 26, 27). This supported the hypothesis from electrophysiological data that different pools of secretory vesicles exist with different release properties (20). Secretory vesicles at the plasma membrane do not exhibit free diffusion but instead demonstrate a caged, or tethered behavior, termed morphological docking. Upon stimulation a proportion of these morphologically docked vesicles fuse with the plasma membrane releasing their cargo (28). Using high speed imaging under TIRFM illumination, Degtyar and colleagues demonstrated that immediately prior to fusion, secretory vesicles undergo a rapid lateral movement (29). The mechanisms behind this are unclear but it may serve to sample a larger area of the plasma membrane, enhancing the probability of SNARE interactions, or be a direct result of the formation of the SNARE complexes themselves. While TIRFM provides high temporal resolution of exocytotic events at the plasma membrane, it cannot observe events over the entire cell surface or at depth, in clusters of cells (17). Spinning disk confocal microscopy has been employed to study vesicle fusion throughout cells clusters and in polarized pancreatic acini (30, 31). These studies are beginning to move the temporal study of vesicle dynamics and fusion from single cells toward whole organs and beyond.


**FIGURE 2 | Summary of diffraction-limited and super-resolution microscopy approaches**. Seven microscopy techniques are detailed: confocal laser scanning microscopy (CLSM), stimulated emission depletion microscopy (STED), continuous wave STED (CW-STED), structured illumination microscopy (3D-SIM), total internal reflection fluorescence microscopy (TIRF), photoactivation localization microscopy (PALM), and stochastic optical reconstruction microscopy (STORM). The table shows the typical operational emission wavelengths, lateral and axial resolutions, the temporal resolution of data acquisition, and the sampling depth range. For PALM and STORM microscopy the typical lateral localization precision is stated. For further information regarding the referenced techniques (63).

While secretory vesicles provide an ideal object to image, being larger and more sparsely distributed than the limits imposed by diffraction, they are only one part of the story of exocytosis. Imaging of proteins is confounded by limited resolution, however, this has not prevented the development of innovative solutions to probe SNARE protein function with high temporal resolution. Fluorescence recovery after photobleaching (FRAP) has been widely used to examine the molecular motion of SNARE proteins on the plasma membrane (32). By selectively bleaching fluorescently labeled SNAREs in a sub region of the plasma membrane the recovery of fluorescence in this region through diffusion can be monitored. Sieber and co-workers used FRAP to measure the diffusion of syntaxin molecules on the plasma membrane to derive a dynamic model of SNARE organization (33). In a subsequent study by the same group FRAP was used to probe the intermolecular interactions of SNAP-25 and syntaxin (34), supporting *in vitro* observations of a 1:1 binary intermediate (11, 13). FRAP provides a global average for the diffusion rate of the population of molecules being studied. Due to the high density of the molecules, their individual motion cannot be resolved using diffraction-limited microscopy. However, using an adaptation of a scanning confocal microscope in which the laser beam is parked to a single point in space, diffusion of molecules through the beam can be monitored. This approach called fluorescence correlation spectroscopy (FCS) has been used extensively in *in vitro* reconstituted studies of SNARE protein organization (35–37). FCS studies showed that synaptobrevin and syntaxin both preferentially sequester in the liquid-disordered phase of giant unilamellar vesicles (35). This argues against a classical raft hypothesis for SNARE protein organization, however, the situation in cellular membranes may be more complicated due to the diversity of lipid species present.

#### **RESOLVING THE MOLECULAR ORGANIZATION AND INTERACTIONS UNDERLYING EXOCYTOSIS**

The proteins responsible for driving the process of exocytosis are un-resolvable due to their size and high density in neuroendocrine cells using diffraction-limited microscopes. In the last 5 years there has been a revolution in cell imaging with the introduction of super-resolution microscope techniques, which have pushed the ability to resolve objects through supramolecular imaging down to single molecule detection (38) (**Figure 2**). These techniques can be broadly divided into two classes; hardware approaches and localization approaches. Stimulated emission depletion microscopy (STED) and structured illumination microscopy (SIM) achieve the enhancement in resolution through alterations in the illumination of the sample (39–41). This achieves a two to sixfold enhancement in the optical resolution of the microscope over the theoretical limit imposed by the diffraction of light. STED can be implemented using either pulsed lasers or constant wave lasers (CW-STED) (42); the former providing enhanced lateral resolution while the latter provides simpler implementation. The second class of super-resolution techniques utilize localization of emission to achieve single molecule imaging (43). These single molecule localization microscopy (SMLM) techniques include photoactivation localization microscopy (PALM), ground state depletion with individual molecule return (GSDIM), and stochastic optical reconstruction microscopy (STORM and dSTORM) (44–47). All of these SMLM approaches take advantage of the ability to switch molecules between light and dark states to decrease the number of fluorescent molecules observed at one time. By repeated recording it is then possible to observe many tens of thousands of individual, fluorescently labeled, proteins and compute their position with a precision of 5–20 nm.

These super-resolution techniques have been utilized to investigate the organization of the plasma membrane secretory machinery. Over the last decade it has become clear that the plasma membrane SNAREs, syntaxin, and SNAP-25 are not uniformly distributed over the plasma membrane, but instead are observed to exists in a clustered morphology (13, 48–53). These studies used diffraction-limited techniques and largely agreed that clusters were of the order of 200 nm in diameter. It is important to note that using these microscopes, as a result of diffraction of the light through the optics, any object of approximately 200 nm or less would appear this size. Regardless, a number of important observations were made with relation to the dependency of the integrity of these clusters on cholesterol and their partial colocalization with secretory vesicles (48–50). In 2007 a major advance was reported using STED microscopy, which showed that syntaxin forms clusters of approximately 50 nm in size. Based on computer modeling this supramolecular assembly was hypothesized to contain around 70 molecules of syntaxin and required functional SNARE domains of syntaxin (33). In a subsequent study STED was used to probe the involvement of phosphatidylinositol (4,5) bisphosphate (PIP2) in syntaxin clustering. This study concluded that cluster formation required the combined clustering of PIP2 and syntaxin acting synergistically together (54). Importantly, the measured size of the clusters falls at the limit of accuracy of the STED microscope used and so this can only act as an upper limit to the size of a cluster. More recently, SMLM has been employed to investigate SNARE organization on the plasma membrane (55, 56). This goes beyond the observation of multi-protein structures to localize the individual molecular components. dSTORM imaging demonstrated a clustered morphology for both syntaxin and SNAP-25 (55, 56). This non-homogeneous distribution was also observed using PALM and recapitulated in live cells using single particle tracking PALM (sptPALM) (56). This latter technique also observed that syntaxin and SNAP-25 molecules cannot freely diffuse on the plasma membrane, but instead are restricted to as yet undefined microdomains. Interestingly using these superresolution molecular techniques, the secretory vesicles were not localized to the higher density plasma membrane SNARE clusters (56). This appears at odds with the previous observation of partial colocalization with SNARE clusters using diffraction-limited techniques (48–50). However, this apparent discrepancy is most likely due to the level of "detail" resolved using the respective technique.

Although super-resolution techniques provide unparalleled resolution they are limited to describing the position of protein molecules with no information on interaction status. However, it is possible to probe interactions of molecules over distances below 10 nm using Förster resonance energy transfer (FRET) microscopy (57). This has been employed to study protein–protein interactions between the SNAREs, providing key insights into how these molecules interact in a cellular environment and how this is regulated (54, 58–60). There are two main approaches for the determination of FRET; intensity based techniques including sensitized emission and acceptor photobleaching, and fluorescence lifetime imaging (FLIM) (17). All of these techniques measure the effect of non-radioactive energy transfer between the donor and acceptor molecules. In terms of SNARE clustering, FRET has successfully been used to examine protein clustering in artificial liposomes, observing clustering through exclusion from high cholesterol regions as well as probing the intra-cluster interactions of syntaxin and SNAP-25 (52, 61, 62).

#### **REFERENCES**


fusion thermally induced by soluble N-ethyl maleimide sensitive-factor attachment protein receptors (SNAREs). *Biophys J* (2004) **87**:3569–84. doi:10.1529/biophysj.104.048637


#### **CONCLUSION**

The development of new microscopy techniques with ever higher temporal and spatial resolution has mirrored the requirements and advancements of the field of neuroendocrine secretion. This biological system serves as an excellent test bed for many of the new emerging technology as a result of our already considerable understanding and also the specific problems and opportunities offered. The next big step is to combine super-resolution, high speed, and functional imaging into single experiments. The iterative process of technological advancement and new biological findings will continue until the ultimate experiment of being able to watch a complete single fusion event with sub-molecular resolution of multiple protein components is achieved.


at nanometer resolution. *Science* (2006) **313**:1642–5. doi:10.1126/ science.1127344


Halemani ND, et al. t-SNARE protein conformations patterned by the lipid micro-environment. *J Biol Chem* (2010) **285**:13535–41. doi:10. 1074/jbc.M109.091058


munc18-syntaxin 1 interaction. *J Biol Chem* (2007) **282**:12097–103. doi:10.1074/jbc.M700227200


**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: 07 June 2013; accepted: 02 October 2013; published online: 17 October 2013.*

*Citation: Graczyk A and Rickman C (2013) Exocytosis through the lens. Front. Endocrinol. 4:147. doi: 10.3389/fendo.2013.00147*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Graczyk and Rickman. 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.*

# Imaging large cohorts of single ion channels and their activity

### **Katia Hiersemenzel, Euan R. Brown and Rory R. Duncan\***

Edinburgh Super-Resolution Imaging Consortium (ESRIC), Institute of Biological Chemistry, Biophysics and Bioengineering, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK

#### **Edited by:**

Stephane Gasman, Centre National de la Recherche Scientifique, France

#### **Reviewed by:**

Silvio O. Rizzoli, Göttingen Graduate School for Neurosciences, Germany Ludovic Richert, Centre National de la Recherche Scientifique, France

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

Rory R. Duncan, Edinburgh Super-Resolution Imaging Consortium (ESRIC), Institute of Biological Chemistry, Biophysics and Bioengineering, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, UK e-mail: r.r.duncan@hw.ac.uk

As calcium is the most important signaling molecule in neurons and secretory cells, amongst many other cell types, it follows that an understanding of calcium channels and their regulation of exocytosis is of vital importance. Calcium imaging using calcium dyes such as Fluo3, or FRET-based dyes that have been used widely has provided invaluable information, which combined with modeling has estimated the subtypes of channels responsible for triggering the exocytotic machinery as well as inferences about the relative distances away from vesicle fusion sites these molecules adopt. Importantly, new superresolution microscopy techniques, combined with novel Ca2<sup>+</sup> indicators and imaginative imaging approaches can now define directly the nano-scale locations of very large cohorts of single channel molecules in relation to single vesicles. With combinations of these techniques the activity of individual channels can be visualized and quantified using novel Ca2<sup>+</sup> indicators. Fluorescently labeled specific channel toxins can also be used to localize endogenous assembled channel tetramers. Fluorescence lifetime imaging microscopy and other single-photon-resolution spectroscopic approaches offer the possibility to quantify protein– protein interactions between populations of channels and the SNARE protein machinery for the first time. Together with simultaneous electrophysiology, this battery of quantitative imaging techniques has the potential to provide unprecedented detail describing the locations, dynamic behaviors, interactions, and conductance activities of many thousands of channel molecules and vesicles in living cells.

**Keywords: PALM, storm, STED, super-resolution,TIRFM, imaging, microscopy, ion channel**

#### **EXOCYTOSIS**

Ion channel biology is central to all physiology and regulated exocytosis, the process of secretion in specialized cells such as neurons and neuroendocrine cells, underlies physiological cell-to-cell signaling (1). Mis-regulation of exocytosis leads variously to a number of different conditions, including diabetes (2), neurological disorders (3), growth and sleep disorders (4), and asthma (5). Regulated exocytosis does not proceed spontaneously, but requires an influx of calcium ions through voltage-gated calcium channels (VGCCs), with the final fusion event at the plasma membrane driven by synaptotagmin-dependent mechanotransduction (6, 7) to the core SNARE complex comprising syntaxins, SNAP-25/23/29 (both target, or t-SNAREs) and synaptobrevins (a vesicular, or v-SNARE) (8–10) as well as SM proteins, such as munc18 (11–14). In both neurons and endocrine cells the voltage-dependent calcium influx is typically shaped by a classical negative feedback role for large-conductance calcium- and voltage-activated potassium (BK) channels (15–17). Thus disruption of either VGCC and/or BK channel function also leads to defects in physiology (18–20) [for example, chronic inflammatory pain is associated (21) with exon-18 splice variants (22) that lack the putative motif ("synprint") (23) in N-type VGCCs that is thought to interact directly with the secretory machinery]. Central to all these models is the requirement for secretory vesicles to be appropriately localized with the secretory machinery and VGCC and BK channels. Our understanding of ion channel biology has been defined largely by electrophysiological approaches, due to their high temporal resolution and single molecule sensitivity. Recent, rapid advancements in super-resolution imaging now offer the opportunity to test directly long-standing hypotheses regarding ion channel locations, interactions, dynamics, and compositions in living cells.

#### **VOLTAGE-GATED CALCIUM CHANNELS AND EXOCYTOSIS**

Voltage-gated calcium channels are the voltage sensors that convert cell depolarization into a cytoplasmic calcium signal, which subsequently triggers regulated exocytosis. A multiplicity of calcium channel subtypes is expressed in excitable cells (including neurons, chromaffin, and pancreatic beta cells as excitable secretory cell types) (24–26). Seven genes encode high voltage-activated channels (HVA) L (Ca<sup>v</sup> 1.1–1.4), P/Q (Ca<sup>v</sup> 2.1), N (Ca<sup>v</sup> 2.2), and R (Ca<sup>v</sup> 2.3) type. Three genes encode low voltage-activated channels (LVA, Ca<sup>v</sup> 3.1–3, known as T-type channels). In neurons and endocrine cells VGCCs regulate a variety of other fundamental cellular processes in addition to controlling vesicle fusion; decades of research have defined electrophysiologically the relative contribution of each subtype to secretion (19, 25). For example, in mouse adrenal chromaffin cells, responsible for catecholamine (e.g., adrenaline) secretion, and in pancreatic beta cells, which release insulin in response to elevated blood glucose, N- and Ltype calcium channels play a major role in both exocytosis and in other cellular signaling such as shaping the action potential and pacemaker currents in chromaffin cells but N-, P/Q-, and R VGCCs are also present. Importantly, our understanding of the spatial organization of the different VGCC subtypes is immature (**Table 1**) and in this respect their interactions and spatio-temporal patterning in the functional- and molecular-vicinity of the fusion complex is of particular interest.

#### **LARGE-CONDUCTANCE CALCIUM ACTIVATED POTASSIUM CHANNELS AND EXOCYTOSIS**

As BK channels, encoded by a single gene, have a key role in modulating the calcium signal that leads to vesicle fusion (18–20) there has been considerable effort in trying to estimate the distances between Ca2<sup>+</sup> and K<sup>+</sup> channels, their inter-channel interactions (physical and functional) (16, 30, 63), VGCC and BK interactions with the exocytotic machinery itself, particularly with syntaxin1 (21, 31–33), and in the distances between these channels and large dense-cored vesicles (64). Mathematical modeling of calcium concentration nano-domains near the intracellular mouth of channels, combined with the determination of the Ca2<sup>+</sup> sensitivity of the secretory machinery (i.e., synaptotagmin) have led to the concept that secretory vesicles reside very close to VGCCs in the membrane, and slightly further away from BK channels (34– 37, 64). Determining the precise concentration of Ca2<sup>+</sup> required to elicit exocytosis is confounded by spatial heterogeneity in the cell that to date has been impossible to measure accurately; however, it is likely that fusion competent vesicles in secretory cells are targeted somehow to the sites at the membrane experiencing high Ca2<sup>+</sup> concentrations – i.e., very close to the mouth of a calcium channel (34).

#### **FUNCTIONAL COUPLING OF ION CHANNELS WITH EXOCYTOSIS**

Our understanding of the functional coupling of ion channels with the secretory apparatus comes from diffraction-limited microscopy, binary biochemical binding determined *in vitro*, and electrophysiology. The first can only resolve, at the very best, clusters of ion channels and approximate distances in images in the order of 250 nm. *In vitro* biochemistry, whilst invaluable, cannot deliver the "where's and when's" of interactions in cells and so overlooks the key elements of spatio-temporal regulation. Electrophysiology can resolve single ion channel activities, or entire cell cohorts of activity, but with limited spatial resolution. It thus remains unknown how: (i) membrane cohorts of single ion channels are spatially distributed, (ii) the proportion of *active* ion channels compared to the total pool (Ca2<sup>+</sup> and K+) reside within functionally meaningful distances of fusion competent- and/or incompetent-vesicles, (iii) how the dynamics of channel activity may correlate with their spatial pattern *and/or* interactions with the SNARE molecular machinery in intact cells, and (iv) whether every channel at the membrane is functional. It is clear that new tools are required to address these questions.

#### **SUPER-RESOLUTION IMAGING AND EXOCYTOSIS**

The membrane-trafficking field has a strong history of using cutting-edge techniques and imaging is no exception. Superresolution microscopy is an emerging powerful tool to further research on ion channels and calcium signaling involved in exocytosis, and have already been applied in studies of the exocytotic machinery (12, 38–41). Our own recent work revealed that the majority of vesicles do not access the necessary compliment of SNARE molecular machinery at the membrane required for fusion (12, 40, 42). Furthermore, vesicle dynamics are also segregated, not only spatially at the membrane but also by vesicle age as we showed that vesicles are prioritized for release according to the time since their assembly (43). This mini-review summarizes the main super-resolution imaging modalities and illustrates their potential uses in quantifying ion channel molecular biology in relation to exocytosis. **Table 2** summarizes the super-resolution imaging modalities described and their (potential) uses in examining ion channel biology.

#### **TOTAL INTERNAL REFLECTION FLUORESCENCE MICROSCOPY**

Total internal reflection fluorescence microscopy (TIRFM) is an optical technique that allows the evanescent illumination of a thin optical section (∼100 nm) near the base of a cell (44). For TIRFM the use of the high NA objective becomes relevant in combination with the low refractive index of aqueous cell media. When the illumination laser encounters the interface between these refractive indices at a shallower angle than the so-called critical angle, then total internal reflection occurs, creating an evanescent wave at the interface. This extends and illuminates with an exponential decay only about 100 nm into the specimen before it becomes too



nd = not done.



weak to excite fluorophores. This creates a high-contrast image of the plasma membrane and adjacent structures with little background fluorescence from the intracellular compartments of the cell. This allows the imaging of ion channels, vesicles, and exocytotic proteins in living samples with high temporal resolution (13, 42, 45, 46). TIRFM is the foundation for a number of superresolution approaches that take advantage of the thin optical section generated in order to reduce out-of-focus signal.

#### **SINGLE MOLECULE LOCALIZATION MICROSCOPIES**

An impressive pallet of fluorescent proteins (FP) has been created, each with different excitation and emission wavelengths. Not only has the color range been refined, but also the use: through modification of the molecular structure, photo-activatable (PA) FPs have been developed, permitting applications in advanced microscopy such as photoactivation localization microscopy (PALM) (47, 48). Techniques such as PALM, ground-state depletion with immediate molecular return (GSDIM) (49), and stochastic optical switching microscopy (STORM) (50) have circumvented the diffraction-limit of resolution to permit the localization of single molecule point-spread-function (PSF) signals that are separated in time. STORM and GSDIM detect PSF-signals from single organic dye molecules; a potentially high density of molecules (e.g., fluorescent-conjugated antibody-detected epitopes)

in the sample is driven into a long-lived, optically inactive "dark state" using a combination of high illumination power and specialized, highly reducing buffers. Single fluorescent molecules then spontaneously and randomly re-emerge to emit photons during subsequent imaging (**Figure 1**). This permits the acquisition of single PSF-signals that are sparsely distributed in a time-resolved dataset. PALM employs PA encodable FP (47, 48). These molecules are inherently non-fluorescent but can be altered conformationally, using short-wavelength illumination in order to render them fluorescent. By determining empirically the dose of activation energy required in order to activate a sparse fraction of all PA-molecules, again, single PSFsignals can be acquired. Rapid imaging trains capture the fluorescent data, before single molecules are irreversibly photodestroyed by bleaching and other processes. Cycles of activation, followed by imaging and photo-destruction are repeated over and over again acquiring 1000 s of images in total. PALM is already starting to be used to define ion channel molecule localization, recently employed to image single aquaporin molecules (51).

Single molecule signals can be identified by their shape, size, expected fluorescent intensity (i.e., they are PSFs) as well as by their quantal "on–off" behavior. Mathematical fitting, localizing the centroid of these PSFs defines the precise *xy* coordinates from

into the objective. This creates an evanescent wave at the interface which effectively excites fluorophores only about 100 nm into the sample. Therefore only those fluorescent proteins or dyes within that region will be excited and

short-wavelength light before it is confirmationally able to absorb excitation and emit light. **(C)** (left) Localized and rendered PALM image of BK channels on the plasma membrane of a HEK293 cell (scale bar: 5µm) with a region of interest (right) showing the localization of molecules in detail (scale bar: 1µm).

where the signals arose; the certainty of localization is affected by brightness, noise, and pixel size. By rendering coordinates, a virtual image is created which shows the coordinate positions of all the molecules on the plasma membrane that have emitted. A variation of PALM combines single-particle tracking, so termed sptPALM (52), allowing the high-precision tracking of many 1000 s of single molecule signals with high temporal resolution.

#### **SIMULATED EMISSION DEPLETION MICROSCOPY**

The underlying limitation with microscopy is the diffractionlimited width of the PSF. Single molecule localization microscopies (SMLMs) circumvent this by determining the centroids derived from single molecule PSFs, and though these are excellent techniques, they are not imaging directly sub-diffraction structures. Simulated emission depletion microscopy (STED) presents a fundamentally different approach, directly manipulating the PSF through the use of stimulated emission to deplete fluorophores before they fluoresce (53). On a conventional scanning microscope, the sample is raster scanned with an excitation beam, which increases the energy of the fluorophores from the ground state *S*<sup>0</sup> to the exited singlet state *S*1. The fluorophores spontaneously relax, emitting photons as they return to the ground state, which is detected as fluorescence. This process typically occurs on a nanosecond timescale. In STED microscopy a second depletion laser, is used immediately after the excitation beam (**Figure 2**). The depletion laser forces the excited fluorophores back

to the ground state via stimulated emission, a process that typically occurs on a picosecond timescale. This effectively forces depleted fluorophores into a dark state before traditional spontaneous fluorescence can occur. By using a phase mask profile with a donut-shaped depletion pattern, that perfectly overlaps the initial excitation beam, the effective PSF becomes defined by the geometry of the donut inner ring and the depletion laser power. The optical resolution can be considerably improved, to around 70 nm, to give a higher resolution data compared to standard diffraction-limited image. Gated-STED (gSTED) (54) combines this with time-correlated detection and pulsed excitation to further increase resolution, reduce noise, and improve utility with living samples enormously by reducing illumination energies (54–56). STED has been used to examine Ca2<sup>+</sup> channel clustering induced by the exocytotic machinery (57) as well as probing the nano-scale structure of the synapse (38, 39, 58). The advent of less cytotoxic

formed by the overlap of the standard PSF and the depletion PSF. **(C)** With

gSTED technologies will increase the utility of this approach in cell biology.

#### **STRUCTURED ILLUMINATION MICROSCOPY**

stain – microtubules. Scale bar, 500 nm.

Structured illumination microscopy (SIM) illuminates samples with a grid pattern and a sinusoidal excitation, providing a known high-frequency structured pattern (59, 60). Mathematical approaches are then applied to reconstruct a real image from the emission data, with the ability to resolve structures at around half the diffraction-limit (i.e.,around 85 nm resolution in general). Importantly, SIM also improves the axial resolution in images and can be used anywhere in 3-D, albeit in around 300 nm *z*-volumes, in multiple spectral colors. SIM has not yet been used to examine ion channel biology, to our knowledge, but clear opportunities exist, perhaps for studying intracellular channels beyond the reach of TIRF-limited approaches.

#### **QUANTIFYING PROTEIN–PROTEIN INTERACTIONS IN CELLS**

Förster resonance energy transfer (FRET) is a physical effect where energy can be transferred in a non-radiative way between a high-energy fluorescent donor molecule and a lower energy proximal acceptor, if the absorption and emission spectra of each, respectively, overlap. As FRET is strongly dependent on the interfluorophore distance and falls off with the sixth-power of distance, it occurs only on the nanometer scale. For this reason, FRET between two fluorescent molecules is commonly interpreted as indicating a direct interaction (61). Quantifying FRET is notoriously unreliable however, especially when using fluorescence intensity as a read-out. When FRET occurs, the energy transfer from the donor to an acceptor provides a rapid route to relaxation for the excited donor molecule, resulting in a significantly shortened fluorescence lifetime. Measuring the excited-state fluorescence lifetime of the donor molecule in each pixel of an image, known as fluorescence lifetime imaging microscopy (FLIM),therefore provides a direct and quantitative approach to this problem. FLIM has been widely used to quantify FRET between the SNAREs and binding partners inside intact cells (20, 21, 42–46, 48) and could be applied similarly to the study of ion channels; for example, it has been long-thought that VGCCs containing the so-called "synprint" (31) site interact directly with syntaxin and synaptotagmin but this has never been measured directly in intact cells.

#### **FLUORESCENCE CORRELATION SPECTROSCOPY**

Although powerful, FLIM suffers from slow temporal resolution, as commonly 10,000–100,000 s of photons must be acquired from each pixel in the image in order to build a large enough sample of time-tagged photons to permit accurate data fitting (62). In addition, in FRET experiments, the absence of a detected FRET signal does not necessarily mean that no interaction has occurred,leading to potential false-negative conclusions. Fluorescence Correlation Spectroscopy (FCS) acquires intensity fluctuations caused by fluorescence emission as molecules diffuse across a tiny excitation volume held stationary in a living sample (63–65). This approach thus calculates the diffusion rates, concentrations, and molecular numbers of molecules under scrutiny. As diffusion depends directly upon mass, a simultaneous decrease in diffusion rate and cross-correlation of signals between two different fluorescent molecules in the excitation volume provides unequivocal evidence of an interaction, on the millisecond timescale.

#### **SUPER-RESOLUTION IMAGING OF ION CHANNELS**

Single molecule localization microscopy can be applied to ion channels (51), to define the precise locations of large cohorts of single molecules. Furthermore, it is possible to combines SMLMs with diffraction-limited imaging, in order to capture data describing the locations of single channels and single vesicles, for example. This approach would allow appraisals such as "nearest-neighbor" analyses, to determine the distances between vesicle centers and

#### **REFERENCES**


their adjacent ion channel molecules. These approaches were used recently, finding that in contrast to previous understanding, the majority of the secretory machinery is segregated spatially from secretory vesicles, with protein molecules moving between membrane "hotspots" distinct from vesicle docking sites (12, 40).

#### **IMAGING ION CHANNEL ACTIVITY – "OPTICAL PATCHING"**

Patch-clamp electrophysiology is the ideal technique to determine the functionality of single ion channel molecules. Combined with fluorescent Ca2<sup>+</sup> indicators, voltage-clamping can be used to deliver a train of tiny depolarizations, designed to open and close ion channels very rapidly. Image processing approaches may then be used, similarly to SMLM approaches, to localize the centroid position of the fluorescent signal reporting the ion channel location in a technique called optical patching (66–68). This approach has been used to great effect to report the locations of largeconductance, long open-dwell time IP3 receptors in cells (66); with the advent of image data "de-noising" strategies (40, 69) and faster camera detectors, optical patching promises to be useful for localizing single VGCCs in living cell membranes, combined with TIRFM.

#### **IMAGING ION CHANNEL TETRAMERS**

Active VGCCs and K channel alpha subunits are known to be tetramers; it difficult, however, to determine whether the channels detected on the cell surface using any of the approaches described above, are assembled alpha subunits. The study of ion channel function has been aided greatly over the years by the discovery that several potent natural toxins bind to specific ion channel targets with exquisite sensitivity and affinity. Recent work has taken advantage of this, using fluorescentconjugates of these toxins to localize ion channels in intact cells. For example, ω-conotoxin is a peptidyl toxin originally isolated from the fish-hunting cone snail *Conus magnus* (70, 71) and is a specific N-type calcium channel blocker. This toxin has been used in the past to label neurons and neuromuscular junctions, revealing ion channel localization in diffractionlimited images (72). These approaches can now be combined with SMLM, as recently described using a Na+-channel toxin in order to define the nano-scale locations of single assembled tetrameric alpha subunits (73). Combined with patch-clamp electrophysiology, these approaches offer the promise of manipulating channel behavior and localizing large cohorts of single molecules, simultaneously.

In summary, ion channel biology is uniquely placed in having diverse single molecule-resolution biophysical approaches available; electrophysiology and the newly emerging super-resolution imaging and spectroscopic approaches. Together, these promise to further our understanding not only of ion channel behavior, but also of channel location, function, and relationship with vesicles and the exocytotic molecular machinery.

islets. *Endocrinology* (1996) **137**:1340–8. doi:10.1210/en.137.4. 1340

3. Bommert K, Charlton MP, DeBello WM, Chin GJ, Betz H, Augustine GJ. Inhibition of neurotransmitter release by C2-domain peptides implicates synaptotagmin in exocytosis. *Nature* (1993) **363**:163–5. doi:10.1038/363163a0


1126/science.1132915


types contributing to chromaffin cell excitability, exocytosis and endocytosis. *Cell Calcium* (2012) **51**:321–30. doi:10.1016/j.ceca.2012. 01.005


Bonifacino JS, et al. Imaging intracellular fluorescent proteins at nanometer resolution. *Science* (2006) **313**:1642–5. doi:10.1126/ science.1127344


spatial resolution of STED microscopy. *Opt Express* (2011) **19**:4242–54. doi:10.1364/OE.19. 004242


*Methods* (2007) **4**:963–73. doi:10.1038/nmeth1104


**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: 30 May 2013; accepted: 16 August 2013; published online: 04 September 2013.*

*Citation: Hiersemenzel K, Brown ER and Duncan RR (2013) Imaging large cohorts of single ion channels and their activity. Front. Endocrinol. 4:114. doi: 10.3389/fendo.2013.00114*

*This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology.*

*Copyright © 2013 Hiersemenzel, Brown and Duncan. 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.*