I. INTRODUCTION

Top Previous Next References

The traffic of secretory vesicles to the plasma membrane of eukayrotic cells is essential for normal cellular function. It also forms the basis of intercellular communication in multicellular organisms through the release of a wide array of extracellularly acting molecules. The fusion of secretory vesicles with the plasma membrane occurs in essentially all cells in the form of constitutive exocytosis that is required for the insertion of new cell membrane (107, 377). Some extracellular molecules (e.g., plasma proteins, antibodies, extracellular matrix components, etc.) are also secreted by a constitutive exocytotic pathway. In many cell types, a second pathway also exists in which exocytosis can be tightly regulated to allow the controlled release of vesicle contents or regulated insertion of new membrane components due to fusion of preformed secretory vesicles only in response to a physiological signal. Regulated exocytosis has been extensively studied in synapses where it is the mechanism by which neurotransmitters are very rapidly released in a controlled manner from synaptic vesicles to mediate neurotransmission (378, 874). Certain neurons also possess dense-core granules (otherwise known as large dense-core vesicles, Refs. 53, 769) that can be triggered to undergo exocytosis independently of synaptic vesicles (52, 284, 446, 801, 871). A wide range of nonneuronal cell types contain regulated secretory vesicles identified as dense-core granules or secretory granules, the contents of which serve a diverse range of physiological functions. These include cells specialized to secrete large amounts of secretory products including, for example, neuroendocrine, endocrine, and exocrine cells but also cells where exocytosis serves a more specialized and transitory function including the sperm (134, 488, 650) and the egg (189, 724, 833). The latter examples demonstrate the essential nature of secretory granule exocytosis even for the continued existence of multicellular species.

Work from a large number of laboratories using biochemical, molecular biological, and genetic approaches has, in recent years, identified many of the proteins involved in the mechanism of synaptic vesicle exocytosis (428, 714, 734) and established their in vivo importance in neurotransmission (93, 95, 245, 434, 441, 490, 581). The interactions between these proteins and the way in which a Ca²⁺ signal leads to synaptic vesicle exocytosis are known in outline (428), although the full details of the processes involved still remain to be resolved. Similar molecular events appear to underlie secretory granule exocytosis (103, 121, 258, 260, 297, 307, 356, 408, 500, 506, 610), and in fact, functional studies of granule exocytosis in certain experimentally favorable cell types (e.g., adrenal chromaffin cells, Refs. 121, 124, 126) have been important in the development of the current description of the various stages in regulated exocytosis. Synaptic vesicle exocytosis has been extensively reviewed in recent years (35, 37, 56, 58, 60, 96, 130, 308, 316, 361, 428, 584, 638, 734, 736, 812, 847) and will not be addressed in extensive detail here. It will, however, be used to illustrate some of the advances in our molecular understanding and in comparisons with secretory granule exocytosis. Our main aim in this review is to concentrate on the diversity of dense-core or secretory granule exocytosis. We describe current views on the mechanistic aspects of secretory granule exocytosis and its triggering and regulation. In addition, we discuss its control and cell-specific specializations related to the physiological role of granule exocytosis in the diverse cell types in which it occurs.

	II. THE WIDESPREAD DISTRIBUTION OF SECRETORY GRANULE EXOCYTOSIS

Top Previous Next References

Secretory granules and their regulated exocytosis have been most extensively studied in a few cell types chosen either as model systems due to certain experimental advantages or due to their crucial physiological or pathophysiological interest. The pathway followed by secretory proteins through the cell was delineated in classical studies by George Palade from the study of pancreatic exocrine cells (578). In more recent years, probably the most studied cell types have been the adrenal chromaffin cell (and its tumor counterpart the PC12 cell line) as a model for both biochemical and electrophysiological investigation (121, 279, 506), the pancreatic -cell (225, 408, 844) due to the importance of insulin secretion, and its dysfunction in forms of diabetes mellitus and hemopoietic cells [mast cells (430), platelets (159, 415, 416), and neutrophils (88, 540, 689)] due to their importance in normal and pathological immune responses (732). Secretory granule exocytosis also occurs, however, in many different neuroendocrine and endocrine cell types for the secretion of peptides and other hormones (53) and in exocrine cells for the secretion of digestive enzymes (104, 596, 826). Specialized secretory cells are found in various predominantly nonsecretory tissues [e.g., in the heart (524), the lung (651), and the kidney (301)]. In addition, granule exocytosis also has specialized functions in other cells such as the sperm (134, 488) and the egg (40, 54, 356, 469).

The wide range of secretory granule-containing cells present in mammalian species is shown in Table 1. The list includes neurons that can possess dense-core granules in addition to synaptic vesicles or may possess dense-core granules alone (769). Certain neuroendocrine and endocrine cells are present in tissues with a purely secretory function such as those in the pituitary, adrenal medulla, or pancreas. Others are present as less abundant cells in tissues specialized for other functions such as the G cells that reside in the gastric mucosa and secrete gastrin (75, 218, 798), the juxtaglomerular cells of the kidney which secrete renin (255, 301, 545), the alveolar II epithelial cells of the lung responsible for secretion of surfactant (304, 305, 651, 841), or goblet cells in the lung and intestine that secrete mucin (718, 799). Cells with other functions such as endothelial cells also have a regulated secretory function (74, 238, 797) for release from their Weibel-Palade bodies. A more specialized form of exocytosis occurs in the egg where cortical granule exocytosis is not to release extracellularly acting signaling molecules but to release components of the fertilization envelope that prevents further sperm entry (833). In the case of the sperm, a single secretory organelle at the head of the sperm, the acrosome, releases enzymes required to enable fertilization (134, 488). Another unusual cell type is the mammary epithelial cell responsible for milk production (120, 127, 413, 433). These cells produce and secrete large amounts of the milk proteins including the caseins and do so by both constitutive and Ca²⁺-regulated exocytosis (788, 789).

	III. BIOGENESIS OF SECRETORY GRANULES

Top Previous Next References

In order for secretory cells to release material by exocytosis strictly on demand, this material cannot be allowed to enter the constitutive exocytotic pathway. Therefore, some mechanism(s) must exist to selectively exclude regulated secretory proteins from nascent constitutive secretory vesicles. It follows, then, that an alternative organelle capable of stimulus-induced exocytosis is required to accommodate these regulated secretory proteins. In general, secretory granules are used for this purpose. The highly condensed granule matrix, which gives rise to the eponymous dense core visualized by electron microscopy, allows efficient storage of large amounts of secretory material in an osmotically inert form (107). Although in most cell types secretory granules appear to represent an entirely new class of organelle, granules in various hemopoietic cells and certain other cell types share several properties with lysosomes (83, 732). For example, in addition to possessing lysosomal markers, such granules are accessible via the endocytic pathway and dense-core formation can occur in multivesicular bodies (624, 732). This suggests that a convergence of biosynthetic and endocytic membrane traffic analogous to lysosomal biogenesis is required for secretory granule formation in such cells. Indeed, Chediak-Higashi syndrome in humans (and its mouse model, beige) is characterized by giant intracellular lysosomes and granules in hemopoietic cells and melanocytes, whereas granules in other cell types are unaffected (87, 486). Similarly, the acidic hydrolases contained within the acrosome of spermatozoa suggest that this organelle may be a modified lysosome (785). Indeed, as conventional lysosomes can undergo stimulus-induced fusion with the plasma membrane in a wide range of cells (19), it may be that even the classical secretory granules evolved from a lysosomal progenitor.

In contrast to the biogenesis of "secretory lysosomes" found in hemopoietic cells, the granules found in endocrine, exocrine, and neural cells are a product of the biosynthetic pathway alone. Initial formation of immature granules occurs at the trans-Golgi network (TGN), although subsequent maturation steps occur beyond this compartment (28, 774). Unlike transport vesicle formation at earlier stages in the secretory pathway, granule formation does not apparently require a coat-driven budding process. Instead, it is thought that membrane deformation may result from the aggregation of secretory proteins in the TGN. Chromogranin A (CgA) appears to be essential for this process in neuroendocrine cells, as downregulation of CgA in PC12 cells reduces granule number and, strikingly, expression of CgA in fibroblasts induces the formation of dense-cored vesicles of around 100 nm diameter (383). Although the mechanistic basis of this has not been demonstrated, presumably the aggregative properties of CgA, coupled with its propensity to interact with membranes (351), allows wrapping of the TGN membrane around the forming CgA aggregate, analogous to the budding of enveloped viruses. Although CgA alone seems to turn on secretory granule biogenesis, its restricted expression in neural and endocrine tissues means that other as yet unidentified proteins must fulfill this function in other cell types. Interestingly, heterologous expression of pro-von Willebrand factor in endocrine and neuroendocrine cell lines has been shown to cause the biogenesis of tubular structures resembling Weibel-Palade bodies, the organelle in which endogenous von Willebrand factor is normally stored within endothelial cells (814). Similarly, heterologous expression of the melanocyte-specific protein Pmel17 in nonpigmented cells induces the formation of pre-melanosome-like structures (64). Consistent with this finding, inhibition of synthesis of melanin, the content of melanosomes, inhibits melanosome biogenesis (296). Nascent granule buds formed by CgA or its putative functional homologs must subsequently be pinched off to form immature secretory granules. Cholesterol is essential for this process as depletion of cholesterol arrests granule biogenesis at a late stage with dense-cored buds observable at the TGN (821). The role of cholesterol in vesicle scission may be direct, by facilitating negative membrane curvature at the bud neck (821). Alternatively, because cholesterol is a key component of lipid rafts, it may play an indirect role in the recruitment of raft-associated proteins that then drive membrane scission. Dynamin II is an attractive candidate for such a protein, because it has been shown to control granule formation in AtT20 cells (857), and because its homolog, dynamin I, is thought to directly drive endocytic vesicle scission (201).

Once formed, immature secretory granules must be processed and remodeled to form mature secretory granules. In endocrine and neuroendocrine cells, this involves both fusion with other immature secretory granules and removal of missorted material via budding (775). Fusion between immature secretory granules contributes to the increased size and density of mature granules in PC12 cells (773) and occurs via the ubiquitous core machinery for membrane fusion (792) (see sect. V). Perhaps surprisingly, immature granules can undergo regulated exocytosis (773), indicating that the process of granule maturation does not confer plasma membrane fusion competence, but rather inhibits homotypic granule-granule fusion. Intriguingly, immature granule fusion requires syntaxin 6 (830), and this SNARE protein is absent from mature granules, being removed by the budding of AP-1/clathrin-coated vesicles (390). This suggests that the transition from an immature granule capable of homotypic fusion to a mature granule capable only of exocytosis is due to the removal of SNAREs involved in granule-granule fusion. In a variety of cell types, it has been shown that newly synthesized granule contents are preferentially released, suggesting that immature granules may be inherently more fusogenic (see Ref. 283 and references therein). However, studies on AtT20 cells suggest that the process of maturation actually increases stimulus-secretion coupling, as newly formed granules are poorly responsive to secretagogues relative to mature granules (226). Again, this change is likely to be mediated by removal of proteins via coated vesicles. In this case, the presence of synaptotagmin IV on immature granules is thought to inhibit the putative exocytotic Ca²⁺ receptor synaptotagmin I (437), with its removal during maturation enabling normal Ca²⁺-dependent release (226). In addition to altering organelle fusion specificity, budding via coated vesicles serves to reduce excess membrane from the condensing granule and to remove missorted proteins, such as lysosomal proteins (390).

	IV. THE EXOCYTOTIC PATHWAY

Top Previous Next References

With the use of a range of functional assays, regulated exocytosis has been functionally dissected into a number of stages (121, 123, 279, 462, 585). In general, regulated exocytosis of secretory granules involves exocytosis of a premade pool of stable granules (Fig. 1) that in some cell types have a very long half-life in the absence of stimulation (194). The pool of secretory granules may be fully releasable [as in mast cells for example (430) where a full all-or-nothing exocytotic response occurs (331)] but in many cell types (such as neuroendocrine, endocrine, and exocrine cells), a large reserve pool exists and only a small portion of the granules (~1%) are initially rapidly releasable. In endocrine cells, ongoing secretion eventually requires synthesis of new granules. In many cells, the granules exist in a range of states from being nonreleasable or in a pool that becomes releasable following a recruitment step, through to being part of a so-called "ready-releasable" pool (18, 325, 519, 522, 559, 585, 654, 766, 767, 851, 852) that can undergo exocytosis within tens of milliseconds. These pools have been defined and studied through the use of permeabilized cell assays of exocytosis (109, 110, 342, 396, 430), single cell electrophysiological approaches using patch-clamp capacitance measurement of cell surface area (306, 521, 559), amperometric measurement of the release of single granule contents (418, 838), or by imaging exocytosis through the use of evanescent wave microscopy (729). These techniques have allowed more direct measurement of exocytosis than is generally possible for synapses where usually the small size of the presynaptic terminals precludes direct measurement of exocytosis. The majority of work on synaptic vesicle exocytosis has relied on overall measurement of synaptic transmission based on a simultaneous pre- and postsynaptic recording and functional assessment of the number of released quanta (95, 349, 459). Alternatively, the overall synaptic vesicle cycle within synaptic terminals has been assessed using uptake and release of the lipophilic FM dyes (20, 69) that became fluorescent following insertion into lipid bilayers and thereby, report plasma membrane area and changes due to exocytosis and membrane retrieval (69, 70, 326, 374, 388, 512, 660, 726, 846). This technique has, however, poor time resolution compared with the speed of neurotransmission (662). Patch-clamp capacitance measurement has been possible for a small number of rather specialized giant synaptic terminals (323, 586, 677, 810, 811). In addition, morphological approaches with electron microscopy have provided a static snapshot of granules (603, 604) or synaptic vesicles (293, 294, 330, 817) and their distribution within the cell. More recently, novel imaging techniques using evanescent wave (also known as total internal reflection) microscopy have been used to examine the movement of single secretory vesicles and exocytosis at the level of the light microscope in PC12 and chromaffin cells (365, 550-552, 728, 730), INS-1 cells, (782), pancreatic -cells (548, 549), and in specialized synapses (863). These new approaches have allowed an analysis of the dynamics of granule recuitement, docking at the plasma membrane, and exocytosis in living cells.

View larger version (151K): [in this window] [in a new window]   Fig. 1. Electron micrographs showing the secretory granules in adrenal chromaffin cells. A: a general view of an unstimulated chromaffin cell. The scale bar represents 1 µm. B: a granule core being released by exocytosis from a cell in the hamster adrenal medulla. The scale bar represents 100 nm.

The functionally defined, sequential stages within the exocytotic pathway (Fig. 2) have been shown to involve ATP-dependent priming steps, physical movement of vesicles to the subplasmalemmal region of the cell, tethering and then docking at release sites on the plasma membrane, conversion to a fully releasable state, triggered membrane fusion, release of granule contents, and finally retrieval of the granule membrane (Fig. 1). In the case of Ca²⁺-triggered exocytosis, Ca²⁺ may stimulate multiple stages before membrane fusion as well as fusion itself and also membrane retrieval by endocytosis (111, 235, 458, 522, 579, 703, 765). Description of some of these stages in exocytosis has resulted in some confusion in the literature due to different uses of terminology or due to the limitations of electron microscopic studies. There are significant technical problems with fixation in electron microscopy of dynamic events like exocytosis (330, 603, 604). In particular, chemical fixation acts too slowly (330, 705) to be able to preserve exocytosis in progress (unless special conditions are used, Ref. 524) and, therefore, exocytotic events are rarely seen in conventional electron microscopy even after treatments that stimulate extensive exocytosis but can be captured by use of rapid freezing techniques (399).

View larger version (11K): [in this window] [in a new window]   Fig. 2. Schematic representation of the steps leading to secretory granule exocytosis. This represents the events observed in neuroendocrine cells undergoing Ca2+-triggered exocytosis. In this case, Ca2+ entry leads to disassembly of the cortical actin cytoskeleton. This allows granule recruitment in MgATP-dependent steps to the plasma membrane where tethering and docking of the granule can occur. Fusion and release of granule contents can be triggered by Ca2+ in the absence of MgATP.

The existence of a labile ATP-dependent process required to maintain exocytosis was shown in various cell types including adrenal chromaffin cells (39) and sea urchin eggs (40). The term priming was originally used to describe an ATP-dependent process, detectable in permeabilized adrenal chromaffin cells that preceded Ca²⁺-triggered fusion and increased the extent of exocytosis due to a Ca²⁺ trigger (340). Priming was itself stimulated by lower levels of Ca²⁺ (78). Priming and fusion were subsequently shown to be stimulated by distinct cytosolic proteins (140, 315, 816). ATP-dependent priming, requiring ATP hydrolysis, encompasses multiple events. These can include reorganization of the actin cytoskeleton (113, 151, 154, 806) to recruit granules to the subplasmalemmal space and modification of the so-called SNARE proteins by -SNAP and the ATPase N-ethylmaleimide-sensitive fusion protein (NSF) (45, 50, 140, 208, 288, 321, 384, 436, 505, 514, 580, 713, 714) (described in detail in sect. VB). In addition, ATP is also required in a house-keeping role to maintain the levels of polyphosphoinositides required for continued exocytosis (227) as illustrated by the identification of phosphatidylinositol transfer protein (PITP) (314) and phosphatidylinositol 5-kinase (313) as cytosolic priming factors that can reconstitute exocytosis after ATP depletion. In the studies of secretory granule exocytosis, priming is a term used to describe any functionally detected ATP-dependent process that occurs before fusion. It is well established that membrane fusion itself does not require ATP hydrolysis (322, 340, 430, 585, 758, 851, 852). Confusion in the literature has come from the term priming also being applied to an ill-defined step in synaptic vesicle maturation that may not be ATP dependent. Within synapses a pool of synaptic vesicles are described as "docked" (129) as they appear to be very close to the presynaptic plasma membrane at the active zone by electron microscopic examination. Based on electrophysiological measurement, only a small proportion (~1-10%) of these docked vesicles is able to undergo exocytosis in response to Ca²⁺ elevation after Ca²⁺ channel opening following a single action potential (282). This pool of docked vesicles can, however, be released experimentally, through an unknown mechanism, by treatment with hypertonic sucrose (654). It has been suggested that the so-called docked vesicles undergo a maturation process after docking to become releasable. Confusingly, this has also been termed priming even though there is no experimental evidence as to whether or not it is an ATP-dependent step. This maturation step is not understood in molecular terms but is likely to involve the assembly of the fusion machinery, following tethering of the vesicles to the membranes, to allow very rapid synaptic vesicle exocytosis and appears to involve the protein unc13 (33).

The term docking is also one that has generated much confusion. As noted above, the concept of docked vesicles initially came from electron microscopic studies of fixed brain samples (293) in which synaptic vesicles close to the plasma membrane (within 10 nm) could be counted. Some of these vesicles remain associated with isolated plasma membrane preparations (153) and so must be tightly bound to the plasma membrane. These vesicles or granules seen by electron microscopy should correctly be described as "morphologically docked." Confusion arose from the assumption that all morphologically docked vesicles are attached to the plasma membrane via the same mechanism (129, 349) and are engaged with the core exocytotic machinery (particularly the SNARE proteins). We now know, from study of various vesicular traffic steps, that multiple mechanisms exist for initial attachment or tethering of vesicles to the target membrane (177, 599, 827). This precedes a true docked interaction in which the vesicles are poised for exocytosis. In general, the morphological distribution of vesicles and granules can be misleading, as this cannot easily be interpreted in relation to protein interactions known mainly from in vitro studies.

The precise events that initiate and complete membrane fusion are not understood, although many of the proteins involved have been identified (discussed in sect. V). It has been established through biophysical measurements, however, that fusion proceeds through the formation of a fusion pore that can open transiently and then potentially rapidly reclose (Fig. 3) (429). The majority of the data on fusion pore characteristics have come from analysis of secretory granule exocytosis. Fusion pores have, however, recently been observed in recordings from posterior pituitary nerve endings that result from fusion of both large dense-core vesicles and from synaptic-like microvesicles (391). Patch-clamp measurements of plasma membrane capacitance in mast cells were the first to allow the detection of changes in cell surface area due to single granule fusion events and also demonstrated a phenomenon known as capacitance flicker, which suggested that reversible fusion events were occurring (241). Later electrophysiological measurements characterized the transient fusion events (91, 493, 542, 543) and determined the current flow through the initial fusion pore (11, 90, 429, 442, 653, 667), which was of a defined magnitude as if the pore operated like a large ion channel or gap junction channel. It appeared that the initial fusion pore could either reclose or alternatively lead to a phase of fusion pore expansion (429) and thereby full exocytosis. Reclosure of the fusion pore can occur even after an expansion phase (12). The rate of fusion pore expansion was found to be a potentially regulated phenomenon and was increased by elevating intracellular Ca²⁺ concentration in rat mast cells (242, 310) or by activating protein kinase C in eosinophils (667). One physiological consequence of fusion through a reversible fusion pore or reclosure after fusion pore expansion is that this could allow control of the amount of release that occurs per secretory granule. It has been speculated that the rate of fusion pore expansion for synaptic vesicles could determine the rate of increase of glutamate concentration in the synaptic cleft and, thereby, determine whether or not synaptic transmission occurs (172). The concentration of glutamate reached is crucial for whether a synapse is active or "silent," and indirect data show that this concentration can be modulated. It seems unlikely that controlling fusion pore expansion would be a mechanism to regulate release of small molecules such as neurotransmitters, however, because these would diffuse too quickly for release to be limited. It could, in contrast, be significant for the release of proteins (579) or larger structures within secretory granules (305).

View larger version (17K): [in this window] [in a new window]   Fig. 3. Comparison of kiss-and-run exocytosis and full fusion. Membrane fusion initially occurs by the formation of a fusion pore. Fusion pore expansion allows release of granule content. This can be limited by rapid reclosure of the fusion pore (kiss-and-run). Alternatively, full emptying of the granule can be followed by a retrieval mechanism involving dynamin and clathrin.

The concept of the fusion pore has been validated using carbon-fiber amperometry, which detected low levels of release of contents from mast cell granules during the initial phase after fusion (during capacitance flickering) and before the postulated fusion pore expansion had occurred (16). Biophysically similar fusion pores have been detected during fusion mediated by the influenza virus hemagglutinin (HA) fusion protein (200, 429, 872). In addition, analysis of fusion mediated by HA mutant constructs has shown that fusion pore formation and subsequent fusion pore expansion are mechanistically distinct events (457). Little is known about the structure of the fusion pore, and in extreme models (15, 143, 494, 495, 515, 872), it has been postulated to be either entirely protein based (as in an ion channel) or purely lipidic but surrounded by a protein scaffold. Two studies of mutant mice have suggested that the fusion pore is a real structure that may eventually be understood through molecular genetic studies. In the first, transient fusion events were found to be more frequent in membrane capacitance recording of mast cells from the ruby-eye mutant mouse (542). It is not yet known, however, which gene contains this mutation. In the second, mice were generated to knockout expression of SCAMP1. This had no major effect on brain development, but in mast cells, again transient, reversible, membrane fusion events were more frequent (243). These studies suggest that modification of protein expression can affect the opening/closure characteristics of the fusion pore, although the molecular basis for the changes in the mast cells from these mice is yet to be determined.

	V. THE CONSERVED CORE MACHINERY FOR REGULATED EXOCYTOSIS

Top Previous Next References

By the 1970s, the idea that stimulus-release coupling occurred via regulated exocytosis of secretory or synaptic vesicles was generally accepted. Although it was clear that Ca²⁺ was a key trigger of this process (219, 373), the downstream mechanism responsible for membrane fusion was entirely unknown. Indeed, subsequent studies of granule exocytosis in various cell types over many years failed to significantly illuminate this issue. Ironically, arguably the greatest advances in our understanding of the mechanism of regulated exocytosis came from complementary genetic and biochemical studies of constitutive secretion (613, 657). In the genetic approach, yeast cells were mutagenized and then screened for temperature-sensitive mutants able to secrete at 25°C but not at 37°C. This resulted in the isolation of the first secretory (sec) mutant sec1 (539). Subsequently, further mutants were isolated and subdivided into 23 complementation groups, termed sec1-23, each representing at least one mutant allele of a gene required for vesicle formation or consumption at defined steps in the secretory pathway (369, 537, 538, 613). In the biochemical approach, membrane traffic through the mammalian Golgi was reconstituted in vitro using isolated Golgi complexes and cytosol (41). Subsequent protein purification identified several soluble proteins required for efficient intra-Golgi transport (657).

The first protein identified using the biochemical approach was NSF (Table 2, Fig. 4), which was purified based on its ability to reconstitute intra-Golgi transport after blockade by N-ethylmaleimide (NEM) (82). NSF was also implicated in endoplasmic reticulum (ER) to Golgi transport (55) and endocytic vesicle fusion (216, 647) in vitro, suggesting that NSF was required for multiple steps in the constitutive secretory and endocytic pathways (657). When the mammalian NSF gene was cloned (839), its coding sequence was found to be 48% identical to that of the yeast SEC18 gene (224). Like NSF, Sec18 is required for multiple steps in the secretory and endocytic pathways (292, 636), and Sec18 can substitute for NSF in mammalian in vitro assays of endosome fusion (845), intra-Golgi transport (839), and granule exocytosis (722). These observations suggested that NSF and Sec18p contribute similar, if not identical, biochemical functions to an evolutionarily ancient molecular mechanism of membrane fusion operative in all cells.

View larger version (31K): [in this window] [in a new window]   Fig. 4. Major proteins involved in regulated secretory granule exocytosis. The role of many of these proteins was originally identified from studies of neurotransmitter release or exocytosis in neuroendocrine cells such as chromaffin or PC12 cells. In some secretory cells, nonneuronal isoforms of the proteins are likely to be involved. The functional importance of all of the proteins shown has been established by genetic manipulation of their expression.

This notion received a major boost when biochemical protein-protein interaction studies revealed that NSF and its essential cofactor, -soluble NSF attachment protein (-SNAP; Sec17 in yeast), interact with brain membranes via a SNAP receptor (SNARE) complex comprising VAMP/synaptobrevin (Snc1/2 in yeast), syntaxin1 (Sso1/2 in yeast), and SNAP-25 (Sec9 in yeast) (714). These synaptic SNARE proteins are proteolytic substrates for the clostridial neurotoxins (Table 3; reviewed in Ref. 497), which are potent inhibitors of neurotransmission, thus suggesting that regulated exocytosis at the synapse occurred by the same basic process as constitutive membrane fusion. The realization that SNARE homologs were associated with distinct intracellular compartments and that yeast SNARE mutants blocked membrane fusion with these organelles led to the idea that SNARE proteins are essential for all intracellular membrane fusion events. Furthermore, the realization that two additional protein families, namely, the Rabs and the Sec1 homologs, were also involved in multiple vesicular transport processes led to the concept of a universal mechanism of membrane fusion (61, 246, 657). This idea is now generally accepted, although debate still rages over the proposed functions of the individual components of the conserved core machinery and over the precise mechanism of membrane fusion. Here we discuss current ideas of the biochemical functions of SNAREs and their regulatory interacting proteins and review the evidence for the involvement of these proteins in granule exocytosis.

SNAREs are membrane proteins that are localized to the various intracellular organelles. Although originally defined biochemically as SNAP receptors, most SNAREs have actually been identified by possession of a characteristic heptad repeat sequnce known as the SNARE motif. Searching for similar motifs in the human genome reveals 35 SNAREs, compared with 22 in the yeast genome (84, 588). Despite the large number of SNARE proteins, the synaptic SNARE complex first identified by Sollner et al. (714) remains the most intensively studied, and most current ideas of SNARE function are based on this, the archetypal complex. It comprises three proteins: the integral membrane proteins VAMP/synaptobrevin (777) and syntaxin 1 (59) and the membrane associated protein SNAP-25 (573). VAMP is a synaptic vesicle protein, whereas syntaxin and SNAP-25 are most enriched in the presynaptic plasma membrane (Fig. 4). These localizations led to the classification of SNAREs as vesicle (v-) or target (t-)SNAREs, and the notion (known as the SNARE hypothesis) that the fidelity of vesicle docking and fusion is determined by the specificity of v-SNARE:t-SNARE interactions (714). More recently, a rival classification system has been proposed based on whether a conserved glutamine (Q) or arginine (R) is present in the SNARE motif (240) in the so-called zero layer (Fig. 5), although in practice Q-SNAREs are almost always t-SNAREs and R-SNAREs are almost always v-SNAREs. In the synaptic complex, syntaxin (t-SNARE) contributes one Q-containing helix, SNAP-25 (t-SNARE) contributes two Q-containing helices, and VAMP (v-SNARE) contributes one R-containing helix (605, 743) (Fig. 5). Characterized SNARE complexes always contain three Q-helices (collectively, the t-SNARE) and one R helix (the v-SNARE), although these may be contributed from three or four proteins (SNAP-25 homologs typically contribute 2 helices) (240). This code of 3Q:1R helices has been shown to be the essential combination which, in a lock and key fashion, allows specific SNARE pairing and consequent membrane fusion between appropriate membranes in a reconstituted liposome fusion assay (483, 583). The Q and R residues themselves, however, are not essential for membrane fusion (291); rather, mutation of these amino acids impairs later disassembly of the SNARE complex by NSF and SNAPs (see sect. VB).

View larger version (54K): [in this window] [in a new window]   Fig. 5. Structure of the neuronal SNARE complex. The structure of the core complex formed by the cytoplasmic helices of VAMP, syntaxin 1, and SNAP-25 was solved by X-ray crystallography. A: the overall structure of the 4-helix bundle showing the position of the zero layer. B: a cross-sectional view of the interactions in the zero layer involving Arg-56 of VAMP, Gln-226 of syntaxin with Gln-53 and Gln-174 of SNAP-25.

The idea that SNARE complexes are essential for membrane fusion in living cells is well supported by genetic studies in the yeast, Saccharomyces cerevisiae, where deletion of SNAREs is usually lethal, and where conditional SNARE mutants are blocked at predictable membrane traffic steps (277). Similarly, genetic ablation of synaptic SNAREs in Drosophila melanogaster, Caenorhabditis elegans, and Mus musculus abolishes evoked neurotransmission (533, 678, 681, 825). It is interesting to note, however, that the effect of inactivating syntaxins tends to be more dramatic than when other SNAREs are blocked. For example, ablation of VAMP-2 or SNAP-25 in mice and n-syb in Drosophila abolishes evoked neurotransmission, but spontaneous transmitter release events are relatively unaffected (94, 209, 678, 825). In contrast, syntaxin null mutants in Drosophila exhibit a complete block in both spontaneous and evoked neurotransmission (681). Likewise, deletion of both syntaxin1 homologs (Sso1 and Sso2) in yeast prevents exocytosis and so is lethal, whereas strains deleted in both VAMP homologs (Snc1 and Snc2) are viable. Such v-SNARE-independent membrane fusion can result from complexes formed between Sec9 (yeast SNAP-25) and dephosphorylated Sso1/2 (yeast syntaxin) (452), supporting the idea that SNAREs (or at least syntaxins) are essential for intracellular membrane fusion.

In addition to studies in genetically tractable organisms, the use of clostridial neurotoxins has provided important evidence for an essential role of SNAREs in exocytosis. The key discovery in this regard was that tetanus and botulinum B neurotoxin were metalloendoproteases that blocked neurotransmission presynaptically by selectively cleaving VAMP (432, 669). Following this work, it was established that the remaining six botulinum neurotoxins cleaved either VAMP, syntaxin 1, or SNAP-25 (Table 3 and references therein). Clostridial neurotoxin light chains (the proteolytic subunits) were already known to block granule exocytosis in permeabilized adrenal chromaffin cells and PC12 cells (6, 76, 77, 443, 481), indicating that SNAREs were also required for exocytosis in such neuroendocrine cells. To test if this was true for all granule exocytosis, many labs applied clostridial neurotoxins to permeabilized secretory cells. This was shown to block secretion from pancreatic -cells (630, 663), pancreatic acinar cells (269), parotid acinar cells (261), AtT-20 cells (4), eosinophils (336), platelets (249), peptidergic hypothalamic neurons (206, 367), and gastric enterochromaffin-like cells (338) and to block the acrosome reaction in human sperm (772). Although not all granule exocytosis is blocked by clostridial neurotoxins, this should not be interpreted as providing evidence for SNARE-independent membrane fusion. Rather, the most likely explanation is that exocytosis in such cells proceeds via toxin-resistant SNARE isoforms (it is known that even minor changes to recognition and/or cleavage sites can prevent proteolysis by clostridial neurotoxins, Ref. 794). For example, the lack of sensitivity of mast cell secretion to botulinum neurotoxin E reflects the involvement of the nonneuronal SNAP-25 homolog SNAP-23 (297, 793).

As an alternative to the use of clostridial neurotoxins, various reports have shown that SNARE-directed antisera or peptides inhibit granule exocytosis. When applied to well-studied model systems such as chromaffin cells, this serves as independent confirmation of neurotoxin data (299, 300). However, such approaches can also reveal new information on SNARE function in exocytosis. In platelets, for example, exocytosis of -granules, dense-core granules, and lysosomes is inhibited by antisera to syntaxin 2 and SNAP-23, but dense-core granule exocytosis is unique in being unaffected by syntaxin 4 antibody (159, 160, 416). Similarly, in neutrophils, while syntaxin 6 antibody prevents exocytosis of both specific and azurophilic granules, SNAP-23 antibody inhibits only specific granule exocytosis (465). This suggests that independent exocytosis of distinct organelles within the same cell may be controlled, at least in part, by distinct SNARE combinations, in keeping with observations on reconstituted liposome fusion (483, 583).

Clearly, SNAREs are critical for granule exocytosis, but what is their role in this process? Several lines of evidence suggest that SNAREs themselves may directly drive membrane fusion. First, tetanus toxin application results in the accumulation of docked vesicles at the synapse, suggesting that SNAREs function after morphological vesicle docking/tethering (see sect. IV). Second, formation of the SNARE complex is very energetically favorable; assembled SNARE complexes are resistant to SDS and have extremely high thermal stability (239, 317). Third, the structure of the synaptic SNARE complex is such that formation of its coiled-coil bundle (Fig. 5) would bring the transmembrane anchors of VAMP and syntaxin (and hence the vesicle and plasma membrane) into close apposition (309, 337, 605, 744). Fourth, and most compelling, membrane fusion between artificial liposomes has been reconstituted using recombinant SNAREs alone (483-485, 829). Taken together, these observations support a model (Fig. 6) where a conformational change in syntaxin exposes its H3 helix, which then serves as a nucleation site for interaction with the VAMP coiled-coil domain and the two helices of SNAP-25. Formation of the full complex then proceeds down its energy gradient, unleashing sufficient energy to overcome the hydration barrier and initiate bilayer fusion.

View larger version (28K): [in this window] [in a new window]   Fig. 6. Schematic scheme for membrane fusion driven by the SNAREs. A: syntaxin is initially unavailable for assembly into the SNARE complex due to its association with munc-18. B: dissociation of munc18 frees syntaxin which adopts an "open" conformation. C: the SNAREs interact and assemble the four-helix bundle. D: as the SNARE complex is fully assembled, the lipid bilayers of the plasma membrane and vesicle are brought close enough for bilayer fusion to occur.

Studies of granule exocytosis have been important in testing if this model applies to physiological membrane fusion. One ingenious approach (629) has been to engineer mutant SNARE proteins that are resistant to clostridial neurotoxin cleavage but are otherwise fully functional. Transfection of such constructs into neurotoxin-treated chromaffin cells and -cell lines restores exocytosis (287, 571, 629, 823), enabling the functional analysis of other mutations to be examined. This has revealed an essential role for the coiled coil domain of VAMP (629), supporting the view that VAMP's function in exocytosis requires incorporation into the SNARE complex. Analogous studies with SNAP-25 have revealed a novel role for the palmitoylated linker domain between its two Q helices in SNARE complex disassembly and exocytosis (824). Furthermore, this system has been used to demonstrate that the conserved glutamine residues that define SNAP-25's two Q helices (Fig. 5) are dispensible for granule exocytosis, suggesting that formation of a fully stable SNARE complex may not be essential for membrane fusion (291). A modification of this approach using soluble fragments of SNAP-25 has further shown that reconstitution of exocytosis in botulinum E-treated PC12 cells requires only the COOH-terminal helix and that this selectively stimulates the late Ca²⁺-induced triggering stage (164, 165), again supporting an active role for SNARE complex formation in driving membrane fusion (166). Furthermore, studies using SNARE-specific antibodies suggest that SNAREs exist in loosely complexed states, with full winding of the complex occurring at a late stage in granule exocytosis (167, 853), in keeping with the model proposed above. However, some data argue against the idea that SNARE complex formation causes membrane fusion. For example, in vitro exocytosis of sea urchin cortical granules can proceed in the absence of detectable SNARE assembly (189, 748). In addition, work on the yeast vacuolar fusion system has suggested that SNARE complex formation acts upstream of membrane fusion (791), with fusion ultimately being catalyzed by vacuolar ATPase subunits (594). Further work is needed to finally define the function of SNAREs in membrane fusion, and studies of granule exocytosis will no doubt be central to such future endeavours.

NSF and -SNAP are soluble proteins that act to disassemble SNARE complexes. Owing to the extreme stability of the SNARE complex (see sect. VA), disassembly requires significant energy input in the form of ATP hydrolysis (714), and this is provided by the homohexameric ATPase NSF. Each identical subunit of NSF comprises an NH₂-terminal domain (N) required for interactions with -SNAP, followed by two AAA⁺ ATPase domains (D1 and D2) with distinct, essential functions (836). The ability to hydrolyze ATP is unique to the D1 domain in both NSF and its yeast ortholog, Sec18 (507, 722, 740, 747, 835, 854), whereas the function of the catalytically inactive D2 domain is to establish the homohexameric conformation (417). These domains are arranged such that the SNARE complex is positioned inside the NSF hexamer, with three -SNAP molecules connecting the SNARE complex to three projecting N domains (309, 337, 836). -SNAP not only recruits NSF to the SNARE complex but also stimulates the ATPase activity of the D1 domain (318, 475, 507, 723), and this SNAP stimulation of ATPase activity is essential for SNARE complex disassembly (50). It is thought that the conformational changes in NSF induced by ATP hydrolysis are transduced through -SNAP to unwind the four helices of the SNARE complex. This process may be mechanistically similar to the unwinding of nucleic acid duplexes by DNA/RNA helicases, to which NSF shares some structural similarity (836). Although it is not yet clear where unraveling of the four helices of the SNARE complex is initiated, the conserved Q residue contributed by syntaxin is essential for SNARE disassembly (666), suggesting that NSF/-SNAP-induced hydration of the zero layer enables unwinding of the helices.

In vivo evidence for a requirement of NSF and SNAPs in membrane fusion comes from the comatose (NSF) mutants in Drosophila melanogaster (580, 700) and the sec18 (NSF) and sec17 (-SNAP) mutants in Saccharomyces cerevisiae (224, 292, 295, 538, 839). These mutants are characterized by temperature-sensitive blocks in presynaptic neurotransmission and membrane traffic through the biosynthetic pathway, respectively. In all cases, the mutant organisms accumulate SNARE complexes at the restrictive temperature (436, 712, 771), confirming that NSF and -SNAP are required to disassemble SNAREs in living cells. Recently, NSF has also been shown to be required for the assembly of exocytotic sites for trichocyst exocytosis in Paramecium (256). Despite ready acceptance of this notion, questions remained over the precise point in the membrane fusion process at which NSF and -SNAP functioned (504). Studies of granule exocytosis have been important in this regard (see below), but rigorous tests have also been performed using yeast and flies. Analysis of in vitro yeast vacuole-vacuole fusion indicated that Sec17 and Sec18 acted in an ATP hydrolysis-dependent stage that preceded membrane contact (476). Although these findings continue to be cited as evidence for a postfusion role of NSF/-SNAP (168), it is nevertheless clear that they strongly support a predocking, priming role. Analysis of comatose mutants has proven controversial, with different groups concluding both pre- and postfusion actions of NSF (376, 435, 436, 771). However, it is important to note that continued neurotransmission requires cycles of exo- and endocytosis, so it can be difficult to define in such recycling systems exactly when postfusion becomes predocking.

No such problems affect analysis of granule exocytosis, where vesicles are preformed and display very long half-lives (194), and where assaying release of endogenous granule contents precludes artefactual effects of vesicle formation/recycling. Functional evidence supports a role for -SNAP and/or NSF in granule exocytosis in chromaffin cells (505), PC12 cells (45), pancreatic -cells (384, 514), spermatozoa (488), and platelets (159, 416). The use of both stage-specific exocytosis assays in permeabilized cells and high-resolution electrophysiological approaches have established that -SNAP/NSF acts in an early ATP-dependent priming reaction that precedes the later Ca²⁺-dependent fusion step (45, 140, 851) (see sect. IV). Furthermore, this priming reaction is likely to be the disassembly of SNARE complexes, since mutant -SNAPs unable to stimulate NSF ATPase activity and consequent SNARE disassembly cannot support exocytosis (50, 288, 513, 851). Because trans-SNARE complexes formed between opposing membranes would be inaccessible to the large NSF hexamer, and because NSF function in vacuole fusion is complete even before membrane tethering, it seems certain that the substrates in the priming reaction are cis-SNARE complexes formed within the same membrane. Indeed, although predominantly plasma membrane proteins, significant levels of syntaxin and SNAP-25 are found on both synaptic vesicles (815) and chromaffin granules (339, 745, 746), and SNARE complexes formed on synaptic vesicles are disassembled by -SNAP/NSF (572). -SNAP and NSF are themselves present on chromaffin granule membranes (44, 128). Taken together, these various observations suggest that -SNAP/NSF act as molecular chaperones, continually disassembling cis-SNARE complexes to release SNARE proteins to engage in trans and hence function in further membrane fusion events (121).

Despite the widespread acceptance of this model, SNARE complex disassembly is not the only function of NSF and SNAPs. In recent years, NSF has been shown to bind to several different receptors or their interacting proteins and to regulate their insertion/removal from the plasma membrane. These include the AMPA receptor (531, 565, 715), the ₂-adrenergic receptor (187), -arrestin (478), and the GABA_A receptor-associated protein (387). This suggests a second function for NSF in membrane traffic that is independent of SNAREs. However, another novel NSF binding protein, GATE-16, interacts with a Golgi v-SNARE, stimulates NSF ATPase activity, and stimulates intra-Golgi transport (664). These properties are similar to the yeast LMA1 protein complex, which binds a vacuolar t-SNARE and stimulates Sec18 ATPase activity (854). This suggests that GATE-16 may be a general factor regulating membrane fusion, but no data are available on whether GATE-16 acts in granule exocytosis. In addition to the ubiquitous -SNAP, two other isoforms have been described (181, 182). -SNAP is brain specific (529, 834) and, despite sharing 83% amino acid identity with -SNAP, displays significant differences in some assays (181, 186), but not others (737). In chromaffin cells, both - and -SNAPs enhance exocytosis induced by micromolar Ca²⁺ (505, 737, 850), but only -SNAP does so at submicromolar Ca²⁺ levels (850). The reason for this is not clear, but it may conceivably involve the putative Ca²⁺ sensor synaptotagmin, which has been claimed to specifically bind -SNAP and thereby recruit NSF (670). The function of -SNAP, which displays only 18% amino acid identity with -SNAP, remains unclear. Although it is effective at binding and stimulating NSF ATPase activity (181, 507, 737), it is inactive in ER-Golgi transport (593) and has no reproducible effect on granule exocytosis in chromaffin cells (505).

Although much attention has focused on NSF, SNAPs, and SNAREs, these are not the only evolutionarily conserved proteins required for intracellular membrane traffic. Genetic studies have established a general requirement for Sec1-related genes in vesicle fusion in a wide variety of organisms, including yeast (SEC1, SLY1, VPS33, VPS45), plants (KEULE), nematodes (UNC18), flies (ROP), and mice (munc18-1) (32, 303, 800). The founding member of this family is yeast Sec1, the original secretory mutant isolated by the classic screen of Novick and Schekman, which is essential for exocytosis in yeast (3, 229, 537-539). The realization that similar proteins were required for exocytosis across phyla led to the identification of Sec1 homologs in mammals (variously named nSec1, munc18, mSec1, or rbSec1) (303). Three putative mammalian Sec1 orthologs have been described: the neuronal/endocrine munc18-1/nSec1 protein and the ubiquitous munc18-2 and munc18-3 isoforms. Although most work on the Sec1 family has focused on exocytosis, their general role in membrane fusion is illustrated by the requirement for Sec1 homologs in ER-Golgi transport (Sly1) and transport to the vacuole (Vps33/Slp1, Vps45) in yeast (1, 196, 564, 813). Sec1 family members from a variety of organisms have been shown to interact with syntaxin homologs, suggesting that the conserved function of Sec1 proteins in membrane fusion may be mediated via syntaxins (360). Indeed, the yeast syntaxins, Sso1 and Sso2, act as multicopy suppressors of sec1-1 mutants (2), and the inhibition of neurotransmitter release in Drosophila caused by overexpression of ROP is relieved by co-overexpression of syntaxin (848), thus supporting this notion. Neuronal Sec1/munc18 binds to syntaxin1a with nanomolar affinity in vitro to form a complex that precludes syntaxin binding to synaptic SNAREs (598). These observations have led to the suggestion that Sec1 proteins act as syntaxin chaperones, preventing SNARE complex assembly until signaled to release syntaxin (360). Although the precise nature of this signal is unknown, various observations suggest that it is linked to the completion of Rab-dependent tethering (see below).

Sec1 proteins have been implicated in granule exocytosis in pancreatic -cell lines (867), platelets (625), pancreatic acinar cells (268), PC12 cells (221), and chromaffin cells (248, 320, 335, 809). In -cell lines, nSec1/munc18 interacts with syntaxin 1, and nSec1/munc18 peptides or antisera increase insulin release (867). Although this supports the above-proposed role of Sec1 proteins as negative regulators of syntaxin function, other studies suggest that this idea is too simplistic. Indeed, loss of Sec1 family members generally results in a complete block in membrane traffic, indicating an additional positive essential role. Studies of nSec1/munc18 in PC12 cells and Vps45 in yeast suggest that, rather than merely preventing syntaxin entry into the core complex, the syntaxin-Sec1 interaction is an essential intermediate that enables activation of syntaxin for efficient incorporation into the SNARE complex (105, 221). This notion is consistent with recent studies in chromaffin cells on the effects of a mutant form of nSec1/munc18 (R39C; Fig. 7) with a reduced affinity for syntaxin. Here, it was found that the mutant resulted in an increased rate of fusion pore expansion (248), which could result from the increased release of activated syntaxin into the fusion machinery. Alternatively, these data may indicate a much later role for nSec1/munc18 in the membrane fusion process than previously appreciated, as suggested by recent studies in yeast. Unlike neuronal Sec1, yeast Sec1 will not bind efficiently to its cognate syntaxin (Sso1/2) in isolation, but only when it is assembled into an exocytotic SNARE complex (132). Clearly, this observation is difficult to reconcile with the proposed function of Sec1 proteins as syntaxin chaperones that regulate subsequent SNARE complex formation; rather, this finding suggests that Sec1 functions at the heart of the membrane fusion process, acting at or after SNARE assembly. In contrast to both of these models, studies of chromaffin cells from munc18-1 knock-out mice have revealed a major defect in granule docking (809), suggesting a much earlier role for this protein. Perhaps nSec1/munc18 has multiple functions, with both predocking and perifusion roles. However, the normal complement of docked vesicles at the synapse in munc18-1 knock-out mice, despite a complete block in membrane fusion (800), argues that there must be additional synaptic docking/tethering factors for this to be true. Clearly, Sec1 proteins have essential and possibly multiple functions in physiological membrane fusion, but much more work is required to understand these in molecular terms.

View larger version (111K): [in this window] [in a new window]   Fig. 7. Structure of the syntaxin/munc18 complex. The structure was solved by use of X-ray crystallography (491). Syntaxin forms a tight bundle seen in green at the bottom which is clasped by Munc18. Residue Arg-39, the mutation of which modifies fusion pore kinetics (248), is shown in space-fill mode in the Munc18 structure.

Rab proteins comprise the largest family within the Ras superfamily of small GTPases. Individual Rab proteins are localized to distinct, characteristic organelles, suggesting that each regulates a particular membrane traffic step. Rabs are thought to act as molecular switches, with the time spent in the active GTP-bound "on" state determining how long vesicles remain competent for docking/fusion, before GTP hydrolysis switches this state off (661). The initial findings implicating these GTPases in membrane traffic came from the identification of Sec4 and Ypt1, mutations in which bock exocytosis and ER-Golgi transport, respectively (665, 687). Interestingly, whereas the number of different SNAREs, SNAPs, NSF, and Sec1 family members is roughly similar from yeast to human, yeast contains only 11 Rab proteins, compared with 60 in humans (84). As the increased complexity during evolution does not apparently require significant amplification of the general fusion machinery, presumably the greatly increased number of Rab proteins reflects a different, regulatory role(s). The general consensus is that rabs are major determinants of the compartmental specificity of membrane transport and achieve this by selective interactions with distinct effector proteins (686, 864). The best-characterized Rab effectors are involved in the initial tethering of vesicles to target membranes. These include the exocyst complex (Sec4 effector in exocytosis) (760), the HOPS protein complex (Ypt7 effector in transport to the vacuole) (685), p115 (Rab1 effector in Golgi transport) (13), and EEA1 (Rab5 effector in endocytosis) (177). Another class of Rab effectors are motor proteins that act to control organelle movement within cells. For example, Rab5 controls movement of endosomes on microtubules and Rab6 binds to the kinesin-like protein rabkinesin-6 (228, 527). In addition to microtubule-based motors, Rabs have also been implicated in actin-based motility by virtue of functional interactions with class V myosins (207). Finally, genetic (and in some cases, physical) interactions between various Sec1 homologs and Rabs (303, 686) suggest that the Sec1 family may represent a third distinct class of Rab effectors. However, a variety of other proteins that do not clearly segregate into these classes have also been proposed as Rab effectors (864), emphasizing the potential complexity of Rab function(s) in the cell.

Rab3 exists as four isoforms A-D in mammals and is the major Rab implicated in regulated exocytosis in neurons and neuroendocrine cells. Rab3A is not an essential protein in mammals but may regulate neurosecretion by limiting the recruitment of synaptic vesicles into the releasable pool at the plasma membrane for subsequent fusion events (272, 274). Rab3 mutants in C. elegans are also viable but have less synaptic vesicles close to the active zone (534), suggesting a role for Rab3 in vesicle tethering in common with other Rab proteins (177, 597).

Overexpression of active forms of Rab3A, or other Rab3 isoforms, has been found to inhibit exocytosis in chromaffin cells (341), PC12 cells (178, 363, 364), rat basophilic leukemia (rbl) cells (640, 704, 790), insulin-secreting cells (628), pancreatic exocrine cells (162), and the cholecystokinin (CCK)-secreting STC-1 enteroendocrine cell line (278). These forms of Rab3A are also inhibitory when injected into neurons (220) consistent with an inhibitory function of Rab3A in regulated exocytosis. Several Rab3A effector proteins are known including RIM (820) and rabphilin 3A (696). Both RIM (353, 742, 820) and rabphilin 3A (25, 180, 469) are involved in secretory granule exocytosis in chromaffin, PC12, insulin-secreting cells, and mouse eggs. Analysis of specific mutations that disrupt interaction of Rab3 with these effectors has excluded them, however, as the inhibitory effectors of Rab3 (178, 193, 362, 676). The inhibitory effects of Rab3 may instead be exerted via the effector Noc2 (319), a protein expressed predominantly, if not exclusively, in neuroendocrine cells (403). Mutations were made in Noc2 based on its sequence homology with the NH₂ terminus of rabphilin 3A and the structure of the complex of Rab3 with rabphilin 3A (contact regions A and B, Fig. 8) (566). These mutations disrupted both binding to Rab3A and the ability of Noc2 to inhibit exocytosis in PC12 cells (319).

View larger version (70K): [in this window] [in a new window]   Fig. 8. Structure of the complex between Rab3A and the NH2-terminal domain of rabphilin 3A. The structure was solved by use of X-ray crystallography (566), and this figure shows Rab3A in its GTP-bound forms (ribbon structure with GTP in space fill) bound to the rabphilin 3A NH2-terminal domain. The two regions of contact between the two proteins are indicated. Contact A involves a helix from rabphilin 3A formed from residues close to the NH2 terminus of rabphilin 3A. Contact B involves the conserved SGAWFF motif of rabphilin 3A.

Other studies have suggested that certain Rab3 isoforms may, in contrast, exert positive effects on exocytosis in some cell types. Evidence for a positive role of Rab3 isoforms in granule exocytosis includes an inhibitory effect of treatment with Rab-GDI, which depletes rabs, on pepsinogen secretion from permeabilized gastric chief cells (619). In addition, depletion of Rab3B, using antisense oligonucleotides in anterior pituitary cells inhibited exocytosis (439) and overexpression of rab3D in pancreatic acini of transgenic mice enhanced exocytosis in response to secretagogues (555). Furthermore, expression of a GTP-binding deficient mutant of Rab3D (rab3A NI35I) in AtT-20 cells reduced the number of docked secretory granules and impaired exocytosis consistent with a positive role for rab3D in promoting granule docking (42). Confusingly, expression of this mutant also reduced the number of docked granules in PC12 cells (461), and conversely, expression of the wild-type Rab3A produced an increase in docked granules. These data on PC12 cells are difficult to reconcile with other data showing that overexpression of wild-type Rab3A in PC12 cells inhibited exocytosis (178, 828). It is possible, however, that Rab3A has multiple roles in exocytosis. Despite the suggestions that different Rab3 isoforms have distinct functions, a direct test of this possibility shows that the four Rab3A-D isoforms all had the same inhibitory effect after overexpression in PC12, chromaffin (178) and insulin-secreting cell lines (354).

Evidence implicating Rab3A or other Rab3 isoforms in exocytosis has also been described for the sea urchin egg (188), the mouse egg (468), parotid gland cells (620), parathyroid gland chief cells (348), anterior pituitary gonadotrophs (757), and for the acrosome reaction in mouse (822), rat (355), and human sperm (862).

Functional data have also implicated Rab4 in the regulated exocytosis of -granules in platelets (695). In addition, more