New Roles of Syntaxin-1A in Insulin Granule Exocytosis and Replenishment*

In type-2 diabetes (T2D), severely reduced islet syntaxin-1A (Syn-1A) levels contribute to insulin secretory deficiency. We generated β-cell-specific Syn-1A-KO (Syn-1A-βKO) mice to mimic β-cell Syn-1A deficiency in T2D. Glucose tolerance tests showed that Syn-1A-βKO mice exhibited blood glucose elevation corresponding to reduced blood insulin levels. Perifusion of Syn-1A-βKO islets showed impaired first- and second-phase glucose-stimulated insulin secretion (GSIS) resulting from reduction in readily releasable pool and granule pool refilling. To unequivocally determine the β-cell exocytotic defects caused by Syn-1A deletion, EM and total internal reflection fluorescence microscopy showed that Syn-1A-KO β-cells had a severe reduction in the number of secretory granules (SGs) docked onto the plasma membrane (PM) at rest and reduced SG recruitment to the PM after glucose stimulation, the latter indicating defects in replenishment of releasable pools required to sustain second-phase GSIS. Whereas reduced predocked SG fusion accounted for reduced first-phase GSIS, selective reduction of exocytosis of short-dock (but not no-dock) newcomer SGs accounted for the reduced second-phase GSIS. These Syn-1A actions on newcomer SGs were partly mediated by Syn-1A interactions with newcomer SG VAMP8.

Pancreatic islet ␤-cells release insulin in a biphasic pattern (1,2). Exocytosis of several pools of insulin secretory granules (SGs) 3 mediated by distinct membrane fusion machineries underlie each of the two phases of glucose-stimulated insulin secretion (GSIS) (1,2). The fundamental components of membrane fusion machinery are three SNARE proteins (syntaxin, SNAP-25 (synaptosome-associated protein of 25 kDa), and VAMP (vesicle-associated membrane protein)) and nSec/ Munc18 (SM) protein, which act to remodel and activate the SNARE complex assembly (3). Each vesicle SNARE (v-SNARE) (VAMP) and target membrane SNARE (t-SNARE) (syntaxins, SNAP25) and SM protein constitute a family of isoforms (4) to enable combinatorial matching of cognate partners that underlie the molecular basis of distinct exocytotic events (4). Interestingly, ␤-cells employ almost all of the major SM⅐SNARE complexes to mediate exocytosis of distinct insulin SG pools. The current thinking is that Munc18a⅐Syn-1A⅐VAMP2 complex mediates the pool of insulin SGs that dock onto plasma membrane (PM) for an indefinite period, called "predocked" SGs, until a strong Ca 2ϩ stimulus evokes fusion (5,6). This pool of predocked SGs accounts for the readily releasable pool mediating first-phase GSIS (1,2). A much larger number of insulin SGs called "newcomer" SGs could be mobilized from the cell interior to undergo fusion but were noted to take only minimal to no residence time at the PM (7,8). Newcomer SGs account for almost all of second-phase GSIS and a substantial amount of first-phase GSIS (7,8). We recently identified the SM⅐SNARE complex mediating newcomer SG fusion as Munc18b⅐Syn-3⅐VAMP8 (9 -11). A smaller population of insulin SGs undergo homotypic SG-SG (compound) fusion (2), which can be potentiated by cAMP-acting glucagon-like peptide 1 (GLP-1) (12) and Ca 2ϩ -acting carbachol (13). SG-SG fusion, mostly in the form of orderly sequential SG-SG fusion in ␤-cells, is mediated by Munc18b⅐Syn-3⅐SNAP25 complex (9,10,14). The remaining SM⅐SNARE complex, Munc18c⅐Syn-4, was initially postulated to mediate biphasic GSIS by acting on predocked SGs in a manner redundant to Munc18a⅐Syn-1A (15,16). Our recent work showed that Munc18c⅐Syn-4 also acts on newcomer SGs (17,18). Furthermore, unlike ␤-cells, which employ redundant SM⅐SNARE complexes for exocytosis, Munc18c⅐Syn-4 in complex with VAMP2 and SNAP23 is the only SM⅐SNARE complex mediating glucose uptake in insulin-sensitive tissues, adipocyte, and muscle (15).
The strongest evidence for the role of Syn-1A in insulin exocytosis has been provided by the study employing a global Syn-1A knock-out (KO) mouse (6), originally generated to examine neuronal plasticity, but it showed remarkably few neuronal phenotypic abnormalities (19). Instead, Syn-1A-KO mice showed profound defects in insulin exocytosis (6), specifically much fewer predocked SGs that were fusion-incompetent, and apparently without perturbation in newcomer SGs. This is of clinical relevance because levels of Syn-1A and cognate proteins are severely reduced in islets of T2D patients, postulated to contribute to insulin secretory deficiency (20). However, the global Syn-1A deletion can potentially affect neuronal and endocrine secretions that might influence ␤-cell function. In fact, recent reports by the original group that created the mouse demonstrated a perturbation in the hypothalamic-pituitaryadrenal axis affecting corticosterone (21) and catecholamine release (22), two hormones that profoundly affect glucose homeostasis by their actions on insulin-sensitive tissues and secretion of islet hormones. Syn-1A is also present in ␣-cells to mediate glucagon secretion (23), which in turn might have paracrine influences on ␤-cells.
It therefore behooves us to unequivocally reassess the role of Syn-1A in ␤-cell insulin exocytosis per se, employing a ␤-cellspecific KO mouse (Syn-1A-␤KO). Our results confirmed that Syn-1A deletion in ␤-cells caused a reduction in number and fusion of predocked SGs, resulting in reduced first-phase GSIS. Unexpectedly, Syn-1A deletion reduced SG replenishment to releasable pools after stimulation and selective reduction in fusion of short-dock newcomer SGs; both underlie the reduced second-phase GSIS. This reduced biphasic GSIS in vivo resulted in hyperglycemia.
There is concern that some rat insulin 2 promoter (RIP)-Cre strains, such as Tg(Ins2-Cre) Mgn , exhibit leaky expression of Cre in the mid-brain and ventral brain regions, including the hypothalamus (24,25). This is important because Syn-1A deletion in hypothalamic neurons can potentially perturb glucose homeostasis (26). We therefore chose another RIP-Cre strain, Tg(Ins2-Cre) Herr (27), which was reported to have a more restricted and punctate pattern of Cre expression in the brain, particularly in the hypothalamus, compared with the robust Cre leakage in other RIP-Cre strains, including the RIP-Cre Mgn and RIP-Cre/ERT. First, we confirmed this Tg(Ins2-Cre) Herr strain by sequencing of the specific 660-bp rat Ins2 promoter. Next, we confirmed that there was no or minimal brain leakage by performing total RNA analysis by quantitative PCR of the hypothalamus of our Syn-1A-␤KO and RIP-Cre control mice using a sensitive quantitative RT-PCR protocol with high dynamic range. Hypothalamic expression levels of Cre were extremely low (supplemental Fig. S1A), with threshold (C t ) values (34.65 Ϯ 1.13) below the cut-off C t value of 30 for positive gene expression in both groups (supplemental Fig. S1B). Consistent with the extremely low Cre expression level, hypothalamic Syn-1A levels were equivalent in both mouse groups, indicating that no deletion of Syn-1A occurred in Syn-1A-␤KO mouse hypothalamus. There was also no difference in expression in the Syn-1A cognate partners, including Munc18a, VAMP2, and SNAP25, or non-cognate SNARE proteins Syn-3 and VAMP8 (supplemental Fig. S1A).
Islets were isolated from Syn-1A-␤KO and control mice (12-14 weeks old) and subjected to Western blotting, which confirmed the reduction in Syn-1A levels without any effect on the levels of cognate Munc18a, VAMP2, and SNAP25 or noncognate SNARE proteins (Fig. 1B). The residual Syn-1A in Syn-1A-␤KO islets was due to Syn-1A present in non-␤-cells shown on confocal microscopy. Here, in control mouse islets, Syn-1A localized to insulin-positive ␤-cells as well as glucagon-positive ␣-cells and somatostatin-positive ␦-cells (Fig. 1C). In Syn-1A-␤KO islets, Syn-1A was no longer present in insulin-positive ␤ cells but still positive in glucagon-positive ␣-cells and somatostatin-positive ␦-cells (Fig. 1D).
To ensure that any defects on ␤-cell insulin secretion are attributable to Syn-1A effect per se, we assessed for possible changes in ␤-cell mass (supplemental Fig. S2A) by several parameters (␤-cells per pancreatic area, islet number and size), which showed no differences between the three mouse groups. We also assessed serum glucagon levels (supplemental Fig. S2B), which were not different between the Syn-1A ␤KO mice and the two groups of control mice, consistent with the intact Syn-1A in the Syn-1A-␤KO mouse ␣-cells (Fig. 1D).
Deletion of Syn-1A in Mouse ␤-Cells Reduces the Number of SGs Docked on the PM and SG Replenishment after Stimulation-Reduction of first-phase GSIS can be predicted to be due to reduction of SG docking and/or fusion of predocked SGs with the PM. To examine SG docking at the PM, we performed EM morphometric analysis on Syn-1A-KO ␤-cells compared with control (Syn-1A flox and RIP-Cre) ␤-cells at basal and after a 15-min 16.7-mmol/liter glucose stimulation (Fig. 4A), the latter to assess SG recruitment to PM to replenish releasable pools after depletion caused by first-phase release. We first ruled out any effect of Syn-1A deletion on insulin SG biogenesis in that the diameters of SGs between the three groups were similar (Fig. 4B), and there was no difference in SG densities in basal conditions or after stimulation (Fig. 4C). Predictably, Syn-1A-KO ␤-cells showed a large reduction of ϳ80% in the number of docked SGs in basal conditions ( Fig. 4D; Syn-1A-␤KO, 0.2 Ϯ 0.06; Syn-1A flox control, 0.91 Ϯ 0.21; RIP-Cre control, FIGURE 2. Syn-1A-␤KO mice exhibit glucose intolerance because of reduced blood insulin levels. A, Syn-1A-␤KO mice are glucose-intolerant because of reduced insulin release into the circulation. IPGTTs were performed on Syn-1A-␤KO (n ϭ 11) versus control mice (RIP-Cre (n ϭ 5) and Syn-1A flox (n ϭ 7)), from which we obtained blood glucose (i) and plasma insulin levels (ii). iii, corresponding AUCs determined above basal levels for glucose (i) and insulin (ii) because basal levels are different between experiments. B, weights of Syn-1A-␤KO mice (n ϭ 20) versus control mice (RIP-Cre (n ϭ 19) and Syn-1A flox (n ϭ 17)) were not different. C, Syn-1A-␤KO mice (n ϭ 14) showed higher blood glucose levels (i) and lower insulin levels (ii) than control mice (RIP-Cre (n ϭ 11) and Syn-1A flox (n ϭ 11)) after an 18-h fast and during the fed state. Results are shown as means Ϯ S.E. NS, not significant. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.
1.05 Ϯ 0.24), which would explain the reduced first-phase secretion, consistent with the results of the global Syn-1A-KO mouse (6). After 15-min glucose stimulation of control ␤-cells to secrete and deplete the releasable pools, there was sufficient recruitment of SGs to the PM to replace the exocytosed SGs and sustain second-phase GSIS. However, glucose stimulation of Syn-1A-KO ␤-cells caused an even greater reduction (than basal conditions) of Ͼ95% in morphologically docked SGs (Syn-1A-␤KO, 0.06 Ϯ 0.02; Syn-1A flox control, 1.23 Ϯ 0.29; RIP-Cre control, 1.35 Ϯ 0.33) (Fig. 4, A and D). Of note, glucose stimulation also caused a larger clearing of SGs further into the cell interior in Syn-1A-␤KO ␤-cells (see image 3 in Fig. 4A) compared with basal conditions. This was assessed by detailed analysis in Fig. 4E showing that Syn-1A-␤KO ␤-cells had a mild reduction in the number of SGs within 0 -0.2 m from the PM under basal conditions, which became very severe after glucose stimulation. In fact, this severe reduction in SGs also occurred within the deeper 0.2-0.4-m concentric shell. Further into the cytoplasm from 0.4 to 1 m, there was a larger accumulation of SGs in Syn-1A-␤KO ␤-cells compared with control ␤-cells. These results indicate that Syn-1A depletion reduced the mobilization of SGs from the cell interior to the PM required to replenish the releasable pools (see Fig. 5) that sustain second-phase GSIS (Fig. 3). This action of Syn-1A on recruitment of SGs to the PM is novel.
Syn-1A-␤KO Mouse ␤-Cells Exhibit Reduced Priming and Mobilization of Insulin SG Pools-We therefore then assessed for the effects of Syn-1A deletion on releasable insulin SG pools by employing patch-clamp membrane capacitance (Cm) measurements of single ␤-cells. Insulin SG exocytosis was elicited   Role of Syntaxin-1A in ␤-Cell Insulin Exocytosis FEBRUARY 10, 2017 • VOLUME 292 • NUMBER 6 by a protocol consisting of a train of 10 500-ms depolarization pulses. Cell Cm changes elicited by the first two pulses have been previously postulated to approximate the size of the readily releasable pool (RRP) of primed and fusion-ready SGs (30). Subsequent pulses estimate the rate of SG refilling or mobilization from reserve pool(s) to the RRP, where SGs are subsequently primed for fusion competence (30). The size of the RRP and rate of SG mobilization are believed to correlate with firstand second-phase GSIS from pancreatic islets (30) . Fig. 5, A and B, shows representative recordings of Cm from Syn-1A flox control and Syn-1A-␤KO mouse ␤-cells. When compared with control ␤-cells, Cm increases in Syn-1A-␤KO ␤-cells were reduced at every depolarizing pulse (Fig. 5C). Fig. 5D shows that the size of RRP of SGs (⌬Cm 1st-2nd pulse ) was reduced by 59% (Syn-1A-␤KO, 1.3 Ϯ 0.2 fF/pF; Syn-1A flox control, 3.2 Ϯ 0.6 fF/pF). As would be predicted from the EM data (Fig. 4), the rate of SG refilling/mobilization (⌬Cm 3rd-10th pulse) was also reduced by 45% (Syn-1A-␤KO, 3.9 Ϯ 0.7 fF/pF; Syn-1A flox control, 7 Ϯ 1.2 fF/pF).

Role of Syntaxin-1A in ␤-Cell Insulin Exocytosis
Because Syn-1A has been shown to bind voltage-gated calcium channel (Ca v s) to form the excitosome complexes with cognate SNARE complex proteins in ␤-cells (31), we examined whether the Syn-1A deletion might cause alterations in Ca v currents. Between Syn-1A-␤KO and control ␤-cells, we observed no significant changes in Ca v current amplitudes (Fig.  5, E and F). This suggests the possibility of redundant alternate syntaxins that might be also acting on ␤-cell Ca v s (32).
Syn-1A Deletion Reduced Fusion of Docked and Newcomer Short Dock SGs-TIRF microscopy (6,10,11,17,18), employed to examine single insulin SG exocytotic behavior in response to physiologic glucose stimulation, has consistently shown heterogeneous SG populations, including not only predocked SGs but also newcomer SGs that undergo little (short dock) to no residence time (no dock) at the PM before fusion (8). Newcomer SGs account for almost all of second-phase GSIS but also a major portion of first-phase GSIS (8). At unstimulated (2.8 mmol/liter glucose) state (Fig. 6A), punctate fluorescence indicating predocked SGs were reduced by 66% in Syn-1A-␤KO ␤-cells (0.059 Ϯ 0.02) compared with Syn-1A flox control ␤-cells (0.173 Ϯ 0.03), which would be consistent with the EM results (Fig. 4). When stimulated with 16.7 mmol/liter glucose, single SG fusion events observed as flashes of fluorescence that rapidly dissipate in a cloudlike diffusion pattern were in three different patterns (8). "Predock" fusion mode (Fig. 6, B (top) and C (black)) refers to SGs already docked onto PM for a period of time before stimulation. Newcomer SGs are SGs appearing de novo after stimulation within the evanescent field that then undergo exocytosis in two patterns. First is immediate exocytosis with a docking state of Ͻ200 ms, which is the minimal interval between two consecutive frames (Fig. 6, B (middle) and C (white)), called "no-dock" newcomer SGs. Second are SGs that dock for some residence time at the PM varying from seconds to minutes before fusion with PM (Fig. 6, B (bottom) and C (gray)), called "short-dock" newcomer SGs.

Syn-1A Mediates Recruitment and Fusion of Newcomer SGs Probably by Interactions with Newcomer SG VAMP8 -The
overriding current thinking is that Syn-1A activation by Munc18a forms a SNARE complex with VAMP2 and SNAP25 to mediate docking and fusion of predocked SGs (5,6). Recently, we reported that VAMP8 is the putative VAMP mediating the recruitment and fusion of newcomer SGs (11) along with cognate t-SNARE Syn-3 (10) and that Syn-3⅐VAMP8⅐SNAP25 SNARE complex assembly could be activated by Munc18b (9). Our recent report showed that Syn-4, thought to prefer VAMP2 to mediate fusion of predocked SGs (15,16), could also bind VAMP8 to mediate fusion of newcomer SGs (18). This led us to postulate that Syn-1A deletion's effects on short-dock newcomer SGs (Fig. 6) might in part be due to the promiscuous binding of Syn-1A to VAMP8. We conducted co-IP experiments with Syn-1A antibody on INS-1 cells maximally stimulated with 16.7 mmol/liter glucose plus 10 nmol/liter GLP-1 to optimally activate SM⅐SNARE complexes (9,11). INS-1 was employed as surrogate for islets to provide an abundance of protein required for co-IP studies (Fig. 7). Under unstimulated conditions (0.8 mmol/liter glucose), Syn-1A coimmunoprecipitated Munc18a and minimal amounts of SNAP25 (and SNAP23) and both VAMPs, VAMP2 and VAMP8 ( Fig. 7Ai; analysis of three experiments shown in Fig.  7Aii, loading controls shown in Fig. 7B). With stimulation, the amounts of SNAP25 and VAMP2 co-immunoprecipitated were increased significantly, by 336 and 161%, respectively, as expected. However, the amount of VAMP8 co-immunoprecipitated was also increased, albeit at a smaller amount of 114%. There was no change in the amounts of Munc18a or SNAP23 co-immunoprecipitated. Because binding of Syn-1A to VAMP8 in INS-1 could be attributed to other accessory proteins, we assessed their direct interactions by co-expressing Syn-1A with VAMP2-GFP or VAMP8-GFP in HEK cells and then performed IP. Syn-1A co-immunoprecipitated similar amounts of VAMP2 and VAMP8 (supplemental Fig. S3A (bottom) shows an analysis of three independent experiments). We conducted the reciprocal study of VAMP2 or VAMP8 IP with GFP anti-body, which co-immunoprecipitated similar levels of Syn-1A (supplemental Fig. S3B (bottom) shows analysis) as well. The abundance of Syn-1A⅐VAMP2 complexes exceeded Syn-2⅐VAMP8 complexes in the more physiologic INS-1 model (Fig.  7Aii), suggesting that either other accessory proteins (i.e. Munc18s and others) in native ␤-cells affect SNARE complex  FEBRUARY 10, 2017 • VOLUME 292 • NUMBER 6 assembly, or the more rapid priming and fusion of newcomer SGs suggest faster Syn-1A⅐VAMP8 complex disassembly. Taken together, it appears that specific exocytotic events (predocked versus newcomer SGs mediating first-and secondphase GSIS) may not be dictated solely by syntaxins (Syn-1A, Syn-3, and Syn-4) but also by the v-SNAREs (VAMP2 and VAMP8) that each Syn interacts with.

Discussion
This work shows that Syn-1A mediated not only the number and fusion competence of predocked insulin SGs (6) but also the recruitment of newcomer SGs to explain the reduction of biphasic GSIS observed in vivo and in vitro from pancreatic islets of Syn-1A-␤KO mice. Exocytosis of predocked insulin SGs are now known to account for only half of first-phase GSIS, whereas the other half arises from newcomer SGs, and secondphase GSIS is almost entirely attributable to newcomer SGs (8,10,11). It is likely that reduction of both exocytotic events would quantitatively account for the disappearance of firstphase GSIS in T2D (33), which has been attributed to Ͼ70% reduction in islet Syn-1A levels (20). Moreover, in T2D patients, second-phase GSIS is also reduced or rendered less efficient to the increasing glycemic demand (33), which probably arises from defective exocytosis of newcomer SGs. Our results suggest that this second-phase defect in T2D might also be contributed by Syn-1A deficiency. It is tempting to simply restore Syn-1A levels in the T2D islets to rescue the deficient GSIS (6). However, a previous study suggested that overexpression of Syn-1A even slightly over normal levels actually reduced GSIS (34); that was probably due to excess formation of fusionincompetent incomplete SNARE complexes.
We partly elucidated the underlying mechanism to explain the exocytotic defects caused by Syn-1A deficiency. As expected, preferential binding of Syn-1A to VAMP2 formed the exocytotic SNARE complex with SNAP25 to mediate exocytosis of predocked SGs (3,6). Consistently, we saw a reduced number of insulin SGs docked onto the PM, and fusion of predocked SGs was almost entirely abrogated. We had reported that Syn-3 preferentially binds VAMP8, and the Syn-3⅐VAMP8 SNARE complex mediated both no-dock and short-dock newcomer SGs but without any effect on predocked SGs (10,11). We postulated that Syn-1A could have promiscuous binding to VAMP8 to in part explain the effects of ␤-cell Syn-1A-KO on short-dock newcomer SGs. Indeed, Syn-1A directly binds VAMP8, and Syn-1A⅐VAMP8 complex formation increased following stimulation. Thus, v-SNARE/t-SNARE pairing of Syn-1A⅐VAMP8 complex, like the Syn-3⅐VAMP8 complex, could mediate the recruitment and fusion of newcomer SGs. The binding of Syn-1A to VAMP2 and VAMP8 was mimicked by Syn-4 (18), which also affected both predocked and newcomer SGs. Peculiarly, Syn-4 depletion in human ␤-cells affected predominantly no-dock newcomer SGs (18), whereas Syn-1A-KO had preference on abrogating fusion of short-dock SGs. These studies taken together suggest that the mode of exocytosis is not solely dictated by t-SNARE syntaxins per se, but also by their respective affinity to SG v-SNAREs, VAMP2 preferring predocked SGs and VAMP8 preferring newcomer SGs (11). However, the Syn-1A⅐VAMP8 complex may be binding a distinct set of accessory proteins that influence docking and fusion kinetics differently from those mediated by Syn-4⅐VAMP8 or Syn-3⅐VAMP8 complexes (discussed further below). It is also possible that Syn-1A might be acting on an undefined VAMP(s) other than VAMP2 and VAMP8.
Why do predocked SGs dock for a long time on the PM, whereas newcomer SGs require little (short dock) to no docking time before fusion? It has been hypothesized that docking is not even required but simply a temporal constraint for SG fusion (7). One explanation is the differential coupling of syntaxins (Syn-1A, Syn-3, and Syn-4) to VAMP8 that in part determines which newcomer SG population, short-dock versus no-dock, is preferred for fusion (10, 18) (this study). It is likely that additional unknown accessory proteins influence the assembly of these different SNARE complexes (3,35), which in turn control the kinetics of the purported SNARE complex "zippering" mechanism mediating the completion of membrane fusion (36), which may be faster in no-dock than in short-dock SG fusion. Taken together, distinct Syn⅐VAMP SNARE fusion complexes must be influenced by accessory proteins (defined and undefined) that modulate their distinct priming and fusion kinetics.
We do not have a mechanism of how Syn-1A deficiency reduces the replenishment of SGs to sustain second-phase GSIS. Because this occurs in the cell interior and not on the PM per se, it is probably independent of SNARE complex formation that occurs at the SG-PM interface. We have just reported that calcium sensor synaptotagmin-7, which Syn-1A is known to interact with, can replenish insulin SGs to the PM (37). GLP-1 is known to increase recruitment of newcomer SGs; thus, it is likely that some of the additional accessory proteins (i.e. RIM, Epac2, and snapin) activated by the cAMP pathway in ␤-cells (38 -40) may interact with Syn-1A to recruit insulin SGs. It is thus evident that these undefined areas of investigation of newcomer SGs are "newcomers" to the broader exocytosis field, and there are many gaps in knowledge that need to be more vigorously pursued.
How do we explain the apparent discrepancy between this study employing the more rigorous Syn-1A-␤KO mouse and the previous study using a global Syn-1A-KO mouse (6)? The earlier study (6) did not distinguish between the larger population of no-dock newcomer SGs (in which both studies showed no effects) and the smaller population of short-dock newcomer SGs in which our study showed a reduction. In the previous study, it was initially assumed that there was no neuronal phenotype, yet the global Syn-1A-KO mouse was later found to have defects in the hypothalamic-pituitary-adrenal axis affecting corticosterone (21) and catecholamine release (22), which in turn could well induce acute or chronic direct and indirect effects on ␤-cells that could influence insulin secretion. We showed in our Syn-1A-␤KO mice that the hypothalamus that could have been affected by Cre expression did not affect Syn-1A levels. Syn-1A is ubiquitously present in almost all endocrine cells, including ␣-cells (23) and ␦-cells adjacent to ␤-cells, whose secretory products are well known to influence ␤-cell insulin secretion (reviewed in Ref. 41). The global Syn-1A-KO (6) could have affected ␣and ␦-cell secretion, which in turn influenced ␤-cell secretion in a manner different from our model. We showed that Syn-1A levels are intact in ␣and ␦-cells and that serum glucagon levels were not perturbed by the ␤-cell Syn-1A deletion.
The two last points we would like to make are with regard to the reduced Syn-1A levels found in human T2D islets (20) that underlie the rationale to pursue this study and our rationale for having used multiple strategies to elucidate the precise exocytosis steps mediated by Syn-1A in the ␤-cell. With regard to the first point, we also found reduced islet Syn-1A levels in two of the most frequently used T2D rodent models, the obese Zucker rat (42) and non-obese Goto-Kakizaki rat (43). Another laboratory (44) showed Syn-1A gene expression as reduced in T2D patient islets (9 T2D patients versus 55 non-diabetic donors). However, a report by Marselli et al. (45) using laser microdissection of islet cells reported that T2D human ␤-cells had normal Syn-1A levels.
In this study, we have opted to employ multiple strategies to take advantage of the major strength of each method and offset the inherent weakness of another; thus, these methodologies would be complementary, and results would be mutually confirmatory and therefore more unequivocal. The often used patch clamp depolarization-induced exocytosis assay does not mimic glucose-mediated exocytosis. Initial depolarizations (first two pulses) would release the predocked SGs located in close vicinity of the Ca 2ϩ channels first (designated as the RRP), and the larger Ca 2ϩ influx from subsequent depolarization pulses would diffuse further into the cytoplasm to mobilize and induce fusion of SGs located further into the cell interior (30), which would cause fusion of predominantly newcomer SGs. Although patch clamp protocols were designed to select out the fusion of predocked SGs within the RRP in a manner "disrupted" from newcomer SGs, which has been a widely accepted assumption, this may no longer be the case. We have reported in our VAMP8 KO mouse ␤-cell study (11) that VAMP8 mediated only newcomer SG fusion but appeared to contribute substantially to the RRP using patch clamp capacitance measurements. Thus, to unequivocally distinguish between the population of predocked and newcomer SG fusion and in response to physiologic stimulus glucose, we have preferred the TIRF microscopy approach to unequivocally show these single SG events at the highest spatio-temporal resolution. Nonetheless, TIRF microscopy is limited to visualizing the events at or close to the PM and could not unequivocally assess the mobilization of SGs per se from the cell interior. Strategies to assess SG mobilization and replenishment from the cell interior are very limited, and we believe the strongest one would be by EM after stimulation to deplete as much of the SGs close to the PM as possible. If Syn-1A deletion causes only a docking and/or fusion defect at the PM, then SGs will accumulate near the PM. However, we found in the Syn-1A-KO ␤-cells that there was a very major reduction in SGs within 0.4 m from the PM after glucose stimulation and accumulation of SGs further inside the cytoplasm (0.4 -1 m). This result indicates that SGs are not being mobilized from the cell interior to the PM. Therefore, Syn-1A deletion caused a mobilization defect in the SGs at least in the context of the islet ␤-cell, which is a novel finding and a new secretory function for Syn-1A.
Experimental Procedures Generation of ␤-Cell-specific Syn-1A-KO Mice-We used conventional embryonic stem (ES) cell targeting to generate pluripotent ES cells containing a floxed allele of Syn-1A. Using homologous recombination with pNeoLoxPTK targeting vector, loxP sites were inserted to flank exons 2 and 3 of Syn-1A (Fig. 1A). Correct targeting was confirmed by Southern blotting analysis, and chimeras were generated by aggregating four independent ES cell clones with 8-cell stage embryos. Male chimeras were used for germ line transmission breeding. Mice expressing Cre recombinase under the control of a promoter that drives expression only in islet ␤-cells, transgene mice Tg(Ins2-Cre) Herr called RIP-Cre mice, here were used for breeding with Syn-1A floxed mice to generate ␤-cell-specific KO of Syn-1A (Syn-1A-␤KO). 12-14-week-old Syn-1A floxed and Tg(Ins2-Cre) Herr mice were backcrossed to C57BL/6J mice for eight generations before offspring were used for experiments. All procedures were in accordance with Canadian Council on Animal Care Standards and approved by the University of Toronto Animal Care Committee.
IPGTT-IPGTTs (2 g of glucose/kg body weight) on 12-week-old male mice were performed after an 18-h fast. Blood samples were then collected from the tail vein without anesthesia. Insulin levels were determined by a radioimmune assay (EMD Millipore Corp., Billerica, MA).
Electron Microscopy-As described previously (10,11), mouse islets were fixed for 1 h with Karnovsky style fixative (3.2% paraformaldehyde, 2.5% glutaraldehyde) in 0.1 mol/liter sodium cacodylate buffer (pH 6.5) with 5 mmol/liter CaCl 2 , postfixed for 30 min with 1% osmium tetroxide. Samples were embedded in 1% uranyl acetate (1 h), dehydrated, and infiltrated with Epon 812 resin. Polymerization was completed by epoxy resin, forming a solid epoxy disk, which was subjected to ultrathin sectioning (80 nm) using a Reichert Ultracut E5 microtome. Slices were collected on 200-mesh copper grids, counterstained (15 min) using saturated 4% uranyl acetate followed by Reynold's lead citrate, and then examined in a Hitachi H-7000 transmission electron microscope (Krefeld, Germany) at an accelerating voltage of 75 kV and photographed with an AMT XR-60 camera and software. For morphometric analysis (ImageJ), concentric shells of 0.2 m from the PM to the cell interior up to 1.5 m were drawn, and then the centers of SGs within the shell were determined and counted.
TIRF Imaging-TIRF microscopy was performed as reported (9 -11) using a Nikon TE2000U TIRF microscope (Nikon Canada, Mississauga, Canada), with images obtained at 5 Hz, 100-ms exposure time. Large round cells were chosen as ␤-cells, which were confirmed by their response to high glucose stimulation. Fusion events observed as flashes of fluorescence indicating emptying of NPY-EGFP cargo (mouse ␤-cells infected with Ad-NPY-EGFP) were manually selected. Two concentric circles (5 and 7 pixels with a pixel size of 267 nm, corresponding to ϳ1.3and 1.8-m diameter) were used to center on selected SGs. Evolution of fluorescence changes over time of single SGs was analyzed by Matlab (Math-Works, Natick, MA), ImageJ (National Institutes of Health, Bethesda, MD), and Igor Pro software (WaveMetrics), whereby dissipation of average fluorescence in the concentric annulus was considered to be release of SG cargo. Before image acquisition, cells were preincubated for 30 min in KRB buffer containing 2.8 mmol/liter glucose and then stimulated with 16.7 mmol/liter glucose for 12 min.
INS-1 cells (70 -75% confluence) were treated as indicated; then 1 mg of protein of INS-1 lysate was subjected to immunoprecipitation with Syn-1A antibody (2 g) cross-linked to protein A-agarose beads (Molecular Probes, Inc., Eugene, OR) performed as described previously (9,11). Co-precipitated proteins were identified by Western blotting. For accurate comparison, VAMP2 and VAMP8 (also SNAP25 and SNAP23) were probed simultaneously and separated well enough on PAGE.
HEK cells were transfected with Syn-1A and then infected with Ad-VAMP2-EGFP or Ad-VAMP8-EGFP virus. 2 days