Neurexin-1α Contributes to Insulin-containing Secretory Granule Docking*

Background: The cell surface, transmembrane protein neurexin may play a role in insulin secretion. Results: Knockdown and knock-out of neurexin-1α, the predominant β-cell isoform, increased insulin secretion and impaired secretory granule docking. Conclusion: Neurexin-1α helps mediate insulin granule docking and thereby constrains secretion. Significance: Neurexin-1α could couple localization and functioning of the insulin docking and secretory machinery to extracellular protein interactions. Neurexins are a family of transmembrane, synaptic adhesion molecules. In neurons, neurexins bind to both sub-plasma membrane and synaptic vesicle-associated constituents of the secretory machinery, play a key role in the organization and stabilization of the presynaptic active zone, and help mediate docking of synaptic vesicles. We have previously shown that neurexins, like many other protein constituents of the neurotransmitter exocytotic machinery, are expressed in pancreatic β cells. We hypothesized that the role of neurexins in β cells parallels their role in neurons, with β-cell neurexins helping to mediate insulin granule docking and secretion. Here we demonstrate that β cells express a more restricted pattern of neurexin transcripts than neurons, with a clear predominance of neurexin-1α expressed in isolated islets. Using INS-1E β cells, we found that neurexin-1α interacts with membrane-bound components of the secretory granule-docking machinery and with the granule-associated protein granuphilin. Decreased expression of neurexin-1α, like decreased expression of granuphilin, reduces granule docking at the β-cell membrane and improves insulin secretion. Perifusion of neurexin-1α KO mouse islets revealed a significant increase in second-phase insulin secretion with a trend toward increased first-phase secretion. Upon glucose stimulation, neurexin-1α protein levels decrease. This glucose-induced down-regulation may enhance glucose-stimulated insulin secretion. We conclude that neurexin-1α is a component of the β-cell secretory machinery and contributes to secretory granule docking, most likely through interactions with granuphilin. Neurexin-1α is the only transmembrane component of the docking machinery identified thus far. Our findings provide new insights into the mechanisms of insulin granule docking and exocytosis.

␤ cells share many similarities with neurons, including the expression of many of the same scaffolding and exocytotic proteins (1,2). These proteins mediate the exocytosis of synaptic vesicles in neurons and insulin granules in ␤ cells (3). Advances in the neurobiology field regarding synaptic maturation and function thus frequently offer important clues regarding overall ␤-cell function and the mechanisms underlying insulin secretion.
We previously found that ␤ cells express NRXNs and one of their major postsynaptic binding partners, neuroligins (19). Separately, NRXN1␣ was one of the most abundant transcripts identified in a systematic study of human tissue mRNAs designed to identify highly expressed, membrane-associated, human islet-specific proteins (i.e. without expression in kidney, liver, or exocrine pancreas) (20).
Studies with double and triple ␣-NRXN KO mice have indicated that ␣-NRXNs in neurons are essential for the organization and stabilization of the presynaptic machinery (21)(22)(23). In vitro work suggests that NRXNs interact with the synaptic vesicle-associated protein rabphilin-3A via CASK (24) and contribute to synaptic vesicle docking (16,25). ␤ cells also express CASK (19) and, in addition, granuphilin, a rabphilin-3 homologue implicated in insulin granule docking (26,27). Based on the similarities between neurotransmitter exocytosis and insulin secretion, we hypothesized that NRXNs in the ␤-cell are constituents of the insulin secretory machinery, perhaps helping to mediate secretory granule docking via interactions with granuphilin.
In the present study we examined the role of NRXN1␣ in ␤-cell function. Our data show that NRXN1␣ is an integral component of the secretory granule docking machinery. Like other proteins that contribute to the granuphilin-mediated docking of secretory granules at the ␤-cell membrane, NRXN1␣ inhibits insulin secretion (28,29). Our findings provide new insights into the mechanisms of insulin granule exocytosis.
To detect NRXN expression by immunoblotting, we raised a polyclonal, pan-NRXN antibody against a previously described NRXN peptide (15). Rabbits were injected with the keyhole limpet hemocyanin-conjugated peptide CAKSANKNKKNKD-KEYYV, and serum was affinity purified (Open Biosystems). The antibody was validated for peptide binding by ELISA and competition assay. The antibody was also validated by Western blot and detects all six CFP-tagged NRXN isoforms (30) (NRXN1␣ and 2␣ shown in Fig. 1B, lanes 8 and 9). The identity of the bands was confirmed with an anti-GFP/CFP antibody (data not shown).
Plasmid Constructs-The cDNA construct encoding CFP was previously described (31). The cDNA constructs encoding eGFP-tagged NRXN1␣ (32) and each of the six isoforms of CFP-tagged NRXN (30) were a gift from Dr. Markus Missler (University of Munster) and Dr. Ann Marie Craig (University of British Columbia), respectively. The NRXN1␣ expression vector and an empty vector control were derived from the pCMV-eGFP-NRXN1␣ vector by removing regions encoding eGFP and eGFP-NRXN1␣, respectively. The cDNA vector pSEAP2 encoding secreted alkaline phosphatase (SEAP) was from Clontech.
Animals-NRXN1␣ KO mice were bred from previously described NRXN1␣ and 2␣ heterozygous KO mice (21) purchased from Jackson Laboratory. Genotyping was performed by PCR using primer pairs described on Jackson's website and was confirmed by RT-PCR using RNA isolated from brain and islets as previously described (19). Primers for RT-PCR are described in supplemental Table S1. Islets were isolated from age-matched KO and WT mice at ages 10 -14 weeks as previously described (33), hand-picked and cultured overnight in 8 mM glucose before use in experiments. All procedures were approved by the University of California, San Diego Institutional Animal Care and Use Committee.
Histology-Pancreas tissue was obtained from adult male Sprague-Dawley rats, C57BL/6 mice, and NRXN1␣ KO mice, fixed for 24 h in Pen-Fix (Thermo Fisher Scientific) and embedded in paraffin. Samples of human pancreas tissue from healthy donors were obtained from the National Disease Research Interchange, fixed in 10% buffered formalin, and embedded in paraffin. Sections (6 m) of human, rat, and mouse pancreases were prepared by the University of California, San Diego histology core. Deparaffinized sections were subjected to antigen retrieval by treating with a citrate buffer. H&E staining was performed using standard methods. Immunostaining with the Vectastain Elite ABC kit and NovaRED substrate (Vector Laboratories) was performed as previously described (37). Immunofluorescence was performed using 5% serum in PBS for blocking and antibody dilutions. Primary antibodies (4 ºC overnight) and secondary antibodies (room temperature for 1 h) were applied in a humidified chamber. Controls for immunostaining and immunofluorescence included tissue stained in parallel with secondary antibody alone. Images were captured using a Nikon Eclipse E800 microscope and SPOT RT S.E. camera at ϫ20.
Western Blotting-Total cell extract was prepared by lysing cells in RIPA buffer containing 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl (pH 7.5), 2 mM EDTA, 1 mM PMSF, and 1% Protease Inhibitor Mixture (Sigma-Aldrich). Plasma membrane was separated from cytosolic fractions as previously described by differential centrifugation and detergent/aqueous partitioning using the Plasma Membrane Protein Extraction kit (BioVision) per the manufacturer's instructions (38). Proteins were quantified using the D C Protein Assay (Bio-Rad). NuPAGE LDS Sample Buffer and NuPAGE Reducing Agent (Invitrogen) were added to protein at a 1ϫ final concentration. Protein was then incubated at 70 ºC for 10 min and electrophoresed on either 4 -12% Bis-Tris or 3-8% Tris-acetate NuPAGE Gels (Invitrogen). Protein was transferred to PVDF membrane, blocked with 5% milk in PBS and probed with antibodies in 5% milk in PBS containing 0.1% Tween 20. Membranes probed with HRP-labeled secondary antibodies were incubated with PS-3 chemiluminescent detection reagent (Lumigen) and the chemiluminescent signal cap- , and rat brain as well as B, human islets and C, mouse islets were quantified by absolute qPCR using a set of standard curves for each reaction with known NRXN copy numbers. Rat islet data are also represented as an inset in panel A. Data are represented as mean Ϯ S.E. from three individual tissue, islet or culture preparations assayed in triplicate. tured by HyBlot CL autoradiography film (Denville Scientific). Membranes probed with IRDye-conjugated secondary antibodies were directly imaged and bands quantified with the Odyssey Infrared Imaging System and Software (LI-COR).
Immunoprecipitation (IP)-Cells were lysed in a buffer containing 150 mM NaCl, 1% Nonidet P-40, 50 mM Tris (pH 8), 1 mM PMSF, and 1% protease inhibitor mixture. Lysate was precleared with rec-protein G-Sepharose (Invitrogen) and then incubated overnight at 4 ºC with 5 g of goat anti-NRXN1 P-15 antibody or purified goat IgG (Millipore). The reactions were incubated with rec-protein G-Sepharose at 4 ºC for 2 h and beads were thoroughly washed with lysis buffer. Samples were combined with LDS Sample Buffer and Reducing Agent for Western blotting as described above.
Real-time Quantitative PCR (qPCR)-Brain tissue was obtained from adult male Sprague-Dawley rats. Freshly isolated islets from adult male Sprague-Dawley rats and Swiss Webster mice were provided by the University of Washington Diabetes and Endocrinology Research Center Islet Cell and Functional Analysis Core. Human islets isolated as previously described (39) were obtained from the Southern California Islet Cell Consortium and the University of Alabama at Birmingham through the National Institutes of Health-sponsored Islet Cell Resources Basic Science Islet Distribution Program and flash frozen after 1 to 3 days in culture. Pancreases were procured from heart-beating, cadaveric, non-diabetic, female donors ages 48 and 49 with body mass indices of 34 and 23, respectively. Islet preparations were deemed Ն 80% purity and Ն 90% viability. Total RNA was isolated and reverse transcribed as previously described (19). Gene-specific primers were used to perform qPCR with Power SYBR Green PCR Master Mix (Applied Biosystems), PerfeCTa SYBR Green FastMix (Quanta BioSciences) or 2ϫ qPCR Master Mix (BioPioneer) on an ABI 7500 Fast Real-Time PCR System (Applied Biosystems). Each sample was analyzed in triplicate along with no-RT and no-template controls. For relative qPCR, values were normalized to 18 S RNA. For absolute qPCR, standard curves of human, rat, and mouse NRXN gene-and isoform-specific amplicons were generated by PCR and gel purified with a QIAquick gel extraction kit (Qiagen). Normalization was carried out using total RNA values and confirmed with qPCR of 18 S RNA. NRXN primers were designed to avoid major splice variant regions. All primers were designed using Primer 3 (40) or PerlPrimer (41) software and are described in supplemental Table S1.
RNAi-For NRXN1 RNAi experiments, two pools of siRNAs targeting NRXN1 (NRXN1 pools 1 and 2) and two non-targeting control pools of siRNA (non-targeting pools 1 and 2) comprised of completely different siRNAs were purchased from Dharmacon Research (Thermo Fisher Scientific). For NRXN2 RNAi experiments, a pool of siRNAs targeting NRXN2 (NRXN2 pool A) was purchased from Sigma-Aldrich and a pool of non-targeting control siRNAs (non-targeting pool A) was purchased from Dharmacon Research (Thermo Fisher Scientific). All sequences are listed in supplemental Table S2.
Insulin Secretion and Glucose Stimulation-INS-1E cells were switched to antibiotic-free media containing 5 mM glucose for ϳ18 h and then to 2.75 mM glucose Krebs-Ringer bicarbonate buffer for 1 h. Fresh Krebs-Ringer buffer containing 2.75, 15, or 16.7 mM glucose was then applied for 1 h as previously described (19). For static studies, 10 similar-sized islets (per well in 96-well plates) were stimulated as above. Insulin in the cell lysate and media was measured using a rat insulin RIA (Millipore) or ultrasensitive rat insulin ELISA kit (Crystal Chem). For SEAP (secreted alkaline phosphatase) co-transfection studies, cell lysate and media were also assayed for SEAP as described (42). Both insulin and SEAP secretion are represented as media content normalized to total intracellular content. Total cellular insulin content in lysate was normalized to total cellular protein. Perifusion studies were performed by the University of Washington Diabetes and Endocrinology Research Center Islet Cell and Functional Analysis Core as previously described (43).
EM-INS-1E cells and isolated islets in normal culture media were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) for at least 4 h, postfixed in 1% osmium tetroxide in 0.1 M cacodylate buffer for 1 h and stained en bloc in 1% uranyl acetate for 1 h. Samples were dehydrated in ethanol, embedded in epoxy resin, sectioned at 60 to 70 nm and stained with uranyl acetate and lead nitrate. Grids were viewed using a transmission electron microscope (1200EX II, JEOL) and photographed using a digital camera (Gatan). Cells were chosen at random, and islet morphometry was analyzed after deidentification of experimental groups. Distances were determined by drawing a straight line from the granule center to the nearest plasma membrane. All measurements were performed in ImageJ (44).

RESULTS
NRXN Protein Is Present in ␤ Cells-We previously demonstrated the presence of NRXN in the rat pancreatic ␤ cell by immunofluorescence (19). ␤-Cell expression of NRXN protein was confirmed by immunostaining of human, rat, and mouse pancreas sections (Fig. 1A). NRXN1␣ protein expression in INS-1E ␤ cells was confirmed by immunoprecipitation of NRXN1␣ from cell lysate followed by immunoblotting using a different, pan-NRXN antibody (Fig. 1B, lane 5). Immunoblot analysis of INS-1E cells demonstrated the presence of ␣-NRXN in the plasma membrane fraction, consistent with the plasma membrane localization of NRXN observed in our previous immunofluorescent staining (19). NRXN was not detected in the cytosolic fraction (Fig. 1B, lanes 1-2). As part of validation of the pan-NRXN antibody, it was also used to immunoblot CFP-tagged NRXN expressed in transfected COS cells (Fig. 1B,  lanes 8 -9). The expected increase in NRXN molecular weight due to the CFP tag is not obvious because of the high molecular mass of ␣-NRXNs (ϳ180 kDa) relative to CFP (ϳ27 kDa).
NRXN Transcript Levels in ␤ Cells-We next sought to identify the NRXN isoforms present in ␤ cells. Because existing antibodies cannot distinguish between the three NRXN genes, we used absolute RT-qPCR to quantify transcripts of each isoform in rat islets, rat brain, as well as the rat ␤-cell lines INS-1E and 832/13 ( Fig. 2A). Brain expresses similar levels of all NRXN isoforms except for a slight enrichment of NRXN1␣. Islets and ␤ cells demonstrated a distinctly different pattern, showing a clear predominance of one or two isoforms. In INS-1E and 832/13 ␤ cells, NRXN 1␣ and 2␤ predominate. In rat islets, NRXN1␣ is the predominant transcript ( Fig. 2A, inset). Human (Fig. 2B) and mouse (Fig. 2C) islets displayed a similar expression pattern, with NRXN1␣ transcript being most abundant, followed by lower expression of NRXN 2␤. Because NRXN1␣ is the most abundant isoform in primary islets, it was chosen as the focus for further investigations.
Coimmunoprecipitation of NRXN1␣ with Docking and Submembrane-resident Exocytotic Proteins-To identify NRXNinteracting proteins in INS-1E cells, we immunoprecipitated NRXN1␣ and analyzed coprecipitating proteins by Western blot. In neurons, NRXN seeds the assembly of the exocytotic machinery through interactions with Mint 1 (13), Velis (13,14), CASK (17,18,45), syntaxin 1 (15), and Munc18 (14). Consistent with NRXN being a constituent of the insulin granule submembrane secretory apparatus in ␤ cells, we found coprecipitation of NRXN1␣ with the submembrane exocytotic proteins syntaxin 1, CASK, and Munc18 (Fig. 3). In contrast, no co-precipitation with synaptophysin was detected even in overexposed immunoblots, suggesting that NRXN does not bind directly or indirectly to this vesicular protein. These results suggest NRXN is a component of the multiprotein, submembrane com- plex that regulates the plasma-membrane docking and exocytosis of insulin granules.
In neurons, NRXNs likely contribute to synaptic vesicle docking via interactions with rabphilin-3A (24). ␤ cells express the rabphilin-3-like protein granuphilin, which associates with secretory granules (27). Granuphilin on the surface of insulin granules directly interacts with syntaxin-1 and participates in secretory granule docking and inhibition of SNARE-mediated insulin release (26,46). Both isoforms of granuphilin, the fulllength granuphilin-a and the shorter splice variant granuphilin-b, are expressed in INS-1 cells at similar levels and are reported to function in the same manner (47). Immunoprecipitation of NRXN1␣ revealed its interaction with both isoforms of granuphilin in ␤ cells (Fig. 3, bottom panel).

Knockdown of NRXN1␣ in INS-1E Cells Increases Insulin Secretion, Insulin Content, and Insulin Gene Expression-
Granuphilin-mediated secretory granule docking is thought to act as a "temporal constraint" on insulin secretion, inhibiting SNARE-mediated fusion of insulin granules with the plasma membrane (26,48). Because of this, experimental knockdown and KO of granuphilin increased insulin secretion (48,49). If NRXN is also necessary for docking, its depletion should similarly increase insulin secretion. We thus used siRNA to knock-down NRXN1 in INS-1E cells and determined the effect on glucose-stimulated insulin secretion. siRNA treatment of INS-1E cells resulted in a 78% decrease in NRXN1␣ transcript levels ( Fig. 4A) without increasing the levels of other ␣-NRXN transcripts (data not shown). A similar depletion of NRXN1␣ protein was also observed (Fig. 4B; the top band identified by N.S. reacted nonspecifically with the pan-NRXN antibody). NRXN1 silencing resulted in a 50% increase in insulin secretion at high glucose (Fig. 4C) and a 62% increase in the stimulation index (the ratio of insulin secretion at high versus low glucose) (Fig. 4D). Basal insulin secretion was not significantly changed. Insulin secretion experiments were repeated with similar results using a completely different pool of both control and NRXN1 siRNAs. Interestingly, the increased insulin secretion (measured as % of total cellular content) after NRXN1 knockdown was accompanied by up-regulation of cellular insulin content and insulin 2 mRNA by 31 and 37% over control, respectively (Fig. 4, E and F). Insulin content and mRNA were measured in INS-1E cells after an overnight incubation in 5 mM glucose. To determine the effect of glucose on the observed increases in cellular insulin content, NRXN1 knockdown cells were incubated for 18 h in different glucose concentrations prior to insulin assay. We found that 2.8 mM glucose, but not 11.2 mM, recapitulated the increase in insulin content (data not shown).
To confirm that NRXN1 knockdown selectively impacts the regulated secretory pathway, we also tested the effect of NRXN1 knockdown on constitutive secretion using secreted alkaline phosphatase (SEAP). SEAP is a recombinant cargo protein commonly used to assess constitutive secretion (42,50). We found that NRXN1 knockdown did not affect SEAP secretion at high glucose (Fig. 4G), suggesting that insulin secretion was selectively modulated by NRXN1.
To complement the knockdown study, we also overexpressed NRXN1␣ to investigate its effect on insulin secretion. Transient transfection of INS-1E cells with a NRXN1␣ overexpression vector resulted in an increase in NRXN1␣ protein (as in Fig. 6) but had no effect on insulin secretion (data not shown).
To determine if NRXN isoforms other than NRXN1␣ also regulate insulin secretion and content, we conducted similar experiments using siRNAs targeting NRXN2␤, the other abundant isoform in INS-1E cells ( Fig. 2A). In INS-1E cells where NRXN2␤ transcript was depleted by 72% (supplemental Fig.  S1A), we found a 67% increase in insulin secretion at high glucose (supplemental Fig. S1B), a 91% increase in the stimulation index (supplemental Fig. S1C), and 26% increase in cellular insulin content (supplemental Fig. S1D). Basal insulin secretion was not significantly changed.
Knockdown of NRXN1 Impairs Secretory Granule Docking-We next asked whether NRXN1, like granuphilin, contributes to secretory granule docking at the ␤-cell membrane. Using EM, we examined the distribution of secretory granules in INS-1E cells with and without NRXN1 siRNA treatment (Fig.  5A). Given their average diameter of ϳ350 nm, secretory granules whose centers reside within 100 -200 nm of the ␤-cell membrane are considered docked (26,51,52). We categorized granules into four groups based on their distance from the plasma membrane. Compared with control siRNA-treated cells, NRXN1 siRNA-treated cells had 32% fewer granules in the docked (100 -200 nm) category but significantly more in the Ͼ400 nm category (Fig. 5B). Consistent with the finding of greater insulin content after NRXN1 siRNA treatment (Fig.  4G), knockdown also increased insulin granule number per cell cross section by 57% (Fig. 5C).
NRXN Protein Levels Decrease at High Glucose-We next asked whether glucose affects NRXN1␣ protein levels. We found that after a 1 h treatment with high glucose, endogenous NRXN1␣ protein levels were decreased by 33% compared with low-glucose-treated controls (Fig. 6, left panel). High glucose also decreased the expression of transfected NRXN1␣ protein by a similar 40% (Fig. 6, right panel). Though this decrease in NRXN1␣ was unexpected, it is noteworthy that glucose transporter 2, another ␤-cell membrane protein, has been reported to undergo internalization and degradation upon glucose stimulation (53).

Loss of NRXN1␣ Increases Insulin Secretion in Isolated Mouse
Islets-To further examine the role of NRXN1␣ in insulin secretion, islets were isolated from NRXN1␣ KO mice and agematched WT controls. RT-PCR was performed using RNA from the isolated islets to confirm the absence of NRXN1␣ transcript and the presence of both NRXN2␣ and 2␤ transcripts (supplemental Fig. S2A). Immunohistochemistry of pancreas sections from these mice revealed normal islet architecture despite loss of NRXN1␣ (supplemental Fig. S2B).
Assessment of insulin secretion from isolated islets was performed to investigate the effect of NRXN1␣ KO on glucosestimulated insulin secretion. Islets from KO mice had a 184% increase in insulin secretion at high glucose (Fig. 7A) and a 143% increase in the stimulation index (Fig. 7B). Basal secretion was not significantly changed. Isolated islets were next perifused to compare insulin secretion rates over time after highglucose stimulation. Overall, the insulin secretion rate (ISR) of NRXN1␣ KO islets was higher than that of WT controls after glucose stimulation (Fig. 7C). Although the increase in the area under the curve (AUC) of first-phase (0 -5 min) secretion by KO islets was not statistically significant, the AUC of second-phase secretion (10 -40 min) by the KO islets was significantly (79%) higher than with WT islets (Fig. 7D). A separate, post-hoc analysis of insulin secretion during the peak of first phase secretion revealed that the ISR of KO islets was significantly greater than the ISR of WT islets (mean 40% increase, p ϭ 0.02) during the interval consisting of the 3 to 5 min time points.
Islets from NRXN1␣ KO Mice Have Reduced Secretory Granule Docking-We next sought to determine if, as was the case in siRNA-treated INS-1E cells, docking is reduced in NRXN1␣ KO mouse islets. Analysis of granule distrubtion by EM revealed that, compared with WT, KO islets had 39% fewer granules in the 100 -200 nm distance category (Fig. 8). Consistent with the finding of greater insulin content (Fig. 4G) and granule number (Fig. 5C) after NRXN1 knockdown in INS-1E and NRXN1␣-transfected (Overexpressed) samples treated with low or high glucose concentration. For quantitation of relative NRXN abundance, shown in the lower two panels, intensity of bands was normalized to the GAPDH loading-control band. Data in the lower two panels are represented as mean Ϯ S.E. from seven experiments. Statistical significance was determined using a Student's t test. *, p Ͻ 0.05 indicates the difference compared with low glucose-treated controls.

Neurexin-1␣ Contributes to Insulin Granule Docking
cells, NRXN1␣ KO islets also had a 46% increase in insulin granule number per cell cross section compared with WT controls (Fig. 8C).

DISCUSSION
The present study establishes a role for NRXN1 in insulin secretion and specifically in insulin granule docking at the ␤-cell membrane. Consistent with our prior analysis of transcripts from INS-1 ␤ cells and human islets (19), with the more in-depth qPCR analysis reported here, and with our prior immunostaining of rat pancreas (19), we found that NRXN protein is present in INS-1E and human, rat, and mouse islet ␤ cells. NRXN transcripts in ␤ cells have a very different, more restricted expression profile than in the brain. In rat, human, and mouse islets, NRXN1␣ transcript expression predominates. NRXN1␣ message is also highly expressed in INS-1E cells. There are also substantial levels of NRXN2␤ transcripts in INS-1E cells and, to a lesser degree, in mouse and human islets.
As we hypothesized, the proteins that interact with NRXN1␣ in ␤ cells seem to mirror those in the neuronal presynaptic active zone. This is further evidence that ␤ cells and neurons employ highly conserved machineries to mediate regulated secretion. As in neurons, NRXN1␣ in ␤ cells binds to components of the submembrane exocytotic machinery. NRXN1␣ also associates with granuphilin, a granule-associated protein that plays a key role in secretory granule docking (26). This, together with the insulin secretion and EM results presented herein, indicate that NRXN is a component of the plasma membrane-associated protein complex that mediates insulin granule docking.
Granuphilin is associated with the insulin granule surface and is essential for mediating docking to the submembrane, syntaxin 1A/Munc18-containing docking site (26,48). Granuphilin-mediated docking is inhibitory to insulin secretion (26,46,49). Although this finding is at odds with studies suggesting that docking precedes or facilitates insulin exocytosis (52,54,55), it is in line with a recent study showing attenuation of insulin secretion by ␣-synuclein, another ␤-cell protein that promotes secretory granule association with the membrane (56). By providing another example of a constituent of the docking apparatus that inhibits insulin secretion, our results support the idea that docking serves to constrain insulin exocytosis, halting secretory granules en route to SNARE complex formation and subsequent insulin release. Insulin secreted by WT (left) and KO (right) islets is depicted as percent of total insulin content. B, glucose stimulation index for WT and KO islets was derived from the experiment shown in panel A. The stimulation index is the ratio of insulin secretion (% of total) at 16.7 mM to secretion at 2.75 mM glucose. C, islets were perifused in Krebs-Ringer solution containing 3 mM glucose and then switched at time 0 to 20 mM glucose. The insulin secretion rate (ISR) of WT (black triangles) and NRXN1␣ KO islets (white triangles) was measured every minute for the first 6 min, every 2 min until minute 10 and then every 5 min until minute 40. D, area under the curve (AUC) was calculated for the first 5 min to determine first-phase insulin secretion (black bars) and then from minute 10 to 40 to determine second-phase insulin secretion (white bars). The ISR at the peak of first-phase secretion (3 to 5 min) was on average 40% greater in KO islets (p ϭ 0.02). Data from static cultures are represented as mean Ϯ S.E. from three samples, and each experiment was repeated three times with similar results. Perifusion data are represented as mean Ϯ S.E. from six islet preparations of each type. Statistical significance was determined using a Student's t test. *, p Ͻ 0.05 indicates the comparison of secretion from WT and NRXN1␣ KO islets. FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9

JOURNAL OF BIOLOGICAL CHEMISTRY 6357
To clarify the role of NRXN1 in exocytosis, we used siRNA to deplete NRXN1 in INS-1E cells. Loss of NRXN1, as was previously observed with loss of granuphilin (48,49), increased glucose-stimulated insulin secretion (Fig. 4D). Also as in the case with granuphilin (47), NRXN1 knockdown did not impact basal insulin secretion, nor did it affect constitutive secretion, indicating that the knockdown did not cause a general impairment of ␤-cell secretory function. These findings were confirmed by studies of islets isolated from NRXN1␣ KO mice in which loss of NRXN1␣ also resulted in increased glucose-stimulated insulin secretion. Results from perifusion of KO and control mouse islets revealed a significant effect on second-phase secretion as well as evidence of increased first-phase secretion. These results resemble those observed with granuphilin KO mice, in which both first and second phase secretion were increased (57). It is possible that the continued expression of other NRXN isoforms (supplemental Figs. S1 and S2A) moderated the effect observed here of NRXN1␣ KO on first (and second) phase secretion.
Overexpression of NRXN1␣ had no effect on insulin secretion, suggesting that the maximal effect of NRXN1␣ is already achieved at endogenous levels, perhaps, for example, due to limited availability of key binding partners at supraphysiological expression levels. In any case, this lack of significant effect on insulin secretion would be expected after increasing expression of a protein, such as we now believe NRXN1␣ to be-normally expressed at levels not rate-limiting for insulin exocytosis. An example of this is glucose transporter 2, knockdown of which impairs glucose-stimulated insulin secretion (58) but overexpression of which has no effect (59).
The parallel between the effects of NRXN1␣ and granuphilin levels on insulin secretion is consistent with NRXN, like granuphilin, being essential to the docking mechanism. Interestingly, NRXN1 knockdown caused an increase in insulin content and transcript levels. Electron micrographs also show increased numbers of secretory granules in NRXN1 siRNA-treated cells and islets from NRXN1␣ KO mice. The increase in insulin levels was most likely not due to increased insulin secretion caused by NRXN knockdown: insulin content increased in INS-1E cells treated with NRXN1 siRNA even at non-stimulatory glucose levels at which NRXN1 silencing did not increase insulin secretion. The mechanism, then, whereby NRXNs influence ␤-cell insulin content remains to be uncovered. In NRXN1 siRNA-treated INS-1E cells and islets from

Neurexin-1␣ Contributes to Insulin Granule Docking
NRXN1␣ KO mice, the increases in insulin secretion cannot be explained by increased insulin content since insulin secretion in our static culture studies was normalized to total cellular insulin content and also the magnitudes of the increases in secretion resulting from loss of NRXN1␣ exceeded those of the increases in content.
It is thought that insulin granule docking occurs via interactions of granuphilin with syntaxin 1 beneath the ␤-cell plasma membrane (46). However, loss of syntaxin 1 had no effect on secretory granule docking (57), suggesting the presence of an alternative submembrane receptor for granuphilin. The findings presented here suggest that NRXN1␣ may be the membrane-associated protein that anchors the docking complex at a subplasmalemmal site. First, NRXN1␣ associates, either directly or indirectly, with granuphilin (Fig. 3). Second, NRXN1 knockdown in INS-E cells and KO in mouse islets impairs secretory granule docking. Third, given the extensive resemblance between regulated secretion in neurons and in the ␤ cell, the known importance of interactions involving NRXN in synaptic vesicle docking at the presynaptic membrane suggests a similar role for NRXN in ␤ cells (16,24,25). In neurons, the interaction between NRXN and the granuphilin-related protein rabphilin-3 is mediated by the intermediary protein CASK (24), a known NRXN binding partner (17,18,45). We have found that NRXN in ␤ cells also interacts with CASK. Overall, as would be predicted by extrapolation of findings in the synapse, our results suggest that secretory granule docking in ␤ cells is mediated by the interaction of granuphilin with NRXN1, either directly or indirectly (perhaps via CASK).
While it seems likely that NRXNs are essential for docking of secretory granule docking, granule docking was not completely abolished in INS-1E cells treated with NRXN1 siRNA and in islets from NRXN1␣ KO mice. Possible explanations for the persistence of docked granules in these studies include incomplete NRXN1 gene silencing in the siRNA-treated cells and functional redundancy with NRXN2␤ or other NRXN isoforms (supplemental Figs. S1 and S2A). Alternatively, the remaining granules that appear to be docked by morphological criteria may be stochastically located close to the plasma membrane without being molecularly docked, as has been proposed to account for the ϳ30% of granules that remain in the 100 -200 nm distance category in granuphilin-null mice (26).
NRXN1 is different from most established mediators of docking and exocytosis in that it is a transmembrane protein with a large extracellular domain that binds to multiple partners. Some of these binding partners, such as neuroligins and ␣-dystroglycan, are expressed on the ␤-cell surface and perhaps also by the islet vasculature (19, 60 -62). NRXN1 thus provides a direct link between the exocytotic machinery and the ␤-cell surface and could influence the localization of the exocytotic microdomains beneath the plasma membrane (63). It is also possible that trans interactions of NRXN with extracellular binding partners contribute to the functional architecture of the islet by juxtaposing ␤ cells with endothelial cells to promote cell ␤-cell development, secretory function (64) and polarized orientation (65) as well as maintaining cell-cell contacts important for signaling and regulated secretion (66). Because NRXNs help initiate and guide synaptic differentiation (67), it is also possible that NRXN1, in addition to its role in exocytosis, contributes to the development and functional maturation of the ␤-cell. We also found that glucose stimulation causes a reduction in NRXN protein. Because NRXN1 down-regulation increases glucose-stimulated insulin secretion, NRXN degradation may enhance the insulin secretory response to elevated glucose levels.
In summary, our data demonstrate that NRXN1 has a functional role in ␤ cells as a component of the machinery for regulated exocytosis and, more specifically, is essential for insulin granule docking. NRXN1␣ interacts with constituents of the submembrane exocytotic and docking protein machinery and with the secretory granule-associated protein granuphilin. Decreased expression of NRXN1 in ␤ cells increases glucosestimulated insulin secretion, most likely by decreasing secretory-granule docking. This increase in insulin secretion after NRXN knockdown provides further evidence that docking acts, as has been proposed by others, as a "temporal constraint" on insulin secretion (48). Under stimulatory conditions, NRXN1␣ protein levels are decreased, and this glucose-dependent regulation of NRXN could allow the ␤ cell an additional level of control over insulin secretion. Further work is needed to determine whether NRXN1␣ has a role in ␤-cell maturation, whether its extracellular interactions determine the docking sites of insulin granules, and whether it is a potential therapeutic target for modulating ␤-cell function in diabetes.