A Novel GLP1 Receptor Interacting Protein ATP6ap2 Regulates Insulin Secretion in Pancreatic Beta Cells*

Background: The transduction of the GLP1 receptor (GLP1R) requires interactions with accessory proteins. Results: ATP6ap2, also known as the renin receptor, was shown to interact with GLP1R and to regulate both GLP1 and glucose-stimulated insulin secretion. Conclusion: ATP6ap2 is a novel GLP1R interactor that modulates insulin secretion. Significance: Our study provides new insights into the fine-tuning GLP1R signaling in beta cells. GLP1 activates its receptor, GLP1R, to enhance insulin secretion. The activation and transduction of GLP1R requires complex interactions with a host of accessory proteins, most of which remain largely unknown. In this study, we used membrane-based split ubiquitin yeast two-hybrid assays to identify novel GLP1R interactors in both mouse and human islets. Among these, ATP6ap2 (ATPase H+-transporting lysosomal accessory protein 2) was identified in both mouse and human islet screens. ATP6ap2 was shown to be abundant in islets including both alpha and beta cells. When GLP1R and ATP6ap2 were co-expressed in beta cells, GLP1R was shown to directly interact with ATP6ap2, as assessed by co-immunoprecipitation. In INS-1 cells, overexpression of ATP6ap2 did not affect insulin secretion; however, siRNA knockdown decreased both glucose-stimulated and GLP1-induced insulin secretion. Decreases in GLP1-induced insulin secretion were accompanied by attenuated GLP1 stimulated cAMP accumulation. Because ATP6ap2 is a subunit required for V-ATPase assembly of insulin granules, it has been reported to be involved in granule acidification. In accordance with this, we observed impaired insulin granule acidification upon ATP6ap2 knockdown but paradoxically increased proinsulin secretion. Importantly, as a GLP1R interactor, ATP6ap2 was required for GLP1-induced Ca2+ influx, in part explaining decreased insulin secretion in ATP6ap2 knockdown cells. Taken together, our findings identify a group of proteins that interact with the GLP1R. We further show that one interactor, ATP6ap2, plays a novel dual role in beta cells, modulating both GLP1R signaling and insulin processing to affect insulin secretion.

GLP1 2 is an incretin hormone secreted from enteroendocrine L cells in the intestines. The ability of GLP1 to enhance insulin secretion upon stimulation by the uptake of glucose has been well documented (1)(2)(3)(4)(5). Furthermore, GLP1 has been shown to increase beta cell proliferation possibly through the TCF7l2/Wnt pathway (6 -8). In addition to its effects in pancreatic beta cells, GLP1 also has diverse functions in a variety of extrapancreatic tissues. In the heart, the cardiac effects of GLP1 analogs have led to the amelioration of myocardial ischemia and to the restriction of infarct size (9), and GLP1 infusion could improve heart function (10). GLP1 in plasma was associated with blood pressure levels in a human population study (11). Further, GLP1 or its analogs were shown to lower blood pressure in rodents and human subjects (12)(13)(14). In the brain, GLP1 analogs induced the proliferation of neuronal progenitor cells, implicating a potential involvement in the repair of neurons (15)(16)(17).
The physiological and pharmacological effects of GLP1 are mediated by the GLP1 receptor (GLP1R), a member of the B class G protein-coupled receptor (GPCR) family (18). GLP1R is widely expressed in pancreatic islets, as well as in the brain, heart, kidney, and gastrointestinal tract (19 -21). Specifically in the pancreas, GLP1R expression was confirmed in beta and delta cells; however, it was 10-fold lower in delta cells when quantified by quantitative real time PCR (qPCR) (21). In the alpha cell population, very low expression of GLP1R was detected (21). Like other B class GPCRs, GLP1R signals through the G s protein complex and activates adenylyl cyclase, which converts ATP into cAMP (22,23). The increased intracellular accumulation of cAMP triggers both the PKA and Epac2 pathways that are the common downstream pathways responsible for a number of GLP1-induced intracellular actions (24 -26). In addition to classical PKA and Epac2 signaling, GLP1R is also found to activate PI3K signaling by transactivating the EGF receptor (27,28).
As is the case with many GPCRs, a wide number of receptor activities and modes of signal transduction have been described for the GLP1R. This phenomenon may be explained in part by an interaction of the receptor with a large number of G protein-dependent and independent accessory proteins (interactors). Receptor accessory proteins are reported to regulate GPCRs to target subcellular trafficking and intracellular signaling (29). For instance, KCTD isoforms 8, 12, 12b, and 16 are accessory proteins of the GABAB receptor and are indispensable for its function. These isoforms associate tightly with the GABAB2 receptor C terminus to increase agonist potency and markedly alter G protein signaling, thus accelerating the onset of signaling and promoting desensitization of the receptor in a subtypespecific manner (30). However, very little is known thus far on GLP1R accessory proteins. It has been reported that the GLP1R interactor scaffolding protein ␤-arrestin-1, a protein involved in GPCR agonist-induced desensitization and endocytosis, is required to stimulate cAMP production and insulin secretion in INS-1 beta cell lines upon physical association with the GLP1R (31). Furthermore, the GLP1 interactor caveolin-1 directly interacts with GLP1R, which may in part be directing the trafficking of GLP1R to lipid rafts (32). In another study, ␤-arrestin-1 was shown to associate with GLP1R and c-Src as a complex, which is required for the proliferative action of GLP1 (33). According to our model, these reports raise the possibility that GLP1R could be coupled to many accessory/interacting proteins to form multimeric protein interactome complexes capable of transducing context-specific downstream signaling pathways that lead to an increasing number of cellular actions.
Recently, using a novel membrane-based split-ubiquitin yeast two-hybrid system (MYTH), we discovered a series of GLP1R interacting proteins in human fetal brain that were shown to attenuate GLP1R activity (34). In this current study, we used MYTH to reveal a GLP1R interactome in both mouse and human islets: two tissues where GLP1R agonists have profound and clinically relevant effects. This MYTH screen identified a host of novel putative GLP1R interacting proteins with major differences seen between human and mouse islet libraries. However, one interactor, ATP6ap2 (ATPase H ϩ -transporting lysosomal accessory protein 2), identified in both mouse and human islet screens was shown to be abundantly expressed in pancreatic islets. Further work demonstrated that ATP6ap2 regulated insulin secretion from pancreatic beta cells in both glucose and GLP1R dependent modalities.
Isolation and Disperse of Mouse Islet-Mouse islets were isolated and dispersed from male CD-1 mice (ϳ2 months of age) as described previously (35,36). The intact islets were cultured in RPMI 1640 medium containing 11.1 mM glucose supplemented with 10% fetal bovine serum, 10 mM HEPES, 1% L-glutamate, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in 5% CO 2 .
The Split Ubiquitin Membrane Yeast Two-hybrid System-The MYTH system ( Fig. 1) methodology was described in detail in previous studies (34,37). Briefly, MYTH is based on the "split ubiquitin system" (38 -41) in which ubiquitin can be split into N-terminal (Nub) and C-terminal (Cub) halves. The reconstitution of the two halves forms pseudoubiquitin, which is recognized by ubiquitin specific proteases leading to proteasomal degradation. In the MYTH system, the receptor (in this case GLP1R) is fused with Cub followed by a transcription factor (TF) to form the "bait," whereas the interactor protein is fused with NubG (mutational Nub Ile 13 3 Gly to reduce the affinity FIGURE 1. Membrane-based split ubiquitin yeast two-hybrid. Left panel, membrane-bound ubiquitin protein was split into two halves: C terminus (Cub) and N terminus (Nub, NubG is point mutation of Nub to avoid self-activation). Cub is associated with a TF that was fused to the bait GLP1R as GLP1R-Cub, and NubG was fused to the prey interactors as interactor-NubG. Right panel, if the bait interacts with the prey, the resulting proximity of the ubiquitin halves induced by the interaction will enable the reconstitution of Cub and NubG to form a functional pseudoubiquitin protein. Reconstitution recruits ubiquitin-specific proteases that cleave TF downstream of Cub, allowing the TF to translocate into the nucleus to initiate the transcription of reporter genes, which serve as readout of MYTH. As a result, the MYTH system does not rely on protein expression within the nucleus as does the traditional yeast two-hybrid system and can be used to study membrane-bound proteins such as GLP1R and its interactors.
of Nub for Cub and avoid self-activation) to form the "prey." When NubG interactors (prey) are transformed into yeast expressing GSRs-Cub-TF (bait), the interaction between bait and prey brings NubG and Cub to proximity with one another to form a functional pseudoubiquitin, resulting in the release of the TF upon recognition and cleavage of the pseudoubiquitin by ubiquitin specific proteases. The TF further translocates into the nucleus and induces the expression of reporter genes that serves as a readout of protein-protein interaction.
Construction of Human and Mouse Islet cDNA Libraries Compatible with MYTH-Mouse islets were isolated from CD-1 mice (ϳ2 months of age) as described previously (35,36,42). Human islets from review board-approved healthy donors were provided by the Clinical Islet Laboratory (University of Alberta, Edmonton, Canada). The total RNA of human and mouse islets was prepared using the RNeasy mini kit (Qiagen) and were used to produce a cDNA library that was compatible to the MYTH system by Dualsystems Biotech, Inc. (Schlieren, Switzerland). pPR3-N (Dualsystems Biotech, Inc.) was used as the prey vector for generating the cDNA library.
MYTH Analysis of GLP1R in Human and Mouse Islet cDNA Library-The MYTH analysis was performed by Dualsystems Biotech Inc. The technology and the bait vector pCCW-ste-hGLP1R-cub were described in previous studies (34,37). Briefly, the bait and prey vector were co-transformed into Saccharomyces cerevisiae host THY.AP4 strain, and colony selection was performed on yeast minimal media/synthetic defined agar plates deficient of tryptophan, leucine, histidine, and adenine negative (ϪTrp/ϪLeu/ϪHis/ϪAde). All positively selected colonies were inoculated in yeast, and the plasmids harboring the interactor sequence were purified. The purified plasmids were amplified in Escherichia coli strain XL-10 Gold. All the plasmids were validated by sequencing. The candidates were compared against pre-existing MYTH screening databases from Dualsystem Biotech Inc., and only unique candidates identified in our current screen were chosen as putative interactors of GLP1R for further study.
cDNA Plasmids, siRNA, and qPCR-The cDNA plasmid pcDNA-ATP6ap2 was purchased from Origene (Rockville, MD) and the Midi-Prep Kit (Qiagen) was used for plasmid purification. Short interfering RNA (SMARTpool siRNA) or scrambled siRNA (control) were purchased from Dharmacon, Thermo Scientific (Waltham, MA). Total RNA from cells was prepared using the RNeasy mini kit (Qiagen) according to the manufacturer's instructions. Purified RNA was converted to cDNA using a Moloney murine leukemia virus reverse transcriptase cDNA kit (Sigma), and real time PCR was performed    using an ViiA7 real time PCR system (Applied Biosystems, Foster City, CA) according to the same protocol described in previous studies (43). A standard curve was generated using mouse genomic DNA for quantification purposes. The measurements of gene expression were normalized to ␤-actin transcripts within the same sample. The sequences of the primers are shown in Table 1.
Co-immunoprecipitation and Immunoblotting-INS-1 cells co-transfected with GLP1R-His-V5-tagged and FLAG-tagged interactor were washed with ice-cold PBS containing 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.76 mM KH 2 PO 4 to a pH of 7.4. Cells were harvested in lysis buffer containing 10% glycerol, 50 mM HEPES, 150 mM NaCl, 2 mM EDTA, 0.25% n-dodecyl-␤-D-maltoside with complete protease inhibitor mixture (Hoffmann-La Roche Limited, Ltd., Mississauga, Canada). M2 Anti-FLAG affinity gel (Sigma-Aldrich) was used to pull down FLAG-tagged interactor proteins. Briefly, the cell extract (supernatant) was incubated with anti-FLAG-aga-rose that was equilibrated with wash buffer (0.1% digitonin, 5 mM imidazole with protease inhibitor mixture) for 2 h at 4°C. The anti-FLAG affinity beads were washed three times with wash buffer and eluted in 2ϫ SDS loading buffer. The precipitated proteins from each sample were loaded and separated on a 10% polyacrylamide gel and transferred to PVDFplus TM membrane for immunoblotting. Anti-V5 (Invitrogen, 1:2500 dilution), anti-FLAG primary antibodies (Sigma-Aldrich, 1:2000 dilution), and HRP-conjugated mouse secondary antibody were used, and the fluorescence signal was detected by Amersham Biosciences enhanced chemiluminescence (GE Healthcare Lifesciences) with images acquired by the Kodak Image Station 4000 Pro (Carestream Health Inc., Rochester, NY).
Immunohistochemistry-Tissues and cells were fixed in 10% neutral buffered formalin, dehydrated in 70% ethanol, and embedded in paraffin. Paraffinized samples were sliced (5 m) and adhered to glass slides, rehydrated, and blocked with 3% H 2 O 2 for 30 min. Following PBS washing, sections were incubated in nonimmune serum-free protein block solution (Dako Canada Inc., Burlington, Canada) for 30 min. Sections were blotted to remove excess blocking solution prior to overnight application of primary anti-ATP6ap2 antibody (Sigma-Aldrich, 1:500 dilution) and anti-insulin (Invitrogen, 1:100 dilution) at 4°C. Images of each section were acquired using Aperio Imagescope version 11.0.2.725 (Aperio Technologies, Vista, CA).

ATP6ap2 Regulates Insulin Secretion
and D2-labeled cAMP. Fluorescence signals in both insulin and cAMP assays were measured using the PHERAstar Plus microplate reader (BMG LABTECH, Guelph, Canada).
Transmission Electron Microscopy-Cells were fixed, and images were acquired as previously described (45). Briefly, the samples were observed under a Philips CM100 electron microscope operating at 75 kV. Images were recorded digitally using Kodak 1.6 Megaplus camera system operated using AMT software (Advanced Microscopy Techniques Corporation). Granule numbers were manually quantified using ImageJ software (45).
High Content Imaging-Images were acquired and analyzed on a Thermo Fisher Cellomics ArrayScan VTI HCS reader using iDEV TM software. The filter settings for each dye were excitation/emission: 577/590 nm for LysoTracker Red DND-99, excitation/emission: 494/516 nm for Fluo4AM, and excita-tion/emission: 350/461 nm for Hoechst 33342 (Molecular Probes, Life Technologies). Each dye was loaded into live INS-1 cells or dispersed mouse islet cells according to the manufacturer's recommendation.
Statistics-Paired t tests were performed to determine statistical significance. p values less than 0.05 were considered statistically significant.

Generation and Analysis of Human and Mouse Islet cDNA
Libraries-To better understand the mechanisms through which GLP1R fine tunes the regulation of insulin secretion, we generated cDNA libraries from both human and mouse islets that were equipped with the physiological machinery necessary for insulin secretion. GLP1R was shown to be abundantly expressed in islet beta cells, less in delta cells, and very low in alpha cells (21). As such, we reasoned that isolated human and mouse islets would serve as reasonable models to study the GLP1R interactome in the setting of the pancreatic beta cell. Using purified RNA from isolated human and mouse islets, we generated human and mouse islet cDNA libraries with the complexity and titer required for MYTH screening. The human islet cDNA library generated contained ϳ5.6 ϫ 10 6 independent clones, ranging in sizes from 0.5 to 5.0 kb with 100% of all vectors containing cDNA inserts. The mouse islet library contained 6.9 ϫ 10 6 independent clones with a size range equal to that of the human islet library generated ( Table 2).

MYTH Analysis of GLP1R in Human and Mouse Islets (Functional Involvement)-
The structure of the bait vector overexpressing human GLP1R and its ability to respond to GLP1 was previously described by Huang et al. (34). By using the bait GLP1R vector, we observed 43 positive interactor proteins from the human islet library and 37 such interactor proteins from the mouse islet library. By eliminating highly abundant proteins and common nonspecific MYTH screen interactors (those interactors that appeared in over 50% and 20 -50% of all performed MYTH screens done by Dualsystems Biotech Inc.), we obtained 31 and 29 unique interacting proteins from the human (Table 3) and the mouse islet libraries (Table 4), respectively. Apart from G␣ s , which is known to be linked to GLP1R function, these putative interactors identified in the MYTH screen have not previously been described with GLP1R. Collectively they represent many known functional groups such as intracellular transport, metabolism and ion transport, or signal transduction (Fig. 2). Some interactors were suggested to be relevant to pancreatic beta cell function, such as zinc transporters (SLC39A7 and SLC39A13), fructose transporters (SLC2A5), and insulin exocytotic SNARE proteins (VAMP3) etc. Among the putative interactors, ATP6ap2 and SELK were identified as two proteins present in both the human and mouse islet library screens (Fig. 2).
Expression of Selected Interactors in Pancreatic Beta Cells-Because ATP6ap2 and SELK were identified from both libraries, we examined their expression using qPCR in three distinct sources of pancreatic beta cells (MIN6, INS1, and isolated mouse islets). Since the GLP1R is primarily expressed at the plasma membrane, we also included 11  OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41

JOURNAL OF BIOLOGICAL CHEMISTRY 25051
membrane-bound putative interactors in addition to ATP6ap2 and SELK found in our screens. Among the 13 interactors examined, ATP6ap2 was consistently the most abundant (Fig. 3) in all three cell types, whereas SELK was not. Some other membrane-bound interactors such as SYNGR4, APH1B, and GNAS showed only very low abundance compared with our control, Kcnj11, the subunit of the K ATP channel required for glucose-stimulated insulin secretion (Fig. 3). The transcript expression profile pattern of these interactors was comparable among all three cell types, whereas only SLC39A7 and Leprotl1 had relatively higher expression levels compared with other interactors in mouse islets. Taken together, based on the expression pattern, ATP6ap2 identified from both islet MYTH screening was most highly expressed across all three cell types and was therefore chosen for further functional analysis.
ATP6ap2 has been shown to be expressed in several tissue including brain, heart, kidney, liver, pancreas, and adipose tissues (43,44) with the highest levels reported in MIN6 cells (BioGPS (46, 47)). To localize ATP6ap2 expression within the pancreas, both immunohistochemistry and immunofluores-cence staining were performed on mouse pancreatic slices and dispersed human islet cells, respectively. In mouse pancreata, ATP6ap2 was expressed in insulin immunopositive cells but not in acinar tissue (Fig. 4A). Further, in dispersed human islet cells, ATP6ap2 was shown to be expressed in both alpha and beta cells (Fig. 4B).
Interestingly, we found that the intensity of ATP6ap2 staining appeared weaker in islets from diabetic donors (Fig.  5A). The percentage of islet cells with strong fluorescence (Ͼ600 relative fluorescence intensity) decreased remarkably in diabetic islet cells compared with normal islet cells, whereas the percentage of those with weak fluorescence increased (Fig. 5A). These observations suggested decreased ATP6ap2 expression in diabetic islets. Further, we examined whether or not there was decreased expression in both alpha and beta cells. After co-staining with insulin or glucagon, we showed that ATP6ap2 expression was decreased primarily in beta cells (Fig. 5B). Importantly, correlating with ATP6ap2, islets from diabetic donors had impaired glucose-stimulated insulin secretion (GSIS) compared with controls (Fig. 5C). Effect of Overexpressing ATP6ap2 on Insulin Secretion and cAMP Accumulation in INS-1 Cells-To further validate the interaction of GLP1R with ATP6ap2 employing the INS-1 cell line, we co-expressed epitope tagged GLP1R-V5 and ATP6ap2-FLAG. ATP6ap2 overexpression was detected using both anti-FLAG and anti-ATP6ap2 (Fig. 6A). GLP1R was detected after affinity purification of ATP6ap2-FLAG but not in GLP1R-expressing cell lysates alone, suggesting the interaction between two proteins in pancreatic beta cells (Fig. 6B) validating the MYTH assay results.
ATP6ap2 has been found to act as an adapter protein of the V-ATPase receptor complex (48) that maintains the acidic environment within vesicles required for the maturation and priming of insulin protein in pancreatic beta cells (49,50). We further examined the effect of overexpressing ATP6ap2 in INS-1 cells. Interestingly, ATP6ap2 overexpresssion did not have any significant effect on insulin secretion under basal glucose conditions nor upon glucose or GLP1-induced insulin secretion (Fig. 6C) under the conditions studied. ATP6ap2 overexpression had no effect on GLP1-induced cAMP forma-  OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 tion (Fig. 6D), which is a key second messenger molecule in GLP1 signaling.

ATP6ap2 Regulates Insulin Secretion
Effect of Knocking Down ATP6ap2 on Insulin Secretion and cAMP in INS-1 Cells-Because we did not observe significant effects in cells overexpressing ATP6ap2, to further elucidate the role of ATP6ap2 within pancreatic beta cells, we knocked down ATP6ap2 by siRNA (Fig. 7A). When ATP6ap2 was effectively knocked down (Fig. 7A), insulin secretion was decreased significantly under high glucose conditions (11.5 mM; Fig. 7B), as was GLP1-induced insulin secretion (Fig.  7B). This reduction in insulin secretion was not associated with cell death (data not shown) nor decreases in total insulin content (Fig. 7C). Furthermore, we showed that this attenuating effect only occurred at stimulatory glucose concentrations (11.5, 16.7, and 20 mM), not at low or moderate glucose levels (2.8 or 5.6 mM) (Fig. 7D). The inhibitory effect of ATP6ap2 knockdown was seen across a range of GLP1 concentrations (Fig. 7E).
GLP1R signaling effects are primarily mediated by the second messenger cAMP. In line with this, we found that downregulation of ATP6ap2 led to a decrease in GLP1-stimulated cAMP accumulation in INS-1 cells (Fig. 7F). However, this effect was only observed at 0.1 and 10 nM GLP1 concentrations but not at a 1 pM GLP1 concentration (Fig. 7F), suggesting that the decrease in cAMP accumulation might be concentrationdependent. To confirm that the observation was specific, we also treated the siATP6ap2 transfected cells with forskolin (direct stimulation on adenylyl cyclase) and GIP (incretin acting on beta cells and sharing similar structure and signaling pathways with GLP1R). We did not observe any difference in cAMP accumulation between control and ATP6ap2 knockdown cells (Fig. 7G). This suggested that down-regulation of ATP6ap2 specifically decreased GLP1R stimulated cAMP accumulation. The fact that we did not observe an ATP6ap2 effect in the overexpression model is likely due to the high abundance of native ATP6ap2 in INS-1 cells. Conversely, knocking down endogenous ATP6ap2 would likely exhibit more profound effects on the cells.
Effect of Knocking Down ATP6ap2 on Insulin Granule Morphology and Acidification-Because decreased insulin secretion and impaired insulin processing were observed in ATP6ap2 knockdown INS-1 cells, we next examined whether there were defects in insulin granule morphology using electron microscopy. We did not observe any change in granule morphology (Fig. 8A) or the number of insulin granules in siRNA treated INS1 cells (Fig. 8B), but the percentage of gray and empty granules was increased by more than 2-fold in ATP6ap2 knockdown cells (Fig. 8C). These results indicated that decreased insulin secretion was not due to the reduction of insulin granules per se but was possibly due to impaired insulin biosynthesis.
To further explore a processing defect, we detected an increased ratio of proinsulin versus insulin peptide in ATP6ap2 knockdown cells under all conditions including basal glucose, stimulated glucose, and GLP1 stimulation (Fig. 9A). This also suggested a possible impairment of insulin processing in ATP6ap2 knockdown cells. To address this, the expression of the prohormone convertases involved in insulin processing  including PC1/3, PC2, and CPE (51-55) were examined. There was no difference in the transcript expression of these three key enzymes (Fig. 9B), suggesting that other factors are contributing to impaired insulin processing. It was reported that acidification within insulin granules was critical for their maturation and priming (56). To examine whether ATP6ap2 is involved in the acidification of the insulin granules, we loaded the cells with acidotropic LysoTracker, a dye that accumulates specifically in acidic organelles. In pancreatic beta cells, the majority of cellular acidic structures are indeed insulin granules; therefore Lyso-Tracker staining can be used to evaluate the acidity of the granule given the fact that insulin granule number did not change in ATP6ap2 knockdown cells compared with the control (Fig. 8B). Consistent with previous studies (57), the cells treated with V-ATPase inhibitor bafilomycin were barely stained with Lyso-Tracker, suggesting that granule acidification was blocked (Fig.  9C). Similarly, LysoTracker staining in ATP6ap2 knockdown cells was shown to be decreased significantly when compared with control cells, suggesting that insulin granule acidification was impaired (Fig. 9C). Taken together, these results suggest that decreased insulin processing and secretion upon ATP6ap2 knockdown could be due to impairment of acidification within individual insulin granules.
ATP6ap2 Effects on GLP1-stimulated Ca 2ϩ Influx-GLP1 R activation increases intracellular Ca 2ϩ mediated by PKA and Epac (cAMP) (26,58). We observed a robust increase in intracellular Ca 2ϩ in control dispersed mouse islets upon GLP1 stimulation (Fig. 10A, upper two panels). However, when we knock down ATP6ap2 in dispersed islet cells (Fig. 10B), the GLP1-induced increase in Ca 2ϩ was completely diminished, suggesting that ATP6ap2 was required for GLP1-induced Ca 2ϩ influx (Fig. 10A, lower two panels).

Discussion
The GLP1R and its associated signaling pathways have gained much attention over the past several years because they are targets for current and future medications to treat type 2 diabetes. In this study, we have used a novel MYTH assay (Fig.  1) to screen mouse and human islet cDNA libraries, enabling us to discover over 50 novel putative interactors of GLP1R. Each of these interactors has the potential to regulate pancreatic beta cell function through GLP1R. The interactor, ATP6ap2, was identified in both mouse and human islet screens and pursued in further studies because of its consistently high expression across beta cell lines and primary islets. Our results showed that ATP6ap2 plays an important dual role; first, it regulates GLP1R signaling through cAMP, and second, it likely facilitates the processing of insulin through granule acidification. We realize that the latter effect appears GLP-1R-independent and may be a more global permissive effect of ATP6ap2 on insulin secre- tion. Future studies will further delineate the mechanistic link between GLP1R signaling and more general permissive beta cell effects.
In this study, we chose to identify potential interactors of GLP1R from MYTH screens on mouse and human islet cDNA libraries. Previous studies conducted by us have focused on a human fetal brain cDNA library, using a similar approach. However, we reasoned that pancreatic islet tissue is a more appropriate model from which to screen for interactors of GLP1R with the intension of understanding GLP1R signaling in the beta cell specifically. The pancreatic islet consists of a composition of cells including beta, alpha, and delta cells; however, the expression pattern of GLP1R predominantly in beta cells has been widely accepted (19 -21), much less in alpha or delta cells (21). However, it is possible that other glucagon receptor subfamily or GPCR interactors that normally are not found in GLP1R-expressing cells may in islet tissue interact with GLP1R, requiring rigorous testing for specificity. Notwithstanding this possibility, the pancreatic islet appears to be a good model tissue in which to study the GLP1R interactome in a beta cell specific setting.
Our study was able to identify a series of novel interactors involved in a variety of intracellular functions (Fig. 2), supporting the dynamic role that GLP1R plays in the regulation of pancreatic beta cell function. Of the common interactors identified from both mouse and human islet screens, the interactor ATP6ap2 was expressed at abundant levels in MIN6, INS-1, and mouse islets. Furthermore, by immunohistochemistry in mouse pancreatic sections, we showed that ATP6ap2 was expressed in endocrine islets but not acinar cells (Fig. 4A). Further analysis on human dispersed islets revealed ATP6ap2 expression in both pancreatic beta and alpha cells (Fig. 4B). ATP6ap2 is known as an accessory protein of the V-ATPase (59) composed of a transmembrane proton-translocation domain (V 0 ) and extramembrane pump domain (V 1 sectors) (60). Although there is no direct evidence demonstrating the effect of ATP6ap2 on insulin secretion in pancreatic beta cells, previous studies have demonstrated that the ATP6V 0 B subunit of V-ATPase increased 2.38-fold in human diabetic versus normal islets by microarray analysis (61). Also, the islet tropic a3 isoform (membrane-intrinsic subunit) of V-ATPase was reported to regulate insulin secretion from pancreatic beta cells (12). We observed reduced expression of ATP6ap2 in human diabetic versus normal islets, and the reduction was not in alpha but beta cells (Fig. 5). The reduced expression of ATP6ap2 was accompanied by an impaired GSIS in the human diabetic islets. In line with these observations, we showed that the down-regulation of ATP6ap2 resulted in decreased GSIS and GLP1-induced insulin secretion, suggesting a regulatory role of ATP6ap2 in insulin secretion from pancreatic beta cells. cAMP-PKA is an important pathway for GLP1-induced insulin secretion, and V-ATPase is required for the full activation of PKA in response to glucose stimulation (14). In our studies, cAMP accumulation was significantly decreased at GLP1 concentrations of 0.1 and 10 nM but not at 1 pM, and this effect was not observed in forskolin-or GIP-treated cells (Fig. 7G), suggesting that the cAMP-PKA pathway might be involved in GLP1-induced insulin secretion. The downstream molecules of cAMP are PKA and Epac. Because GLP1R activation increases intracellular Ca 2ϩ via PKA and Epac (26,58), one would expect ATP6ap2 knockdown could affect Ca 2ϩ influx associated with GLP1. Indeed, we found that the GLP1-induced increase in Ca 2ϩ was abolished in ATP6ap2 knockdown cells, suggesting a requirement of ATP6ap2 for GLP1R Ca 2ϩ signaling. This could also in part explain the decreased insulin secretion in cells with ATP6ap2 knocked down. ATP6ap2 has been shown to be involved in the maintenance of acidity within secretory vesicles (62). Previously, it was demonstrated that whole body knock-out of ATP6ap2 was lethal in mice (63); whereas, tissue specific knock-out studies of ATP6ap2 in cardiomyocytes or in podocytes resulted in detrimental defects after birth including heart failure or renal failure, respectively (64,65). These studies confirmed the requirement of ATP6ap2 association with V-ATPase, as well as the functional role of ATP6ap2 in maintaining the acidity of microenvironments within intracellular vesicles. Previous studies have shown the involvement of acidic secretory vesicles in insulin processing and maturation (66). Specifically, proprotein convertases 3 (PC1/3) and 2 (PC2) responsible for insulin processing from proinsulin are strictly pH-dependent (51,67). In line with this, the gene expression of insulin-processing enzymes PC1/3, PC2, and CPE were not changed in ATP6ap2 knockdown cells. However, granule acidification was impaired, causing increased pH, which might inhibit the activity of the insulin-processing enzymes. Furthermore, an increased proinsulin versus insulin ratio was detected, suggesting impairment in insulin processing and maturation within secretory vesicles from ATP6ap2 knockdown cells. To support this, we observed reduced expression of ATP6ap2 in islets from type 2 diabetic donors, whose GSIS was largely decreased. Our data provided a link between ATP6ap2 and diabetes where loss of ATP6ap2 impaired insulin processing via increasing granule pH.
In summary, our study provided a novel insight into GLP1R signaling and was the first to identify GLP1R interactors in pancreatic islets. Our data also suggested that these interacting proteins of GLP1R could be involved in the regulation of insulin secretion and GLP1R signaling in pancreatic beta cells. Because GLP1R has been used as potent drug target in the treatment of diabetes, our findings could contribute to the development of novel effective therapeutic strategies for this disease.