Optimized Proteomic Analysis of Insulin Granules From MIN6 Cells Identifies Scamp3, a Novel Regulator of Insulin Secretion and Content

Pancreatic β-cells play a crucial role in maintaining blood glucose levels, and their dysfunction is a hallmark of type 2 diabetes (T2D). Located within the islets of Langerhans, these cells release insulin, which stimulates glucose uptake by adipose tissue, liver, and muscle (1). In T2D, β-cells exhibit a reduction in glucose-stimulated insulin secretion (GSIS) and dysregulated insulin biogenesis and trafficking (2,3).

In β-cells, insulin is stored within the insulin secretory granules (ISGs), which are ∼300 nm in diameter and contain a dense core (4). In response to nutrient stimuli, ISGs are triggered for exocytosis and transported to the plasma membrane, where they fuse and release insulin.

ISG biogenesis begins at the trans-Golgi network, where proinsulin is sorted alongside prohormone convertases, cosecreted hormones, and carrier proteins (4) into immature ISGs (iISGs). Within iISGs, prohormones are cleaved into mature forms (5), granule lumen become acidified, and undesired proteins are removed (6). Mature insulin then crystallizes with zinc cations to form insulin hexamers, which constitute the dense core of the ISGs (7).

ISGs are dynamic structures, and numerous proteins residing within them play crucial roles in insulin biogenesis and secretion. For example, deficiencies in chromogranin-B lead to impaired proinsulin processing and reduction in GSIS in vivo (8) and in vitro (9). Knockout models of protein interacting with C kinase-1 and islet cell autoantigen of 69 kDa also resulted in impaired insulin secretion and defective glucose metabolism (10). As such, dysfunction of these ISG resident proteins contribute to metabolic disease and T2D. Thus, interrogating the ISG and discovering novel ISG-resident proteins will enhance our understanding of insulin granule biogenesis, trafficking, and glucose-mediated secretion of ISGs.

Four previous studies attempted to capture snapshot proteomes of ISG within β-cells (11–14). However, these studies assembled five proteins in consensus: insulin-1 (Ins1), insulin-2 (Ins2), chromogranin-A, prohormone convertase 2, and carboxypeptidase E (CPE). Given this, it is difficult to interpret the collective proteomic data, due to the co-enrichment of contaminating proteins or fragments of subcellular organelles such as the Golgi network, lysosomes, and most commonly, mitochondria (11–14). In this study, we address this caveat by using an optimized protocol to enrich for ISGs and a dual, unbiased, and targeted proteomics approach to minimize contamination from other subcellular compartments.

Previous studies on ISG isolation methods and proteomics analyses have focused solely on rat insulinoma INS-1 or INS1-E cell lines (11–14). Here, we modify an ISG enrichment method described by Chen et al. (15) to enrich ISGs from the mouse insulinoma β-cell line MIN6. MIN6 cells are arguably a better model for human metabolism because they possess a larger pool of ISGs, respond similarly to human islets in response to lipotoxicity (16), and can form pseudo-islets in culture (17). The modified three-step purification protocol involves sequential Optiprep and Percoll gradient enrichment, followed by sucrose purification of mature ISGs (mISGs). Liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis of fractionation samples identified 8021 unique peptides. Using protein correlation profiling (PCP) analysis, we identify 432 potential ISG proteins. Parallel analysis using proteomic-specific software allowed us to categorize proteins into defined subcellular organelles, assigning 211 proteins to the ISG with high confidence. To validate ISG targets, we conducted insulin colocalization analysis in immuno-stained MIN6 cells and human islets, confirming the ISG localization of Scamp3, synaptophysin, synaptotagmin-13, and syntaxin-7. Finally, siRNA-mediated Scamp3 knockdown (KD) in INS-1 cells elicited a defect in insulin content and GSIS, implicating the novel ISG protein in healthy β-cell insulin regulation for the first time, to our knowledge.

Human pancreas tissue was obtained from deceased organ donors as part of the St. Vincent’s Institute islet isolation program, through Donatelife, to obtain research consent and adhere to strict ethical guidelines developed by the National Health and Medical Research Council, Australia (reference no. E76; ISBN: 9781925129595). Donor information can be found in Supplementary Table 1. Tissue from male and female donors was used, and sex was not considered a factor in statistical analyses of the data.

MIN6 cells (derived from male mice) were purchased from AddexBio and cultured in standard culture medium, as described previously (18). INS-1 cells (derived from male rats) also were cultured in standard cultured medium, as previously described (18). Culture medium was changed every 2–3 days and cells were passaged every 4–5 days. MIN6 cells used were between passages 15 and 26. INS-1 cell used were between passages 19 and 33.

We optimized a two-step subcellular fractionation protocol described by Chen et al. (15). MIN6 cells were grown on 15-cm-diameter petri dishes to 80–90% confluence. Plates were washed with PBS, then cells were scraped into 2 mL of ice-cold PBS and centrifuged at 300g for 3 min. The pellet was resuspended in 1 mL of buffer A (0.3 M sucrose, 1 mmol/L EDTA, 1 mmol/L magnesium sulfate, 10 mmol/L 2-(N-morpholino)ethanesulfonic acid–potassium hydroxide (MES-KOH), pH 6.5) containing EDTA-free protease inhibitors (Roche), then passed through a 21G needle and a 25G gauge needle (10 times each). Homogenates were centrifuged (1,000g, 5 min), and the supernatant was loaded atop a discontinuous Optiprep gradient of five 3-mL layers: (from top): 8.8%, 13.2%, 17.6%, 23.4%, and 30%, diluted in buffer B (2 mmol/L EGTA, 20 mmol/L MES-KOH, pH 6.5). Samples were centrifuged (100,000g, 80 min) in a P50AT2 fixed-angle rotor in a Hitachi 100NX ultracentrifuge set at half acceleration and half deceleration. We removed 1-mL fractions from the top to obtain a total of 16 fractions. Percoll was used in the second step to further enrich and purify ISGs. mISG-enriched fractions 11 and 12 were loaded in 10 mL of 27% Percoll solution, diluted in buffer A in a 16-mL polypropylene tube, then centrifuged (35,000g, 60 min) in an A22-24 × 16 (Thermo Fisher) fixed-angle rotor in a Sorvall Lynx 4000 centrifuge (Thermo Fisher). Again, 1-mL fractions were removed from the top to bottom to obtain 12 fractions. In the third step, 12 1-mL fractions, after Percoll enrichment, were suspended in 9 mL each of sucrose wash buffer (0.3 mol/L sucrose, 5 mmol/L HEPES, 1 mmol/L EGTA, pH 7) and centrifuged (23,700g, 15 min) to pellet ISGs and remove Percoll from solution. Supernatants were discarded and the pellet was resuspended in 4% sodium deoxycholate in water and boiled (100°C, 10 min) for mass spectrometry (MS) analysis.

LC-MS/MS samples were prepared as previously described (19). Peptides were reconstituted with 5% formic acid in MS-grade water, sealed and stored at 4°C until LC-MS/MS acquisition. Using a Thermo Fisher RSLCnano ultrahigh-performance liquid chromatograph, peptides in 5% (v/v) formic acid (3 µL injection volume) were injected onto a 50 cm × 75 µm C18Aq (1.9 µm; Dr. Maisch) fused analytical column with an ∼10 µm pulled tip, coupled online to a nanospray electrospray ionization source. Peptides were resolved over gradient from 5 to 40% acetonitrile for 70 min, with a flow rate of 300 nL/min (capillary flow). Electrospray ionization was done at 2.3 kV. Tandem MS analysis was carried out on a Q-Exactive HFX mass spectrometer (Thermo Fisher) using data-independent acquisition (DIA). DIA was performed as previously described using variable isolation widths for different charge-to-mass ratio ranges (20). Stepped normalized collision energy of 25% ± 10% was used for all DIA spectral acquisitions.

Raw MS data were analyzed using quantitative DIA proteomics software, DIA-NN (version 1.8) (21). Complete mouse proteome from the UNIPROT database was used for neural network generation, enabled for deep spectral prediction. Protease digestion was set to trypsin (fully specific) allowing for two missed cleavages and one variable modification. Oxidation of methionine and acetylation of the protein N-terminus were set as variable modifications. Carbamidomethyl on cysteine was set as a fixed modification. Match between runs and remove likely interferences commands were enabled. The neural network classifier was set to double-pass mode. Protein interferences were based on genes. Quantification strategy was set to any liquid chromatography (high accuracy). Cross-run normalization was set to retention time–dependent. Library profiling was set to smart profiling.

Protein concentrations of fractions or insulin granule pellets from fractionation methods were analyzed using a Pierce BCA Protein Assay (Thermo Fisher) and subjected to SDS-PAGE. All antibodies used are listed in Tables 1 and 2. Protein signals were detected by chemiluminescence (Millipore) on a ChemiDoc MP Imaging System (Bio-Rad). Insulin SDS-PAGE was performed using a modified SDS-PAGE protocol described previously (22).

Cells and human pancreata were prepared for immunofluorescent staining following the standard procedure previously described (18,23). Cell cultures were fixed with 4% PFA. Fixed and sectioned human pancreata were provided by the St. Vincent’s Institute.

Microscope slides were imaged using the Leica LCS SP8 confocal microscope. Slides were imaged using a ×93 magnification glycerol lens (cells) or a ×40 oil lens (human pancreata) and white and stimulated emission depletion lasers. Leica LAS X software was used for image collection, and colocalization analysis done with Fiji ImageJ (24).

Insulin concentrations for INS-1 content and subcellular fractionation (Optiprep) were measured using the Ultra-Sensitive Mouse Insulin ELISA Kit (Crystal Chem). Proinsulin concentrations were measured by Rat/Mouse Proinsulin ELISA Kits (Mercodia).

Scamp3 KD cells were generated by transfecting INS-1 cells with siRNA-mediated oligomers. INS-1 cells were cultured on glass microscope coverslips for immunofluorescent imaging or 6-well and 12-well plates for SDS-PAGE and GSIS, respectively. Cells were transfected using two commercial rat Scamp3 siRNAs (TriFECTa DsiRNA; identifiers rn.Ri.Scamp3.13.2 no. 613664206 and rn.Ri.Scamp3.13.3 no. 613664207, and a nontargeting siRNA (TriFECTa DsiRNA; lot no. 0000865210) complexed with Lipofectamine 2000 (Thermo Fisher) at a concentration of 10 nmol/L. Cells were fixed or harvested 48 h after transfection to validate Scamp3 KD by SDS-PAGE and confocal imaging. GSIS assays were performed on INS-1 cells 48 h after transfection.

The RNeasy Kit (Qiagen) was used to extract total RNA. cDNA synthesis was performed using 500 ng of RNA with iScript Reverse Transcription Supermix (Bio-Rad). RT-PCR was labeled using SYBR select master mix (Thermo Fisher), and fluorescence was quantified by LightCycler 480 II (Roche) normalized to rat GAPDH. The following primer sequences were used: Scamp3(rat), forward, 5′-CAGGAGAAGAGCCAGAGTGC-3′, and reverse, 5′AGAGACTGCAGGGATAGGGG-3′; Pdx1(rat), forward 5′AAATCCACCAAAGCTCACGC-3′, and reverse, 5′AAGTTGAGCATCACTGCCAGC-3′; GAPDH (rat), forward, 5′-CAAAATGGTGAAGGTCGGTGTG-3′, and reverse, 5′-TGATGTTAGTGGGGTCTCGCTC-3′; insulin-1 (rat), forward, 5′-CCTTTGTGGTCCTCACCTGG-3′, and reverse, 5′TGCCAAGGTCTGAAGATCCC-3′; and insulin-2 (rat), forward, 5′-GCAGGTGACCTTCAGACCTT-3′, and reverse, 5′-CAGAGGGGTGGACAGGGTAG-3′.

GSIS in Scamp3 KD and control cells was performed as previously described (18). The HTRF (homogeneous time-resolved fluorescence) Insulin Ultra-Sensitive Detection Kit (Cisbio) was used to measure secreted insulin.

All output raw MS data were analyzed using Rstudio (version 4.1.1.). Variance stabilization normalization (vsn) was performed on the data using the vsn package. PCP analysis was done using the functions hclust, numClusters, and cutree (Rstudio). To assign proteins to organelles and complexes within cells, we used support vector machine learning (SVM) implemented with the svmOptimisation and svmClassication functions in the R package pRoloc (25,26). For intraclass correlation coefficients, our model fitted sample fraction identifications (fractions 1–12) as a random effect and normalized expression number for known ISG proteins as the outcome variable. The intraclass correlation coefficient (ICC) was calculated as the ratio of fraction variance to the total variance. To assign proteins within our samples as markers for other organelles distinct from ISGs, we used the publicly available Mus musculus library, available through the package pRolocdata (27). GraphPad Prism (version 8.0) was used for statistical analyses. Results were considered significant at P < 0.05.

Resources and any data required to reanalyze the data reported in this article are available upon request from the corresponding author. This article does not produce original code. Resources including the MS proteomics data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD051136; username: [email protected]; password: GnZTdMqQ".

Our subcellular fractionation protocol uses a three-step purification of MIN6 postnuclear cell lysates on a discontinuous gradient (8.8–30%) (Fig. 1A and B). Insulin concentrations were quantified in 16 Optiprep fractions, revealing two distinct peaks of insulin (Fig. 1C) corresponding to iISGs (fractions 5–7) and mISGs (fractions 11–13). SDS-PAGE analyses of the fractions confirmed enrichment of mature insulin in the later fractions and proinsulin in earlier fractions (Fig. 1D). Furthermore, through SDS-PAGE analysis, there was a clear enrichment of mISGs lacking contaminants from the cytoskeleton (β-actin) and endosomes (namely, EEA1). Enrichment of PCSK2 (an ISG marker) was also present and matched the ISG enrichment. Fraction 13 exhibited maximal insulin enrichment but also contained contaminating mitochondria (Fig. 1D). Fractions 11 and 12 were chosen for subsequent purification because these had less mitochondrial contamination but high insulin enrichment (Fig. 1D).

Optiprep fractions 11 and 12 were loaded atop a 27% Percoll solution and ultracentrifuged (Fig. 1E), then washed in sucrose to enable SDS-PAGE analyses and tandem MS. SDS-PAGE analysis using antibodies against markers of subcellular compartments for ISG (insulin; PCSK2), cytoskeleton (β-actin), autophagosomes (LC3A/B), mitochondria (ATP5A), and lysosomes (Lamp1) (Fig. 1F) confirmed insulin enrichment in later Percoll fractions, with minimal protein alignment from other organelles removed during the first-step fractionation.

All 12 sucrose-washed fractions from each experiment underwent MS-based proteomic analysis. Protein quantification was performed using label-free analysis in the MaxQuant package, and the protein-level data were subsequently analyzed by PCP and supervised machine learning using the R package pRoloc (26,27). Protein abundance across samples was normalized using the vsn package (28). After this, the covariance of five known marker proteins (Ins1, Ins2, Cpe, Pcsk2, and CgA) within each experiment (n = 7) was quantified by ICC to evaluate experiment reliability. ICCs were calculated based on random effects using a linear mixed model implemented in the package lme4 (29). Two of seven replicate experiments exhibited low and outlying ICC values (<0.3) and, therefore, were excluded from downstream analysis (Supplementary Table 2).

Through PCP analysis, proteins with similar abundance patterns are clustered together based on their Euclidean distances that are measured iteratively between each protein. Hierarchical clustering of all proteins (Supplementary Table 3) resulted in identification of 432 significant potential ISG protein candidates (Supplementary Table 4). Of interest, a single cluster contained well-established ISG marker proteins, including Ins1 and Ins2, islet amyloid polypeptide, prohormone convertases 1 and 2 (Pcsk1, Pcsk2), CgA, secretogranin-2 and -3, Cpe, zinc transporter 8, many proton ATP-ase subunits, syntaxins, synaptotagmins, vesicle-associated membrane proteins (Vamp2, Vamp3, and Vamp7), and Rab-GTPases (Rab3a, Rab27a) (Supplementary Table 4), suggesting high efficiency of the three-step purification method and validity of the proteomics data. Representative protein profiles from a single replicate illustrated a high enrichment of this ISG-specific protein cluster (Fig. 2A), further validating the purification method.

Subsequent Gene Ontology enrichment and STRING analysis of this ISG cluster (Fig. 2B) revealed a high strength index of proteins involved in insulin processing, secretory granule localization, as well as granule trafficking (teal in Fig. 2B). However, some subsets of proteins within this cluster were identified as contaminants, including ribosomal subunits (yellow), lysosomal cathepsins (purple) and subunits of mitochondrial electron transport chain proteins (red), shown in Fig. 2B. Although known ISG proteins were identified through PCP analysis, many other proteins within this cluster may also be ISG resident. For example, synaptotagmin-5, and synaptotagmin-like 4 are potential synaptotagmin isoforms present on ISGs in INS-1E cells (14). Additionally, the identification of synaptotagmin-13 (Syt13), synaptophysin (Syp), and syntaxin-7 (Stx7) suggest their potential residence on ISG membranes. Other proteins of interest are protein disulfide isomerases (e.g., Pdia6, Pdia3), which facilitate insulin processing (30).

PCP analyses offer comprehensive insights beyond single-snapshot proteomics of ISGs and enable unbiased matching of co-abundant proteins across subcellular fractions. Whereas previous ISG proteomes yielded only five consensus proteins (31), our PCP analysis identified these proteins along with other well-established ISG-resident proteins: ZnT8, ProSAAS, islet amyloid polypeptide, secretogranin-2, and secretogranin-3 (9,32–35). Subsequently, these 10 ISG proteins were selected as candidates for subcellular localization analyses. Marker proteins were assigned for other non-ISG subcellular compartments of the β-cell using the Mus musculus library available in pRoloc. The 10 ISG proteins chosen were manually added as markers. Using an SVM learning (a form of supervised machine learning) within pRoloc (26,27), unmarked protein profiles were correlated with marker protein profiles to generate a spatial proteome. Proteins assigned to the ISG exhibited high SVM score distributions (Supplementary Fig. 1) across experimental replicates, indicating strong assignment accuracy. Notably, these proteins had distinct profiles compared with proteins assigned to other subcellular localizations.

With confidence, we identified a total of 211 ISG proteins across five experiment replicates (Supplementary Table 5). Notably, Syp, Syt13, and Stx7 were among the ISG-associated proteins identified. Additionally, Scamp3, initially missed in the PCP analysis, was found in all five replicates in our targeted hyperplexed localization of organelle proteins by isotope tagging analysis (Supplementary Table 5). Through both analyses, the remaining proteins unassigned to the ISGs clustered with other subcellular protein complexes (e.g., mitochondria, proteasomes) (Fig. 2A). These proteins are noncontaminating, highlighting the potential for the identification of novel proteins in other subcellular localizations. Overlap of proteins identified in this study compared with previous ISG proteomes from INS-1 and INS-1E cell lines can be found in Supplementary Table 6.

We further validated Syt13, Stx7, Syp, and Scamp3 as ISG-resident proteins using immunocytochemistry. Colocalization analyses (Fig. 3A and B) showed that all were present in MIN6 cells and had significant positive correlation with insulin. Chromogranin-A, a well-established ISG cargo protein, exhibited high colocalization with insulin (Fig. 3A) and served as a positive control, whereas endoplasmic reticulum (PDI) and cis-Golgi (GM130) markers were negative controls and showed lower levels of colocalization with insulin. Syp was observed closer to the plasma membrane, whereas Scamp3 localized predominately in the perinuclear region with sparse cytosolic distribution.

To validate the association of these proteins in human β-cells, we conducted immunostaining of pancreatic islets of nondiabetic human donors (Fig. 3C). These proteins were found primarily in β-cells, demonstrating significant colocalization with insulin (Fig. 3D), with additional staining present in other cell types of the pancreatic islets for Syt13, Syp, and Scamp3. We then selected Scamp3 as a candidate for further functional studies.

First, we confirmed the expression and ISG colocalization of Scamp3 in INS1 cells (Supplementary Fig. 2A and B). We then investigated the functional consequence of Scamp3 KD in INS1 cells by depleting endogenous Scamp3 with silencer RNAs targeting two unique sequences of rat Scamp3 mRNAs (TriFECTa, IDT). At 48 h after transfection, INS-1 cells exhibited a significant reduction in Scamp3 mRNA levels (Supplementary Fig. 2C) resulting in ∼80% KD of Scamp3 protein in both siRNA conditions (Fig. 4A and B) compared with cells transfected with nontargeting control siRNA. Scamp3 KD INS1 cells had reduced cellular insulin content and GSIS, with no changes in proinsulin content (Fig. 4C–H) or Ins-1, Ins-2, and Pdx1 mRNA levels (Supplementary Fig. 2D–F). Although control cells displayed a typical insulin secretory response with a twofold increase in insulin secretion under the stimulated (16.7 mmol/L) to basal (2.8 mmol/L) glucose condition, neither Scamp3 KD condition exhibited significant differences between basal and stimulated conditions (Fig. 4G). This resulted in a significant decrease in the fold changes of stimulated and basal conditions, indicating a blunted secretory response in Scamp3 KD cells (Fig. 4H).

We further analyzed the distribution and abundance of Scamp3 in human β-cells in non-T2D and T2D islets (Fig. 4I). Although a nonsignificant decrease in Scamp3 fluorescence intensity was noted in T2D islets compared with non-T2D islets (n = 5/group; Fig. 4J), a slight reduction in Scamp3 and insulin localization was observed in T2D islets (Fig. 4K). These findings offer initial insight into potential changes in Scamp3 expression and localization linked to T2D.

Our study presents an efficient approach to isolating ISGs, successfully enriching for them using a three-step subcellular fractionation procedure. This method is applicable to various cell lines or primary cells. Through unbiased quantitative PCP analysis, we established the first MIN6 cell ISG proteome, identifying a total of 432 proteins associated with ISG. Additionally, machine-learning analysis enhanced the proteome, refining 211 proteins assigned with high confidence to ISGs. We validated the presence of 4 ISG-resident proteins of interest in MIN6 cells and human donor pancreatic islets. Through siRNA KD and functional analyses of Scamp3, we validated this novel ISG-associated protein’s involvement in the glucose-regulated secretory pathway in β-cells. Scamp3 was not clustered with ISG proteins after PCP, due to alignment with intracellular transport and cytoplasmic proteins. Inclusion of machine-learning analysis bolsters the data by elucidating other proteins that may be heterogeneously expressed and may be associated with ISGs at different points of the secretory pathway. Expanding the use of this framework may reveal alterations in ISG morphology or proteins, enhancing our understanding of changes in granule populations in β-cell dysfunction and development of T2D.

Previous proteomics analyses of ISGs in INS1 and INS-1E cell types have identified 50–140 candidate ISG proteins (11–14), with one study using PCP. Here, we used an unbiased PCP analysis to hierarchically cluster proteins with highly similar profiles, validating ISG separation techniques and identifying potential novel ISG resident proteins. Our approach identified many previously identified ISG proteins. Gene Ontology enrichment analysis of these clusters further support our findings, indicating high-strength indices for insulin processing, protein localization to secretory granules, and insulin metabolic processes.

PCP revealed the presence of many previously known ISG-associated integral membrane proteins, including members of the vSNARE family (Vamp2/3 and Vamp7), which are known contributors to insulin granule exocytosis (36,37). Additionally, our analysis identified numerous Rab-GTPases, including Rab3a, Rab27a, and Rab8a, known small G proteins that affect ISG movement (38).

Identified synaptic vesicle proteins such as membrane-localized synaptotagmins regulate ISGs exocytosis (39); Syt13 in humans is abundantly expressed in the pancreas (40), regulates islet formation, and its knockout affects α- to β-cell ratios in islets (41). Syt13 is downregulated in murine models of diabetes (db/db) and obesity (ob/ob) (42), demonstrating its clear involvement in β-cell development and GSIS. Synaptotagmins interact with syntaxin proteins (e.g., Snap23, VAMPs) to form SNARE complexes that regulate granule fusion during insulin exocytosis in β-cells. Syntaxin-7 is identified as an ISG-associated protein and forms SNARE complexes with Vamp4, Stx8, and VTI1B, facilitating proinsulin degradation within the iISGs (43). Additionally, Stx7 is involved in SNARE complex formation with Vamp7 and Stx8 during autophagosome formation (44).

Synaptophysin, typically associated with neurons and synaptic vesicle function, emerged as a novel ISG-associated protein (Fig. 3A and B). Although its roles in β-cells, possibly involving γ-aminobutyric acid secretion via synaptic-like microvesicles, is debated (45), we have observed its upregulation in murine islets under conditions of high-fat diet and in mouse models of diabetes (42).

One of the most intriguing ISG-associated proteins we uncovered is SCAMP3. Although the SCAMP protein family is widely expressed and found on post-Golgi vesicle membranes (46), SCAMP3’s role in β-cells remained elusive. However, other SCAMP proteins have been implicated in GSIS (47). Scamp3 is a ubiquitously expressed integral membrane protein found in post-Golgi trafficking pathways (48) and involved in membrane budding with associations with other adaptor proteins to form coated vesicle carriers (49). Other SCAMPs are linked to pore formation in later steps of secretory granule exocytosis, and Scamp3 KD in PC12 cells reduces fusion pore opening and the number of docked granules (50). Our study revealed clear association of Scamp3 within ISGs in rat, mouse, and human β-cells, and knocking down Scamp3 in INS1 cells significantly reduces GSIS (Fig. 4H) and insulin content (Fig. 4D–E). We observed no changes in mRNA transcript levels of either insulin isoform or Pdx1 with Scamp3 KD, indicating a role for Scamp3 in post-transcription pathways in β-cells (Supplementary Fig. 2D–F). A single transcriptome-wide association study has shown the presence of splice sites in the SCAMP3 gene that associate with T2D susceptibility in human pancreatic islets (51). In addition, previous proteomics studies have shown that mouse pancreatic islets exposed to high glucose levels result in a downregulation of Scamp3 compared with islets exposed to low-glucose conditions (52).

In summary, we provide the first evidence, to our knowledge, of Scamp3's role in ISG biogenesis, trafficking, or secretory pathways. Future studies should focus on how Scamp3 regulates these processes in the β-cell.

Acknowledgments. The authors thank the core facilities at the Charles Perkins Centre at the University of Sydney. This study was made possible with access to the Sydney Mass Spectrometry and Sydney Microscopy & Microanalysis facilities. The authors thank the organ donors and their families for their generosity. We also thank the members of St. Vincent’s Institute in the islet isolation program and Donatelife for providing research consent and provision of human pancreata.

Funding. This work was supported by the National Health and Medical Research Council (project grant GNT1139828 to M.A.K.). N.N. is supported by the University of Sydney Postgraduate Award. St. Vincent’s Institute receives support from the Operational Infrastructure Support Scheme of the Government of Victoria, Australia.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.A.K. and B.Y. conceptualized the study, contributed to validation, and funding acquisition. N.N. and B.Y. contributed to the methodology. N.N., H.W., A.M.S., and M.L. conducted the formal analysis. N.N., B.Y., C.F., H.W., and M.L. contributed to the investigation. M.A.K., H.E.T., and T.L. contributed to study resources. N.N., B.Y., M.L., and A.M.S. curated the study data. N.N. wrote the original manuscript draft. M.A.K., B.Y., A.M.S., and M.L. reviewed the manuscript. B.Y. contributed to data visualization and supervised the study. M.A.K. contributed to project administration and is the guarantor of this work and, as such, had full access to all the data in this study and takes responsibility for the integrity of data and accuracy of the data analysis.