Genome-wide CRISPR screen identified a role for commander complex mediated ITGB1 recycling in basal insulin secretion

Objectives Pancreatic beta cells secrete insulin postprandially and during fasting to maintain glucose homeostasis. Although glucose-stimulated insulin secretion (GSIS) has been extensively studied, much less is known about basal insulin secretion. Here, we performed a genome-wide CRISPR/Cas9 knockout screen to identify novel regulators of insulin secretion. Methods To identify genes that cell autonomously regulate insulin secretion, we engineered a Cas9-expressing MIN6 subclone that permits irreversible fluorescence labeling of exocytic insulin granules. Using a fluorescence-activated cell sorting assay of exocytosis in low glucose and high glucose conditions in individual cells, we performed a genome-wide CRISPR/Cas9 knockout screen. Results We identified several members of the COMMD family, a conserved family of proteins with central roles in intracellular membrane trafficking, as positive regulators of basal insulin secretion, but not GSIS. Mechanistically, we show that the Commander complex promotes insulin granules docking in basal state. This is mediated, at least in part, by its function in ITGB1 recycling. Defective ITGB1 recycling reduces its membrane distribution, the number of focal adhesions and cortical ELKS-containing complexes. Conclusions We demonstrated a previously unknown function of the Commander complex in basal insulin secretion. We showed that by ITGB1 recycling, Commander complex increases cortical adhesions, which enhances the assembly of the ELKS-containing complexes. The resulting increase in the number of insulin granules near the plasma membrane strengthens basal insulin secretion.


INTRODUCTION
Insulin secreted from pancreatic beta cells is used to stimulate glucose uptake for cellular fuel and storage. Insulin secretion is continuous but peaks soon after meals, upon glucose stimulation [1]. Glucose is the most potent stimulus of insulin secretion. Inappropriate responses to glucose by beta cells can cause hypoglycemia and hyperglycemia, with the latter representing a hallmark of diabetes. Much has been learned about glucose-stimulated insulin secretion (GSIS) in the past 6 decades [2]. However, much is still unknown about insulin secretion, particularly during fasting [3]. A prerequisite for insulin secretion is the biogenesis of insulin granules and their trafficking to the plasma membrane [4]. Insulin granules, both newly formed in the Golgi complex and recycled from endosomes, migrate along cytoskeletal tracks to the plasma membrane. The cortical microtubule stabilizing center (CMSC) that anchors microtubules is critical for the delivery of insulin granules to the proximity of the plasma membrane [5]. Upon arrival, insulin granules dock on the plasma membrane by the clustering syntaxin1 and Munc18-1 family [6]. Docking is critical for granules to be secreted following priming and a rise of intracellular Ca 2þ , which triggers the fusion of the plasma membrane and the granule membrane [7]. Identifying the mechanism of glucose-stimulated insulin secretion (GSIS) has been a major research focus [7]. Products from glucose metabolism induce insulin secretion through at least two pathways: the ATP-mediated triggering pathway and other metabolite-mediated amplifying pathway. The triggering pathway has been extensively described [7]. It consists of the ATP-sensitive Kþ channel and voltage-dependent Ca2þ channels. In contrast, the amplifying pathway is not well defined. Several glucose metabolites have been proposed to activate the amplifying pathway, including acetyl-CoA [8], phosphoenolpyruvate [9], NADPH [10], and adenylosuccinate [11]. However, the mechanism by which they amplify GSIS is not clear [12].
Half of the daily insulin secreted is released during fasting when blood glucose is at basal levels. Basal insulin is important for maintaining fasting blood glucose by inhibiting hepatic gluconeogenesis and glycogenolysis. Regulation of basal insulin secretion is not well studied compared to GSIS [3]. Similar to asynchronous release and spontaneous release of synaptic vesicles, basal insulin secretion may involve multiple mechanisms [13]. The cellular composition of synaptotagmins (Syt) with different Ca 2þ sensitivity can influence basal insulin secretion. For example, the high basal insulin secretion of immature beta cells is a result of higher Ca 2þ sensitivity due to lower expression of Ca 2þ -insensitive Syt4 [14]. However, a dynamic change of Ca 2þ sensitivity in adult beta cells has not been reported. The number of docked insulin granules also regulate basal insulin secretion. For example, loss of Rim2a, a protein important for insulin granule docking, leads to lower basal insulin secretion [15]. Other mechanisms that regulate basal insulin secretion remain to be discovered.
Integrin b1 (ITGB1) plays an important role in beta cell development and function. Constitutive loss of Itgb1 in beta cells impairs their proliferation [16], while conditional loss of Itgb1 in beta cells impairs GSIS [17]. ITGB1 targets insulin granules to the vascular facing cell cortex through focal adhesion-mediated positioning of the presynaptic scaffold proteins ELKS and Liprin [18,19]. ELKS is a multivalent protein that can interact with multiple components of the secretory machinery, including Liprin-a isoforms, RIMs, bassoon, LL5b and voltagedependent Ca2þ channels (VDCC) [20]. In pancreatic beta cells, ELKS is localized at the secretion sites of insulin granules, consistent with its scaffolding role in the secretion sites [21]. Although ELKS is important for synaptic vesicle docking in neurons [22], loss of ELKS in beta cells does not change insulin granule docking [23], possibly due to functional redundancy [22]. ELKS increases VDCC levels in vascular adjacent membrane [23]. Through its interaction with LL5b, ELKS can direct insulin granule distribution by anchoring the cortical microtubule stabilizing complex (CMSC) to focal adhesions [5].
Genome-wide CRISPR screen is a promising approach to better understand insulin secretion [24]. The CRISPR system consists of a Cas9 protein and a guide RNA, often in the form of single guide RNA (sgRNA). Each sgRNA targets Cas9 to a specific sequence. In a genome-scale screen, a library of sgRNAs (usually consists of 100,000 sgRNAs) targeting all genes are expressed individually with Cas9 to generate a library of mutant cells. Cells with abnormalities are collected to determine the underlying causes. The sgRNAs that are enriched or depleted in these cells identify the responsible genes. For functional defects that do not change viability like GSIS, however, a high throughput method to identify individual cells with functional defects is necessary for such screens. Currently, no highthroughput single cell secretion assay in beta cells is available. The lack of such an assay for insulin secretion has prevented the application of Cas9-based screen in insulin secretion. Therefore, we developed a system for assaying single cell secretion in beta cells and performed a genome-wide CRISPR screen to identify candidate genes that, when mutated, increase or decrease the ratio of secretion at high and low glucose concentrations. We identified a role for the Commander complex in insulin granule docking and insulin secretion under basal glucose conditions. Mechanistically, the Commander complex promotes basal insulin secretion at least in part through recycling ITGB1 from Rab4-positive endosomes to the plasma membrane. Defective ITGB1 recycling decreases the number of cortical focal adhesions and the ELKS-containing complexes. Defective ITGB1 recycling also reduces the number of docked insulin granules and their secretion.

Lentivirus production
HEK293FT cells (a gift frorm David Jacobson lab, Vanderbilt) were used for virus packaging. Cells were cultured in DMEM GlutaMax with 10% FBS, 100IU/mL penicillin, and 100 mg/mL streptomycin to 50e70% confluency in 100 mm dishes. The media was then switched to DMEM GlutaMax with 10% FBS for transfection. The following DNA mixture (18.65 mg lentiviral plasmid, 13.9 mg packaging plasmid (pCMV-dR7.74psPAX2), and 5.56 mg envelope plasmid (pMD2. G) was used together with Takara CalPhosÔ Mammalian Transfection Kit (#631312) as instructed. Cells were incubated with the DNA mixture for 5e7 h and then the transfection media was aspirated and replaced with fresh DMEM GlutaMax with 10% FBS. Medium containing lentivirus was harvested 72 h later and frozen at À80 C. The viral medium was thawed for transduction of MIN6 cells.
2.3. Cell culture, viral transduction and selection MIN6 cells were obtained from Addexbio and grown in AddexBio Advanced DMEM Medium(C0003-04) containing 15% FBS, 0.05 mM 2-mercaptoethanol, and penicillin (100 U/ml)/streptomycin (100 mg/ mL). Cells were cultured at 37 C with 5%CO 2 . The cells were transduced at seeding to enhance the efficiency. Polybrene (10 mg/mL) was added with 3rd-generation lentiviruses containing the abovementioned lentiviral plasmids. The virus was left on the cells for 24 h and then replaced with fresh media. The virus was then allowed to express for 2e3 days. After transduction, reporter expression was confirmed by equivalent EGFP expression and Cas9 expression was confirmed by western blot.
2.4. Glucose-stimulated insulin secretion assays (NanoLuc and ELISA) The MIN6 cells were seeded in a 12-well plate to 80e90% confluency. Cells were starved for 1 h in 1 mL KRB (114 mM NaCl, 5 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 20 mM HEPES, 2.5 mM CaCl2, Original Article 25.5 mM NaHCO3 and 0.2% BSA) containing 1 mM glucose. They were then cultured with freshly prepared KRB with 1 or 25 mM glucose for another hour. In glucose response curve experiment, cells were cultured in freshly prepared 1 mM, 5.6 mM, 11 mM and 25 mM KRB for another hour after starvation. The conditioned buffer was collected and centrifuge at 2,000 rpm for 5 min to obtain supernatants. The cells were lysed with RIPA buffer (R0278, Sigma) followed by centrifugation to obtain supernatants. The supernatants were then measured for NanoLuc using Promega Nano-GloÒ Luciferase Assay System (Promega) or for insulin using a mouse insulin ELISA kit (Mercodia), and DNA using AccuBlueÒ dsDNA Quantitation kit (Biotium) following manufacturer's instruction. The reactions were read using a BioTek SynergyÔ H4 Hybrid Microplate Reader. NanoLuc and Insulin content was normalized to total DNA content (AccuBlueÒ dsDNA Quantitation, Biotium) in each sample. At least three independently generated sets of samples were analyzed under each condition.
2.5. SNAP staining of MIN6-6 cells in GSIS, imaging and flow cytometry analysis MIN6-6 cells were starved for 4 h in DMEM with 10% FBS and 1 mM glucose followed by a 15-minute treatment with SNAP-SurfaceÒ Block (5 mM) in the same medium. After washing out the blocker, cells were sequentially treated with DMEM with 10% FBS, 1 mM glucose and 5 mM SNAP-SurfaceÒ Alexa FluorÒ 546 for 20 min and DMEM with 10% FBS, 25 mM glucose and 5 mM SNAP-SurfaceÒ Alexa FluorÒ 647 for another 20 min, with washes after each treatment. For imaging, cells were pre-seeded on coverslips. After staining, cells were fixed in 4% PFA, mounted on slides with AquaMount (Thermo Scientific), and imaged with Zeiss LSM880 laser scanning spectral confocal microscope (Carl Zeiss). For flow cytometry, a positive control group was treated with 40 mM KCl instead of 25 mM glucose, and a negative control group was treated with 300uM Diazoxide in the presence of 25 mM glucose. Cells were harvested and washed before fixation in 4% PFA. Fluorescence intensity of Alexa Fluor 647 and Alexa Fluor 546 was determined using a 5-laser BD LSRII, and data were analyzed using FlowJo software (BD Biosciences).
2.6. Genome-wide CRISPR screens Mouse CRISPR Knockout Pooled Library (Brie) [25] was purchased from Addgene. Approximately 150 million MIN6-6 cells were transduced at low multiplicity of infection (0.3e0.4). Transduced cells were selected and expanded in puromycin-supplemented media over 14 d to allow time for gene-editing and depletion of the target proteins before experiments. Two independent screens were performed. For each screen, 300 million cells were starved, blocked and sequentially stained with SNAP Alexa Fluor 546 and 647 in 1 mM and 25 mM glucose medium, respectively as described above. The cells were then washed twice in cold PBS, dissociated by 0.25% Trypsin with EDTA (Gibco), and fixed in 4% PFA-PBS for 15 min at room temperature with intermittently vortexing. Cells were then washed twice with cold PBS and resuspend in PBS þ1% BSA. The cells were FACS sorted into the top 5% or bottome 5% according to their Alexa Fluor 647/Alexa Fluor 546 ratio using a 5-laser BD FACSAria III. Approximately 10 million cells were recovered for each population. Sorted cells were then pelleted by centrifugation, and the cell pellet was frozen at À80 C before genomic DNA isolation. Approximately 80 million unsorted cells (1,000 Â coverage per library element) were saved for screen data normalization. Genomic DNA was extracted using NK cell lysis buffe (50 mM TriseHCL, 50 mM EDTA, 1% SDS, pH ¼ 8.0) from sorted or unsorted cells, respectively. The genomic DNA were sent to CD Genomics for sequencing of sgRNAs. Readcount tables and gene enrichmentanalysis were performed using the MAGeCK algorithm (https://sourceforge.net/projects/mageck/).

Real-time qPCR
Commd3 knock down and control cells were washed two times with cold PBS prior to lysis by TRIZOL reagent. RNA was isolated using Direct-Zol columns and concentrated using RNA Clean and Concentrator-25 columns. Two (2) mg of isolated RNA was reverse transcribed using oligo-dT and MMLV reverse transcriptase. The cDNAs were subjected to PCR on a BioRad CFX96 machine and amplicons were detected using SYBR green. Specific SNX14 primers and HPRT1 primers (reference gene) were used (see Table S1). PCRs were run in a LightCycler 96 Real-Time PCR system (Roche). All the qPCR primers used in this study are listed in Table S1. 2.10. FluoZin-3 assay FluoZin-3 assay was performed 2 days after seeding parental MIN6 cells and MIN6-6 cells (50:1) at <5% confluency. These MIN6-6 cells allow for more accurate determination of focal plane and TIRF angle for the assay. Cells were incubated at 37 C in low glucose (5.6 mM glucose) KRB (114 mM NaCl, 5 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 20 mM HEPES, 2.5 mM CaCl2, 25.5 mM NaHCO3 and 0.2% BSA) for 1.5e2 h with a change of buffer after 1 h. Cell-impermeant FluoZin-3 dye (Thermo Fisher Scientific, Cat#: F24194) in KRB buffer was added to a final concentration of 20 mM. Glucose was added together with the dye to a final concentration of 5.6 mM or 25 mM. Focal plane and TIRF angle were refined after dye addition and the recording (60 ms exposure, no delay) started within 2 min after addition of the dye for 10 min. The videos were processed by substracting each frame from the previous frame. Every five subtracted images were grouped as a max projection through time using the Grouped Z-project function in ImageJ and analyzed at 300 ms time resolution. Individual secretion events or 'flashes' were identified using ImageJ software [26] using the point tool. FluoZin-3 movies were blinded then analyzed manually. Cell borders were identified by ROI based on DIC images. Cell number and secretion events within each identified cell were calculated in ImageJ. The detailed protocol has been published previously [27]. Supplementary video related to this article can be found at https://doi. org/10.1016/j.molmet.2022.101541 2.11. Immunofluorescence staining Cultured MIN6 cells on coverslips were treated with experimental conditions for 24 h. The cells were rinsed with ice-cold PBS and fixed with 4% PFA for 10 min at room temperature followed by permeabilization in 0.3% Triton X-100. The cells were then incubated with ITGB1 antiboy (

Confocal and TIRF imaging
Images were acquired using a Zeiss LSM880 wih Airyscan using a 63 Â oil lens. DAPI, Alexa Fluor 488, Alexa Fluor 555, and Alexa Fluor 647 images were acquired sequentially using 405-, 488-, 561-, and 633-nm laser lines. Image quantification was performed using the ImageJ software. For quantification of the size and number of focal adhesions and ELKScontaining complexes, cells were fixed and stained with ITGB1 antibody and a focal adhesion marker (vinculin or paxillin antibody) or ELKS antibody. Basal plane images were converted to 8-bit and grayscale. Cells were outlined and thresholded to avoid artificial over-or underfilling of areas. Average intensity of ITGB1 per cell, thresholded area per cell and outlined cell area were computed while the number of focal adhesions or ELKS-containing complexes per cell was counted manually. The size of these complexes was calculated by dividing the total thresholded area per cell by the number of the cognate complex per cell [28,29]. Two methods were used to determine the number or density of docked insulin granules. In WT and Commd3 KD MIN6-6 cells, TIRF detectable phogrin-EGFP marked vesicles were defined as docked insulin granules. The cells were imaged using a Nikon TIRF2 System for TE2000 with a TIRF 100 Â 1.49 NA oil immersion lens. The granules were counted with the ImageJ software.
To determine the number of stably docked vesicles, pCMV-NPY-EGFP transfected WT and Commd3 KD MIN6 cells were preincubated with 5.6 mM glucose for 1h. After replacing the medium with freshly made KRB buffer containing 5.6 mM, the cells were imaged on a TIRF microscope for 5 min at 1 frame/s. The first and last 5 frames were averaged. The averaged images were compared to identify insulin granules that remained in the same vicinity. These granules were considered as stably docked.
To determine the release probability of insulin granules, pCMV-IAPP-Emerald transfected WT and Commd3 KD MIN6 cells were preincubated in medium with 5.6 mM glucose for 1h. After replacing the medium with freshly made KRB buffer containing 5.6 mM, the cells were imaged on a Nikon TIRF2 System for TE2000 with a TIRF 100 Â 1.49 NA oil immersion lens for 10 min at 1 frame/s. A secretion event was defined as a sudden disappearance (with in 1 s) of a docked vesicle.

Statistics
Results are expressed as the mean AE SEM unless otherwise specified. For data sets where the distribution was appropriate (normal distribution as determined by the D'Agostino & Pearson omnibus normality test), statistics were calculated by one-way ANOVA with Tukey's multiple comparisons test. When the data distribution was non-normal, a KruskaleWallis nonparametric and multiple comparison test was used (>2 data sets). Means were compared using Student's t test when two experimental conditions were compared. GraphPad Prism 9 was used for statistical analyses and graphical representations. Significance was defined at p 0.05.

Development of a MIN6 reporter cell line for bulk and single cell GSIS assay
A high throughput single cell assay of exocytosis of dense core granules (insulin granules) is essential for a genomewide CRISPR/Cas9 screen for genes that regulate GSIS. We chose to use a lumenal SNAP tag in combination with membrane impermeant fluorescently labeled SNAP substrates to label exocytic Insulin granules. SNAP is a 20 kDa engineered form of the DNA repair protein O6-alkylguanine-DNA alkyltransferase that reacts specifically and rapidly with benzylguanine (BG) derivatives to form an irreversible covalent conjugation [30]. As the secretory granules fuse with the plasma membrane to secrete insulin, they also expose their lumen to the extracellular environment. To anchor the SNAP tag in the lumen of Insulin granules, we inserted it into the lumenal region immediately downstream of the dibasic proteolytic site of the granule-specific integral membrane protein, phogrin [31]. Although its function in insulin secretion is still not clearly defined, phogrin has been used to anchor various sensors to the lumen of secretory granules [32e34]. To easily identify cells that express SNAP-Phogrin, we also fused enhanced green fluorescent protein (EGFP) to the cytoplasmic tail of phogrin. This allows selection and visualization of cells that express SNAP-Phogrin-EGFP fusion protein.
Many beta cell clones need to be tested to identify one with appropriate glucose responsiveness. This is because most beta cell lines consist of cells with heterogenous glucose responsiveness [35,36]. To facilitate clone selection, we also linked the coding sequence for a modified preproinsulin (unprocessed insulin) to that of SNAP-Phogrin-EGFP with a viral 2A sequence. The viral 2A peptide causes ribosome skipping, allowing production of multiple independent proteins from one transcript [37]. The preproinsulin gene was modified by replacing the coding sequence for the non-functional C chain with that of nanoluciferase (NanoLuc) [38,39], the smallest luciferase with the strongest bioluminescence. NanoLuc will be packaged together and cosecreted with insulin. Thus, nanoLuc activity in the medium, which can be easily measured with high sensitivity, is a surrogate for secreted insulin. Therefore, NanoLuc allows assaying of GSIS activity in a large number of samples [38]. The reporter gene is under the control of the rat insulin promoter (Rip) as depicted in Figure 1A and was inserted in a lentiviral vector. We chose MIN6 cells as the target for our CRISPR screen. MIN6 cells are derived from mouse insulinoma [40]. Early passage (<40) MIN6 cells are glucose responsive and have simple culture conditions [40,41]. Early passage MIN6 cells were transduced with viral Original Article particles of the construct and a Cas9-expressing construct, Lenti-Cas9-2A-Blast [42] (Figure1B). After two weeks of selection, EGFP-positive cells were FACS sorted into two 96 well plates to generate clones. NanoLuc assays were performed to characterize glucose induction index of each subclone ( Figure 1C, Fig. S1A). Seven subclones with a glucose induction index >2 fold were expanded and tested for the expression levels of Cas9 by western blot. The #6 subclone was chosen for additional experiments because it showed both a high GSIS index ( Figure 1C) and high Cas9 expression ( Figure 1D). We named this clone MIN6-6. When MIN6-6 cells were incubated sequentially in 1 mM glucose and 25 mM glucose for 30 min each, the nanoLuc activity in the latter was more than 4 fold higher, or a GSIS index >4. With a 1h incubation, the GSIS index was 10, similar to the GSIS index of 11 determined by an insulin ELISA assay (Figs. S1BeC). These results indicate that MIN6-6 is glucose responsive and the nanoLuc secretion assay is a viable surrogate for insulin secretion.
We next determined whether SNAP labeling is also glucose responsive. After queching potential surface SNAP with a non-fluorescent cell impermeant SNAP substrate, MIN6-6 cells were exposed to medium with 1 mM or 25 mM glucose containing a cell-impermeant SNAP surface dye Alexa Fluor 546. After 20 min, cells were fixed and imaged with a confocal microscope. The signal of the SNAP label was much higher in MIN6-6 cells in high glucose medium than these in low glucose medium ( Figure 1E). In addition, the SNAP labels were mostly intracellular and colocalized with EGFP positive granules, probably due to compensentary endocytosis. Parental MIN6 cells without the reporter showed little SNAP labeling either in low or high glucose medium (Figs. S1CeD), indicating labeling specificity. Last, we determined whether MIN6-6 cells are compatible with the high throughput GSIS assay in single cells. After quenching, MIN6-6 cells were sequentially incubated in low glucose medium with SNAP surface dye Alexa Fluor 546 then in experimental media with SNAP surface dye Alexa Fluor 647 for 20 min each before being fixed. The  (Figure1F). These data confirmed that SNAP labeling intensity positively correlates with insulin secretion. MIN6-6 cells are therefore suitable for a genomewide CRISPR screen for genes affecting insulin secretion.

3.2.
A genome-wide CRISPR-Cas9 screen identifies a role for the commander complex in insulin secretion To screen for genes important for insulin secretion, we first established a MIN6-6 mutant library using the Mouse CRISPR Knockout Pooled Library (Brie), containing 78,637 gRNAs targeting 19,674 genes and 1,000 non-targeting control sgRNAs [25]. After a two-week puromycin selection, the cells were harvested. Some cells (w8x10 7 ) were used  for genomic DNA to determine the sgRNA distribution, while some were used in two independent screens to identify regulators of insulin secretion. For the screen, the cells were cultured in 1 mM glucose DMEM for 4h, followed by a 20 min incubation with SNAP-Surface Block. They were then sequentially labeled with SNAP Surface Alexa Fluor 546 in 1 mM glucose DMEM for 20 min, followed by SNAP Surface Alexa Fluor 647 in 25 mM glucose DMEM for another 20 min. Cells were dissociated and fixed for fluorescence-activated cell sorting (FACS). Cells displaying the highest or lowest 5% Alexa Fluor 647/546 MFI ratio were collected (Figure 2AeB). The genomic DNA from these cells was isolated and the sgRNA representation was quantified by deep sequencing. Quality control showed >88% sgRNAs were present. Gini index showed an even distribution in the control sgRNA library. In the selected pools, higher cell numbers resulted in a more even distribution (Figs. S2AeC). The abundance of many sgRNAs were substantially increased in both replicates compared with unsorted control population. Nontargeting control sgRNAs exhibited no enrichment ( Figure 2C), confirming that the screens were effective and specific. Genes were ranked based on the enrichment of their targeting sgRNAs using the MAGeCK (model-based analysis of genome-wide CRISPR-Cas9 knockout) algorithm [43]. The sgRNA of a few dozens of genes were enriched in the pool with increased glucose stimulation index. Some were known negative GSIS regulators such as Glud1 and Rab7. However, only a few genes whose sgRNAs were enriched in the pool with decreased glucose stimulation index (Fig. S2D). The small number of hits may be due to the highly stringent cell selection. We focused on gene hits in the pool of high glucose stimulation index because these may be drug targets. Several of these genes encode enzymes catalyzing lipid or carbohydrate metabolism, mediators of gene expression, proteins important for vesicular trafficking, as well as genes lacking annotated functions. Several members of Commander complex were among the top gene hits. These include COMM Domain Containing 5 (COMMD5), VPS35 Endosomal Protein Sorting Factor Like (VPS35L), COMM Domain Containing 6 (COMMD6), COMM Domain Containing 3 (COMMD3), COMM Domain Containing 10 (COMMD10), VPS29 Retromer Complex Component (VPS29), COMM Domain Containing 8 (COMMD8), and COMM Domain Containing 2 (COMMD2) (Figure 2DeE). The identification of multiple components of the Commander complex provides strong support for its role in insulin secretion.

Loss of Commd3 decreases basal insulin secretion
The Commander complex is important for endosomal recycling of a subset of proteins. It comprises 15 different proteins, including COMMD 1e10, CCDC93, CCDC22, VPS35L, VPS26C, and VPS29. COMMD 1e10 are thought to always exist in a complex, and loss of one lowers the expression of the rest [44]. Mutations in Commd genes are linked to human RS/3C syndrome, a developmental disease in humans [45]. Commd4 and Commd7 have been implicated in T2D in two genome-wide association studies [46,47]. However, the role of Commander complex in pancreatic beta cells has not been characterized. We first validated that loss of the Commander complex in beta cells improves glcuose stimulation index. We generated two Commd3 KD populations in MIN6 and MIN6-6 cells using CRISPR/Cas9 with two different sgRNAs. Western blot analysis indicated that COMMD3 protein levels were decreased in both populations, with sgRNA2 population having 80e90% reduction ( Figure 3B). In addition, 4 other Commander complex components, COMMD1, COMMD5, CCDC22, and CCDC93 were also reduced by 60e85% in the sgRNA2 population ( Figure 3A), consistent with previous studies [48,49]. These cells are referred to as Commd3 KD cells. Interestingly, qRT-PCR analysis of genes for the Commander complex in control and Commd3 KD cells revealed a 40e60% reduction in the Commd3 KD cells compared to control MIN6 cells ( Figure 3B). The results suggest a feedback control at the level of either transcription or mRNA stability. Commd3 KD cells showed a high glucose stimulation index, validating the screen ( Figure 3C). Moreover, this increase of GSIS index is mainly due to a decreased basal insulin secretion but not stimulated insulin secretion ( Figure 3D). The glucose response curve showed less insulin secretion at 1 mM and 5.6 mM glucose but a normal insulin secretion at 11.1 mM and 25 mM glucose, indicating COMMD3 regulates basal insulin secretion (Fig. S3A). We confirmed that this is not specific to Commd3 KD, we also generated Commd5 KD cells. Similar to Commd3 KD cells, Commd5 KD cells also have a lower glucose induction index due to lower basal insulin secretion (Figs. S3CeE). To determine whether the low basal secretion is due to increased beta cell maturity, we compared the transcript levels of multiple maturation markers by qRT-PCR. No change was detected in the Commd3 KD cells (Fig. S3B). Therefore, the Commander complex enhances basal insulin secretion without affecting beta cell maturity.

Loss of Commd3 reduces the number of exocytic events and docked insulin granules
To understand the underlying mechanism of decreased basal insulin secretion, we used FluoZin-3 based live imaging on a total internal reflection fluorescence (TIRF) microscope [27,50] to examine insulin granule docking and secretion in control and Commd 3 KD MIN6 cells. FluoZin-3 increases fluorescence emission upon binding to Zn 2þ , a component of insulin crystals in the insulin granules. And due to the cell impermeability of FluoZin-3, it only binds Zn 2þ after secretin. A "flush" of fluorescence occurs with each exocytic event ( Figure 4A). FluoZin-3 was added into the KRB buffer with 5.6 mM or 25 mM glucose and the cells were imaged for 10 min at a rate of 60 ms/frame. At 5.6 mM glucose, we detected insulin secretion in a lower fraction of Commd3 KD cells (20%) than in control MIN6 cells (30%) (Figure 4Be E). In addition, there were fewer secretion events on average in active Commd3 KD cells than in control cells (Figure 4BeD,F). At 25 mM glucose, there was no difference in either the fraction of active cells or average secretion events per cell between Commd 3 KD and control cells (Figure 4BeD). These results support that loss of Commd3 decreases basal but not glucose-stimulated insulin secretion. A decrease in basal insulin secretion could result from a decreased fusion probability or a reduced number of docked granules [51]. To investigate if the decreased basal secretion is due to fewer docked insulin granules in Commd3 KD cells, we measured the number of vesicles near the basal membrane in the SNAP-phogrin-EGFP expressing MIN6-6 cells using TIRF microscopy. We found a significant reduction of the density of cortical insulin granules in Commd3 KD MIN6-6 cells compared to the control MIN6-6 cells at 5.6 mM glucose. After 1h of glucose stimulation, however, no difference was detected in the density of cortical insulin granules between Commd3 KD and control MIN6-6 cells ( Figure 5AeB). As vesicle docking is dynamic [52], some of the cortical insulin granules may be transiently docked, we therefore compared the number of insulin granules that remain docked for 5 min in pCMV-NPY-EGFP transfected control and Commd3 KD MIN6 cells using TIRF microscopy. There was a significant decrease of stably docked insulin granules in Commd3 KD cells compared to control cells (Figure 5CeD). To determine whether a decreased release probability also contributes to the lower basal secretion, we determined the number of exocytic granules in pCMV-IAPP-Emerald transfected control and Commd3 KD MIN6 cells. Exocytosis is indicated by a sudden increase followed by a disappearance of fluorescence (with in 1 s) [53,54]. There was no significant difference between Commd3 KD and control MIN6 cells (Figure 5EeF, Video 1). Together, these results suggest the decreased basal insulin secretion in Commd3 KD cells is due to fewer docked insulin granules.
Original Article 3.5. Loss of Commd3 causes retention of ITGB1 in Rab4-positive recycling endosomes Having demonstrated that Commd3 KD decreases basal insulin secretion due to fewer docked insulin granules, we next investigated the underlying molecular mechanism. Commander complex associates with SNX17, which mediates recycling of membrane proteins with a FxNPxY or FxNxxY sorting motif, such as ITGB1, LDLR and LRP1 [55].
Ldlr is expressed in MIN6 cells and is important for cholesterol uptake Red bars, mean. One-way ANOVA and multiple comparison tests. ****P < 0.0001, **P < 0.01, *P < 0.05. [56]. Cholesterol is important for the formation of Syntaxin clusters on the plasma membrane, where secretory vesicles preferentially dock [57]. Defective LDLR recycling therefore could impair insulin granules docking. However, b cell function is normal in LdlrÀ/À mice [56]. Therefore, we excluded defective LDLR recycling as the mechanism for the decreased basal secretion in the Commd3 deficient cells. Singla et al. reported that knocking down Commd3, Ccdc93, Vps35l or Vps26c causes ITGB1 accumulation in WASH-positive intracellular vesicles in HeLa cells [48]. Mice with ITGB1 ablation in beta cells have decreased beta cell mass and impaired GSIS [16,58,59]. The role of ITGB1 recycling in insulin secretion has not been reported. Therefore, we investigated whether Commd3 KD affects ITGB1 endocytic trafficking in MIN6 cells by using EGFP-tagged Rab4, 5, 7, and 11 to mark different endocytic compartments. At 5.6 mM glucose, there was extensive colocalization of ITGB1 and Rab4 in Commd3 KD cells, in contrast to the low colocalization in control MIN6 cells (Figure 6AeB), indicating that ITGB1 proteins are stuck in the Rab4-positive early endosome or recycling endosomes. These results are consistent with a recycling defect in ITGB1 trafficking. At 25 mM glucose, however, the colocalization of ITGB1 and Rab4 was low in both Commd3 KD cells and control MIN6 cells, indicating the activation of a different recycling mechanism. There was no difference in ITGB1 colocalization with Rab5, Rab7 or Rab11 between Commd3 KD cells and control cells in both basal glucose and high glucose conditions (Fig. S4).
3.6. Defective ITGB1 recycling impairs insulin granule docking and basal insulin secretion We next investigated whether the decreased basal insulin secretion and insulin granule docking in Commd3 KD cells is due to defective SNX17-mediated ITGB1 recycling. We mutated the FxNPxY motifs of human ITGB1 ( Figure 6C) and performed cotransfection to co-express Rab4-EGFP with either the wild type or the mutant ITGB1 in MIN6 cells. We found that, just as in the Commd3 KD cells, there was an increase of ITGB1 in Rab4-positive vesicles in cells expressing the mutant ITGB1 compared to those expressing the wild type ITGB1 (Figure 6DeE). As ITGB1 can dimerize with at least 12 a subunits [60], we examined whether endogenous a subunits are also retained in the Rab4-positive vesicles. Integrin AV transcript is the most abundant a integrin in MIN6 cells [61]. In MIN6 cells expressing mutant ITGB1, ITGAV was also sequestered in the Rab4-positive vesicles as expected (Figure 6FeG). To determine whether the mutant ITGB1 affects insulin secretion, we co-transfected expression vectors for mCherry and mutant ITGB1 in MIN6 cells. We then performed the FluoZin-3 assay to quantify the insulin secretion events under 5.6 mM and 25 mM glucose in transfected and untransfected cells. At 5.6 mM glucose, we found a much lower percentage of cells that secrete insulin compared to untransfected cells. The average secretion events in cells expressing the mutant ITGB1 were also decreased compared to untransfected cells  ( Figure 7AeB). However, at 25 mM glucose, the percentage of secreting cells and secretion events per cell were similar between transfected and untransfected cells (Figure 7CeD). We further investigated whether expression of the mutant ITGB1 reduces insulin granule docking. We measured the number of docked vesicles near the basal membrane in MIN6 cells co-expressing wild type or mutant ITGB1 with mCherry. At 5.6 mM glucose, the insulin granule density near the basal membrane was decreased in MIN6 cells expressing the mutant ITGB1 compared to these expressing the wildtype ITGB1. After 1h of glucose stimulation, however, no difference was detected in the number of insulin granules between cells expressing the mutant and wild type ITGB1 (Figure 7EeF). Taken together, these results indicate that Commd3 KD cells have defective ITGB1 recycling at basal glucose conditions, which decreases insulin secretion and insulin granule docking.
3.7. Defective ITGB1 recycling reduces cortical focal adhesions How defective ITGB1 recycling reduces insulin granule docking and insulin secretion is unknown. Integrin recycling regulates focal adhesion remodeling [62], which is important for GSIS [63]. Therefore, we investigated if the defective ITGB1 recycling from Commd3 deficiency affects focal adhesions. In Commd3 KD cells, the ITGB1 in the basal PM was significantly lower than control cells as shown by immunofluorescence ( Figure 8A 3.8. Reduced cortical focal adhesions diminish the assembly of ELKS complexes Focal adhesions plays an important role in polarized insulin secretion [18]. The underlying mechanism remains unclear but may relate to the ITGB1 mediated assembly of a complex consisting of Bassoon, Piccolo, RIMs, ELKS and liprins that are important for vesicle docking and exocytosis [64,65]. ITGB1 has been shown to recruit ELKS, an important scaffold of the complex, through its interaction with LL5b and Talin-KANK1 [66,67]. ELKS is important for insulin granule docking and secretion in beta cells [21,23]. Therefore, we compared cortical ELKS in Commd3 KD and control cells. The number of ELKS clusters was decreased and their size smaller at basal glucose compared to control cells. After 20 min high glucose stimulation, however, the number and size of ELKS clusters were indistinguishable in Commd3 KD and control cells ( Figure 9A,C-E). Similarly, in cells expressing the recyclingdefective ITGB1, there were fewer and smaller ELKS clusters than in untransfected cells or cells expressing the wildtype ITGB1 at basal glucose condition. However, the difference disappeared after 20 min of high glucose stimulation ( Figure 9B,FeH). These data suggest that decreased ELKS-containing complexes may be responsible for fewer docked insulin granules and decreased basal insulin secretion. Taken together, our results demonstrated a previously unknown function of the Commander complex in basal insulin secretion. We showed that by recycling ITGB1, Commander complex increases cortical adhesions, which enhances the assembly of the ELKScontaining complexes. The resulting increase of the number of insulin granules near the palsma membrane strengthens basal insulin secretion.

DISCUSSION
In the present work, we established a highly glucose responsive beta cell reporter line (MIN6-6) with multiple engineered functionalities. This line allows for CRISPR/Cas9 mutagenesis, quantification of bulk insulin secretion by a straightforward NanoLuc assay, and en masse quantification of glucose-stimulated insulin granule exocytosis in single cells. We conducted a genome-wide screen for genes affecting glucose stimulation index. Several components of the Commander complex were among the top hits of cells with increased glucose stimulation index. We demonstrated that loss of function of the Commander complex decreases basal glucose secretion with little effect on GSIS. The decreased basal insulin secretion was due, at least in part, to fewer docked insulin granules. The reduction in docked granules partially results from impaired ITGB1 recycling and the consequent reduction of cortical focal adhesions and ELKS-containing complexes. The screen identified several genes known to regulate insulin secretion, demonstrating the validity of the assay. For example, an activating mutation in Glud1 causes congenital hyperinsulinism [68]. Beta cells with increased Glud1 activity have increased basal but blunted glucose-stimulated insulin secretion [69,70], consistent with an increased glucose stimulation index in cells with lower Glud1 activity. Similarly, shRNA-mediated Rab7 knockdown increases glucosestimulated insulin secretion [71], consistent with the screen results. Interestingly, a number of genes encoding non-respiratory components of the mitochondria were also among the top hits in the pool of increased glucose induction index. Whether they affect basal or stimulated insulin secretion remains to be determined. Only limited genes were identified in the pool of decreased glucose induction index, however. This may be a result of the heterogeneity of the glucose responsiveness of the cells despite their clonal origin, the small shift in the distribution of the glucose induction index ( Figure 1F), and the stringent selection we implemented ( Figure 2B). A less stringent selection may allow identification of more genes. The screen identified multiple components of the Commander complex, which consists of 15 subunits. COMMD1-10, along with CCDC93, can all interact with CCDC22 to form the CCC complex. The CCC complex interacts with the retriever complex, consisting of Vps35l, Vps26c and Vps29, to form the Commander complex [45]. Although some of the COMMD proteins may have preferred binding partners [72], they can all bind to each other promiscuously [44]. All 15 genes are expressed in MIN6 cells. Interestingly, knockdown of Commd3 led to not only a reduction of protein levels of other components, but also downregulation of mRNA levels of all other genes in the complex, suggesting feedback regulation at transcription or mRNA stability (Figure 3AeB). Previously, it has been reported that hepatocytespecific knockout of Commd1, Commd6, or Commd9 to a lesser extent, destabilizes all Commander components [49]. Similar results have been seen in HeLa cells [48]. Our results indicate that the downregulation may be at least partially due to a decrease in mRNA levels. We found that the increased glucose stimulation index in Commd3 mutant cells results from lower basal insulin secretion. Both NanoLuc assay of bulk insulin secretion and FluoZin-3 assay of insulin secretion in single cells indicated a lower basal secretion and normal glucosestimulated secretion in Commd3 mutant cells ( Figure 3D, Figure 4Be  D). Although lower basal secretion is often a sign of maturity [73], there was no difference in the expression of several maturity markers (Fig. S3B). Instead, we found fewer docked insulin granules at basal state in the Commd3 mutant cells (Figure 5AeD). This contrasts with granuphilin mutant beta cells where a decrease of docked granules slightly enhances basal insulin secretion [74]. The difference is likely due to the dual roles of granuphilin in docking and exocytosis [74,75]. COMMD3 probably does not play a direct role in either docking or exocytosis. The normal glucose-stimulated insulin secretion in Commd3 mutant cells indicates it does not affect glucose-stimulated granule docking nor exocytosis [76]. COMMD3 has been reported to regulate ITGB1 recycling and deficiency of COMMD3 causes ITGB1 accumulation in WASH complex in HeLa cells [48]. We found in Commd3 mutant MIN6 cells, ITGB1 accumulated in Rab4 vesicles at basal glucose ( Figure 6A). This was recapitulated by expressing a recycling-deficient ITGB1 in Commd3-intact MIN6 cells. Expression of the recycling-deficient ITGB1 also reduced insulin secretion and insulin granule docking at basal condition but not at high glucose stimulation. ITGB1 can be recycled via different pathways depending on its partner, activation state, and cellular physiology [77]. Our results suggest that Commander-mediated recycling from Rab4-positive endosomes is the predominant pathway in beta cells at basal glucose conditions and that high glucose activates other ITGB1 recycling pathways. The role of ITGB1 in basal insulin secretion is not clear. In two studies from the Wang lab on mice with beta cell specific loss of Itgb1, it was initially shown that the male mutant mice have normal fasting glucose levels but glucose intolerance, while female mutants are normal in fasting glucose and glucose tolerance [17]. In a follow up study, however, it was shown that primary islets from males, but not female, have decreased basal insulin secretion [58], similar to our results in MIN6 cells. The discrepancy may be due to the more complex regulation of insulin secretion in vivo. Integrin mediates cell adhesion through membrane associated focal adhesions that connect to actin and the microtubule cytoskeleton [78]. As expected, impairment of ITGB1 recycling decreases membrane ITGB1 clusters and focal adhesions at basal glucose conditions (Figs. 8 and S5). Several studies have found a role for focal adhesions in GSIS [63,79,80]. Yet, how reduced focal adhesions leads to a reduction of basal insulin secretion is unknown. Genetic or pharmacological inhibition of focal adhesion kinase (FAK) reduces insulin granule docking without affecting basal insulin secretion [80,81]. It is possible that the focal adhesion effect on basal insulin secretion is FAK independent. How a reduced number of focal adhesions decreases insulin granules docking is also unknown. In the case of FAK deficiency, the diminishment of insulin granule docking is thought to be a result of impaired actin depolymerization [80,81]. In Commd3 deficient MIN6 cells, our results suggest a decreased number and size of cortical ELKScontaining complexes as a potential mechanism. In high glucose condition, ELKS promotes exocytosis by anchoring insulin granules to the vicinity of VDCC [23]. It does not appear to occur at basal glucose conditions as no change of granule release probability was observed in Commd3 KD cells (Figure 5EeF). Other described mechanisms of ELKS action may operate at basal glucose. The ELKS-containing complexes may promote the delivery of insulin granules by anchoring CMSCs to the focal adhesions [67], target the granules to the exocytic complexes through LL5b-Kank1 interaction with focal adhesions [5], and/or facilitate the docking of insulin granules [22]. Although the role of ELKS in GSIS has been well established [65], its role in basal insulin secretion needs further investigations. In summary, in a screen to identify genes that affect the glucose induction index of insulin secretion in beta cells, we identify a role for the Commander complex in enhancing basal insulin secretion. Mechanistically, we show that the Commander complex promotes docking of insulin granules at basal state. This is mediated, at least in part, by its function in ITGB1 recycling. Defects in ITGB1 recycling reduce its membrane distribution and the number of focal adhesions. Detailed molecular mechanisms connecting focal adhesions to insulin granule docking and secretion at basal glucose conditions remain to be elucidated.