C2CD4A and C2CD4B expression in mouse and human islets
Human C2CD4A and C2CD4B are 83% homologous. Like their murine homologues, the two human genes are predicted to have evolved from a common ancestor (see phylogenetic tree, ESM Fig. 2a). Conservation of genomic architecture (synteny) at this locus argues for direct homology between the human and murine forms of each gene. The zebrafish (D. rerio) possesses two C2cd4-like genes homologous to H. sapiens and Mus musculus: C2cd4a and C2cd4c (ESM Fig. 2a). Scrutiny of gene expression databases (http://Biogps.org, accessed March 2020), and previous publications [33, 34], reveals approximately tenfold higher levels of C2cd4b than C2cd4a mRNA in mouse islets and purified mouse beta cells (ESM Table 1). In contrast, roughly equal levels of C2CD4A and C2CD4B mRNA are present in human islets  and purified beta cells . Expression of both genes was detected in the pituitary in both human and mice (data source: GTEx data, analysed in The Human Protein Atlas; www.proteinatlas.org/ENSG00000198535-C2CD4A/tissue and www.proteinatlas.org/ENSG00000205502-C2CD4B/tissue; accessed November 2020). C2cd4a and C2cd4b were both upregulated in pancreatic islets of high-fat-diet--fed DBA2J mouse model of diabetes following 30 and 90 days of the diet vs RC-fed mice (ESM Fig. 3a,b) . No evidence of C2CD4A or C2CD4B upregulation was observed in the islets of individuals with type 2 diabetes vs normoglycaemic control participants .
Examination of the human VPS13C/C2CD4A/C2CD4B locus revealed multiple regulatory elements (Fig. 1), consistent with recent findings . A SNP, rs7163757, has recently been shown by fine mapping as the likely causal variant for type 2 diabetes risk at this locus . Correspondingly, CRISPR activation (CRISPRa) or CRISPR interference (CRISPRi) at the rs7163757 site has the most marked effects on neighbouring genes , consistent with this being the effector variant (SNP). To explore in more detail the potential role of the enhancer around the diabetes risk variant rs7163757  and whether it may play a role in gene expression in disease-relevant tissues (notably the islet and brain/pituitary), we introduced a reporter bearing 1303 bp nucleotides of the human sequence into the zebrafish genome, controlling the production of GFP from a minimal cFos promoter (ESM Fig. 2b). Expression was restricted to the endocrine pancreas and brain at all stages (ESM Fig. 2c,e), and was detected in all islet cell types, most strongly in delta cells (ESM Fig. 2d). No significant differences were seen in the pattern of expression in embryos expressing either the C- or the T variants, with 10% (8/80 or 10/100, respectively) of embryos showing fluorescence in the pancreas.
Whole-mount fluorescent in situ hybridisation of 30 h-post-fertilisation (hpf) embryos revealed that endogenous c2cd4a is expressed in the forebrain (ESM Fig. 2f), ventral spinal cord (ESM Fig. 2g) and pancreas (ESM Fig. 2c,h). Double fluorescent in situ hybridisation revealed presence of the c2cd4a transcript in sst2+, ins+ and gcgb+ cells (ESM Fig. 2d,i–k).
Role of c2cd4a in the zebrafish larvae
Zebrafish possess a single gene, C2cd4a (formerly c2cd4ab), that is homologous to the two mammalian counterparts (ESM Fig. 2a). As a convenient proxy for insulin secretion in the larvae of fish inactivated for c2cd4a or in WT controls (ESM Fig.4 a,b), glucose-stimulated Ca2+ dynamics were monitored in vivo by imaging the fluorescence of a gCaMP6 transgene. Fasting and postprandial blood glucose levels were comparable between WT and c2cd4a mutant animals (ESM Fig. 4c) and glucose-induced Ca2+ changes did not differ between genotypes (ESM Fig. 4d–h). These findings argue against a role for c2cd4a in beta cell function during zebrafish development.
Effects of C2cd4b deletion on glucose homeostasis are more marked in female than male mice
Given the substantially higher expression of C2cd4b than C2cd4a in mouse islets (ESM Table 1), we studied mice in which the former gene was deleted (Fig. 2a). Inter-crossing of heterozygous animals produced pups at the expected Mendelian ratio (Fig. 2b) and resulted in complete elimination of C2cd4b mRNA from isolated islets of C2cd4b−/− mice (Fig. 2c). Female C2cd4b-null mice gained weight at the same rate as control littermates whether maintained on RC or a HFD (Fig. 2d) and no differences were apparent in fed or fasting blood glucose levels either on an RC diet (ESM Fig. 5a,c) or an HFD (Fig. 2f and ESM Fig. 5e).
Beta cell mass, as assessed by histochemical analysis of pancreatic slices, was unaffected by deletion of C2cd4b in both females (ESM Fig. 6a–d) and males (ESM Fig. 6e–h).
Intraperitoneal glucose tolerance was examined for animals maintained on an RC diet or HFD from 8 to 22 weeks of age. Females displayed abnormalities at 12, 20 and 22 weeks of age when fed an RC diet (ESM Fig. 7a,c,e,g), and at 8 and 23 weeks on an HFD (ESM Fig. 8a,c,e and Fig. 3a; p < 0.001 at 22 weeks on an HFD) vs control littermates. While oral glucose tolerance was normal on in RC-fed female mice vs controls (Fig. 3c), defects were observed in HFD-fed mice (Fig. 3g). For female mice on the RC diet, these changes were associated with defective insulin secretion in vivo (Fig. 4a; p = 0.02), as also indicated by unchanged insulin levels after glucose injection despite elevated plasma glucose levels in female knockout mice maintained on an HFD (Fig. 4c). Furthermore, insulin sensitivity was not different between C2cd4b null and WT female mice (ESM Fig. 9a,c).
When maintained on an RC diet, male C2cd4b null mice gained weight at the same rate as WT littermates. When maintained on an HFD, in contrast to females, male mutant mice gained substantially more weight from 14 weeks of age vs WT littermates (Fig. 2d,e; p < 0.001 betwee18 and 21 weeks of age), and had raised fasting blood glucose levels (Fig. 2g; p = 0.03). However, significant differences in body fat and lean mass were not apparent (Fig. 2h,i) vs WT mice. Intraperitoneal glucose tolerance was normal in males maintained on an RC diet at most ages examined, with genotype-dependent differences only at 16 weeks (ESM Fig. 7b,d,f,h and Fig. 3b). Although unaltered in younger male mice after maintenance on an HFD (ESM Fig. 8b,d,f), glucose tolerance was impaired at 23 weeks of age by C2cd4b deletion (Fig. 3f,h) as compared with controls.
As observed in females, insulin sensitivity (ESM Fig. 9b,d) was unaltered in male C2cd4b-null mice vs littermate controls.
Effects of C2cd4b deletion on beta cell function in vitro
Glucose-stimulated insulin secretion was not different between islets from WT or C2cd4b-null female mice maintained on an RC diet (ESM Fig. 10a) but was slightly elevated in those from female null mice maintained on an HFD (ESM Fig. 10c). Likewise, in the isolated islets from female C2cd4b-null mice, we observed no alterations in glucose or KCl-stimulated stimulated Ca2+ dynamics (ESM Fig. 11a,b) or in beta cell–beta cell coupling (ESM Fig. 11c–f). Correspondingly, no changes in voltage-activated Ca2+ currents were apparent in patch-clamp recordings (ESM Fig. 12a–c).
In line with the above findings, massive parallel sequencing (RNA-Seq) of islets from C2cd4b−null or control animals confirmed the lowering of C2cd4b expression in islets of null mice vs WT mice, and revealed a significant (~75%) increase in C2cd4a expression (ESM Table 2), though this was not confirmed by independent qRT-PCR analysis (data not shown). However, no other mRNAs were significantly affected by C2cd4b deletion (ESM Table 2).
Glucose or KCl-stimulated insulin secretion from isolated islets were also unaltered in male C2cd4b-null mice vs littermate controls (ESM Fig. 10 b,d).
Effects of C2cd4b deletion on pituitary function
Given the absence of clear defects in insulin secretion in isolated C2cd4b-null islets, and the expression of both C2cd4a and C2cd4b in the pituitary (data source: GTEx data, analysed in The Human Protein Atlas; www.proteinatlas.org/ENSG00000198535-C2CD4A/tissue and www.proteinatlas.org/ENSG00000205502-C2CD4B/tissue; Fig. 5a), we next assessed whether deletion of this gene might affect the production of sex hormones and, thus, provoke gender-specific differences in glucose homeostasis. E2 and testosterone levels were unaltered in both male and female C2cd4b-null mice maintained on an RC diet or HFD vs WT animals (Fig. 5b,c). In order to remove the negative feedback loop through which hormones secreted from the gonads repress FSH and LH release from the pituitary gland, animals between 12 and 15 weeks of age were gonadectomised prior to these experiments. Compared with WT littermates, female C2cd4b-null mice displayed ~50% lower circulating FSH levels, with no difference in LH levels (Fig. 5d,f; p = 0.003). No differences in LH or FSH levels were apparent between WT and C2cd4b-null male mice (Fig. 5e,g).
C2cd4a-null mice display no metabolic abnormalities
To determine whether C2cd4a inactivation might also have an impact on glucose homeostasis, we next examined metabolic phenotypes in male and female C2cd4a-null mice (ESM Fig. 13a–c). In contrast to their C2cd4b-null counterparts, C2cd4a-null mice displayed no evident metabolic abnormalities up to 22 weeks of age, with neither weight gain, glucose homeostasis nor insulin secretion differing between WT and null mice for either sex (ESM Fig. 13d–i). Likewise, as measured in vitro, glucose and KCl (depolarisation)-stimulated insulin secretion were unaltered in islets isolated from male C2cd4a-null mice (ESM Fig. 13j).
C2CD4A but not C2CD4B C2 domains support Ca2+-dependent intracellular translocation
We next sought to explore the mechanism(s) through which C2CD4B or C2CD4A may influence beta cell (and, potentially, pituitary gonadotroph) function. Both proteins have been suggested to lack a functional C2 domain , consistent with a reported localisation in the nucleus in COS7 cells . In contrast to earlier findings reporting nuclear subcellular localisation, when overexpressed as GFP- or FLAG- tagged chimaeras in rodent (MIN6, INS1[832/13]) or human (EndoCβH1) beta cell lines, C2CD4A and C2CD4B were found at the plasma membrane and in the cytosol and the nucleus (ESM Fig.14 and ESM Fig.15). In the majority of cells, C2CD4A and C2CD4B were primarily localised to the cytoplasm and nucleus. Co-localisation with readily identified intracellular sub-compartments, including the secretory granule (insulin), trans-Golgi network (TGN46), endosome/lysosome (LAMP1) or endoplasmic reticulum (ER; KDEL), was not apparent in the above cell lines (ESM Fig. 16; data only shown for MIN6 cells but results were consistent in all three cell lines).
The above findings suggested that the C2 domain of either protein may bind to Ca2+ and contribute to localisation at, and/or shuttling between, subcellular compartments in living cells. To test this hypothesis, we explored phospholipid-dependent recruitment of these proteins to the plasma membrane in INS1(832/13) beta cells  expressing either a control construct, in which Syt1 (bearing five C2 domains) was fused to GFP  or equivalent C2CD4A or C2CD4B constructs (N-terminal linkage). In response to an increase in intracellular free Ca2+, provoked by 50 μmol/l extracellular Ca2+ and the calcium ionophore ionomycin (50 ng/ml), Syt1–GFP translocated from intracellular (likely ER-bound) sites to the plasma membrane. This movement was readily visualised by simultaneous live-cell wide-field and total internal reflection of fluorescence (TIRF) imaging (Fig. 6a,b,d). A similar, but smaller change in the localisation of C2CD4A–GFP in response to Ca2+ was also observed. In contrast, no response was detected for C2CD4B-GFP (Fig. 6c,e–g).
Identification of C2CD4A and C2CD4B binding partners by mass spectrometry
The above experiments demonstrated that C2CD4A, and possibly C2CD4B, may participate in Ca2+-dependent signal transduction. To gain further insight into possible mechanisms of action, we performed an unbiased proteomic screen using immunoprecipitation and mass spectrometry to identify potential binding partners. Normalising to the negative control, and ranking in order of protein abundance, we generated lists of possible interacting proteins for human C2CD4A, C2CD4B (ESM Tables 3,4) or both (ESM Table 5). MIN6 cells transfected with human C2CD4A or C2CD4B were used for this analysis, given the low transfection efficiency of human-derived EndoCβH1 cells. Interacting partners included proteins involved in Ca2+ binding (torsin-2A [TOR2A] and EF-hand calcium-binding domain-containing protein 5 [EFCAB5;  and www.genecards.org/), NF-κB signalling (sequestosome-1 [SQSTM1] and programmed cell death protein 11 [PDCD11]) and protein trafficking (proprotein convertase subtilisin/kexin type 9 [PCSK9], neural precursor cell expressed, developmentally down-regulated 4, E3 ubiquitin protein ligase [NEDD4], Ras-proximate-1 [Rap1] and GTPase-activating protein [GAP2]). Receptor-type tyrosine-protein phosphatase-like N (PTPRN; insulinoma-associated protein 2 [IA-2]) and PTPRN2; phogrin/IA-2β/islet cell antigen 512 [ICA512]) bound to both C2CD4A and C2CD4B. These protein tyrosine phosphatase-like transmembrane proteins are granule-resident and implicated in granule trafficking and exocytosis .