Sexually dimorphic roles for the type 2 diabetes-associated C2cd4b gene in murine glucose homeostasis

Aims/hypothesis Variants close to the VPS13C/C2CD4A/C2CD4B locus are associated with altered risk of type 2 diabetes in genome-wide association studies. While previous functional work has suggested roles for VPS13C and C2CD4A in disease development, none has explored the role of C2CD4B. Methods CRISPR/Cas9-induced global C2cd4b-knockout mice and zebrafish larvae with c2cd4a deletion were used to study the role of this gene in glucose homeostasis. C2 calcium dependent domain containing protein (C2CD)4A and C2CD4B constructs tagged with FLAG or green fluorescent protein were generated to investigate subcellular dynamics using confocal or near-field microscopy and to identify interacting partners by mass spectrometry. Results Systemic inactivation of C2cd4b in mice led to marked, but highly sexually dimorphic changes in body weight and glucose homeostasis. Female C2cd4b mice displayed unchanged body weight compared with control littermates, but abnormal glucose tolerance (AUC, p = 0.01) and defective in vivo, but not in vitro, insulin secretion (p = 0.02). This was associated with a marked decrease in follicle-stimulating hormone levels as compared with wild-type (WT) littermates (p = 0.003). In sharp contrast, male C2cd4b null mice displayed essentially normal glucose tolerance but an increase in body weight (p < 0.001) and fasting blood glucose (p = 0.003) after maintenance on a high-fat and -sucrose diet vs WT littermates. No metabolic disturbances were observed after global inactivation of C2cd4a in mice, or in pancreatic beta cell function at larval stages in C2cd4a null zebrafish. Fasting blood glucose levels were also unaltered in adult C2cd4a-null fish. C2CD4B and C2CD4A were partially localised to the plasma membrane, with the latter under the control of intracellular Ca2+. Binding partners for both included secretory-granule-localised PTPRN2/phogrin. Conclusions/interpretation Our studies suggest that C2cd4b may act centrally in the pituitary to influence sex-dependent circuits that control pancreatic beta cell function and glucose tolerance in rodents. However, the absence of sexual dimorphism in the impact of diabetes risk variants argues for additional roles for C2CD4A or VPS13C in the control of glucose homeostasis in humans. Data availability The datasets generated and/or analysed during the current study are available in the Biorxiv repository (www.biorxiv.org/content/10.1101/2020.05.18.099200v1). RNA-Seq (GSE152576) and proteomics (PXD021597) data have been deposited to GEO (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE152576) and ProteomeXchange (www.ebi.ac.uk/pride/archive/projects/PXD021597) repositories, respectively. Graphical abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-020-05350-x.


Introduction
Type 2 diabetes risk is the product of both environmental and genetic factors. More than 200 loci have now been described as affecting the diabetes risk score [1]. While most of these impact insulin secretion [2], the identified variants usually lie within or between neighbouring genes and in only a few cases have the causal gene(s) been firmly established [3,4].
Chromosome 15q hosts a risk locus close to the VPS13C, C2CD4A and C2CD4B genes [5], which is associated with impaired proinsulin processing. Deletion of Vps13c (which encodes a lipid transporter [6]) selectively from the pancreatic beta cell [7] has little effect on glucose homeostasis in the mouse. This finding argues that the other genes at this locus contribute towards the effect(s) of risk variants in humans.
Previous expression quantitative trait locus (eQTL) studies have demonstrated altered expression of VPS13C and C2CD4A [7], as well as C2CD4B [8], in islets from individuals carrying risk alleles. Recently, Kycia et al [9] reported that expression of C2CD4B, but not C2CD4A or VPS13C, was affected by risk alleles. However, the direction of effects of risk alleles differed between the reports, with expression lowered uniquely in females in the study of Mehta et al [7] but increased in the study by Kycia et al [9]. However, both of these studies involved relatively small sample numbers. These limitations emphasise the need for interventional studies involving gene inactivation in tractable model systems, such as rodents or fish, as an alternative means of understanding the roles of these genes in metabolic homeostasis.
C2CD4A and C2CD4B (also called NLF1 and NLF2) [10] encode low molecular mass (39 kDa) proteins of presently unknown function. Unlike the homologous C2CD4C gene (expressed from a distinct locus on chromosome 19 in Homo sapiens), neither C2 calcium dependent domain containing protein (C2CD)4A nor C2CD4B possess a canonical Ca 2+ /phospholipid-binding C2-domain [11]. A partly functional Ca 2+binding domain may be present in C2CD4B (see electronic supplementary material [ESM] Fig.1). Given the essential role for Ca 2+ in the control of insulin and other hormone secretion [12], an interaction with Ca 2+ might provide a means through which C2CD4A or C2CD4B influence these processes.
Originally described in endothelial cells as having a largely nuclear distribution, and inducible by cytokines [10], the role of C2CD4B has not been explored previously in beta cells or other metabolically relevant cell types. Nonetheless, silencing of the single homologous C2cd4a/b gene in the zebrafish (Danio rerio) led to a decrease in beta cell mass [13]. In contrast, silencing of the Drosophila melanogaster C2CD4A homologue, spenito (also known as nito), increased circulating levels of the insulin-like molecule IIp2HF [14]. Inactivation of C2cd4c has no detectable effect on pancreatic development in mice [11].
A recent study [15] has indicated that regulation of C2cd4a expression in islets by forkhead box O1 (FOXO1) may be important for the control of insulin secretion. Thus, animals inactivated selectively in the beta cell for C2cd4a displayed abnormal insulin secretion in response to exposure to glucose plus arginine (the effects on secretion stimulated by glucose alone were not reported), as well as glucose intolerance in vivo and abnormal expression of beta cell signature and 'disallowed' genes [16]. However, these studies used the Ins2-dependent rat insulin promoter (RIP) to drive Cre expression, a strategy complicated by off-target recombination and ectopic expression of human growth hormone [17].
Here, we use the more direct approach of inactivating C2cd4b and C2cd4a globally in the mouse, and of deleting their homologue, C2cd4a, in the developing zebrafish (D. rerio), to explore the role of these genes in glucose homeostasis.

Methods
For detailed Methods and the reagents list, please refer to the ESM Methods.
Glucose homeostasis Animals were fasted overnight prior to experiments. For IPGTTs, glucose (1 g/kg body weight) was injected into the abdomen. In OGTTs, glucose (2 g/kg body weight) was administered directly into the gut via oral gavage. Blood glucose levels were recorded using an automatic glucometer (Accucheck; Roche Diagnostics, Burgess Hill, UK). For insulin tolerance tests, animals were fasted for 5 h prior to experiments. Insulin was injected into the abdomen: Male and female mice (both C2cd4b-null and WT) on the regular chow (RC) diet were injected with a 1 or 0.75 U/kg body weight of insulin, respectively. HFD-fed male and female mice were injected with 1.5 or 0.75 U/kg body weight of insulin, respectively. Blood glucose levels were measured post injection at the time points indicated. To measure insulin secretion in vivo, animals were fasted overnight and glucose (3 g/kg body weight) was injected into the abdomen. Blood insulin levels were measured using an Ultra-Sensitive Mouse Insulin ELISA Kit (90,080; Crystal Chem, Zaandam, the Netherlands).
Insulin secretion from isolated islets Islet isolation [19] and measurements of insulin secretion [20] were measured as described previously, using ten size-matched islets incubated in triplicate in HEPES-buffered Krebs-Ringer medium (KREBH) and either 3 mmol/l glucose, 17 mmol/l glucose or 20 mmol/l KCl at 37°C. Insulin was measured using an Insulin Ultra-Sensitive Kit (Mercodia, Uppsala, Sweden).
Generation of C2CD4A and C2CD4B FLAG-and -green fluorescent protein-tagged constructs Human C2CD4A and C2CD4B cDNA sequences were cloned in-frame into plasmid P3XFLAG-CMV-14 (Addgene, www.addgene.org) to provide a C-terminal 3×FLAG epitope tag. Green fluorescent protein (GFP)-tagged proteins were generated by inserting the human C2CD4A and C2CD4B cDNA sequences into the Cterminus of GFP using plasmid pEGFP-C1 (Addgene).
Study approval All mouse in vivo procedures were conducted in accordance with the UK Home Office Animal (Scientific Procedures) Act of 1986 (Project licence PA03F7F0F to IL) and approved by the Imperial College Animal Welfare and Ethical Review Body. All zebrafish work was approved by the ethical committee of the University of Liège (protocol no.  or under European Union and German laws (Tierschutzgesetz), and with the approval of the TU Dresden and the Landesdirektion Sachsen (approval licence number: TVV 45/2018).
Zebrafish maintenance and generation of transgenic lines Zebrafish (D. rerio) were obtained from the Zebrafish International Resource Center (Zfin, Eugene, OR, USA) or from the European Zebrafish Resource Center at Karlsruhe Institute of Technology (www.ezrc.kit.edu/), and then bred at the host laboratory (University of Liège [ULg], Liège, Belgium, or Universität Dresden, Dresden, Germany, respectively). Genetically modified zebrafish were developed at the host laboratories. Zebrafish were raised and cared for according to standard protocols [22]. The transgene rs7163757C-cfos:eGFP was constructed by introducing a 1303 bp region carrying the genome-wide association study (GWAS)-identified gene variant rs7163757-C upstream of a c-fos minimal promoter driving enhanced GFP (EGFP) expression (pGW_cfos-EGFP) [23,24] via Gateway LR recombination (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). The transgene was injected into zebrafish embryos and GFP expression patterns were analysed during development, in adulthood and in offspring, using a confocal microscope.
Whole-mount in situ hybridisation and immunohistochemistry on zebrafish embryos Double fluorescent whole-mount in situ hybridisations were performed as previously described [25] with the antisense RNA probe for the different genes prepared as described [26]. Immunohistochemistry on whole-mount embryos was performed as previously described [27]. The proteins analysed were GFP, insulin, glucagon and somatostatin. Images were acquired with a confocal microscope and processed with ImageJ (https://imagej.nih.gov/ij/; accessed December 2019) and figureJ (https://imagej.net/ FigureJ).
Pericardial glucose injection and live imaging in zebrafish larvae Glucose injection and live imaging of zebrafish primary islets were performed as described previously [28].
Blood glucose measurements in adult zebrafish Adult zebrafish were fasted for 24 h before glucose measurements. For post-prandial measurements, fasted fish were fed with live brine shrimp. Blood was collected from euthanised fish with a microcapillary needle and glucose levels were measured using a glucometer.
Immunofluorescence analysis of pancreatic slices Mouse pancreases were dissected at 24 or 25 weeks of age. Pancreases were fixed in 4% (wt/vol.) paraformaldehyde and embedded in paraffin for the measurement of insulin and glucagon via immunofluorescence.
Homogeneous time-resolved fluorescence (HTRF) assay The Insulin Ultra-Sensitive Kit (ref. 62IN2PEH; Cisbio, Codolet, France) was used according to the manufacturer's instructions to measure released or total insulin levels in mouse samples. Each sample was measured in duplicate and incubated with europium cryptate and XL665 antibodies overnight before measuring the Fӧrster resonance energy transfer (FRET) efficiency.
Whole-cell voltage-clamp electrophysiology Measurements were performed on single beta cells isolated from wild-type (WT) and C2cd4b-knockout mice. Starting from a holding potential of −80 mV, and voltage-dependent Ca 2+ currents (VDCC) currents were generated through application of sequential 10 mV depolarising steps ranging from −70 mV to 70 mV (500 ms).
Sample preparation for RNA sequencing Islets from five male mice/genotype at 22 weeks of age on regular chow diet were used for RNA purification. Generation of doublestranded cDNA and library construction were performed using NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB, Hitchin, UK). Sequencing was performed by the Imperial BRC Genomics Facility (Imperial College London, UK) as 75 bp paired end reads on a HiSeq4000 according to Illumina specifications. Data are archived in the GEO repository (www. ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE152576).
Quantitative reverse transcription PCR Quantitative reverse transcription PCR (RT-qPCR) was performed in triplicate using SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) and primers for mouse C2cd4a and βactin genes.
Immunoprecipitation and mass spectrometry Mycoplasmafree MIN6 cells grown in standard culture conditions [31] were transfected in duplicate with FLAG-tagged-C2CD4A or -C2CD4B expressing plasmids or with a FLAG tag-only expressing plasmid using Lipofectamine 2000 (Thermo Fisher Scientific). Immunoprecipitation and affinity purification-MS (AP-MS) analysis were performed, as previously described [32], with minor modifications. The mass spectrometry proteomics data have been deposited to the ProteomeXchange repository (www.ebi.ac.uk/pride/archive/projects/ PXD021597).
Statistical analysis Blinding was carried out for all mouse in vivo experiments. Data were analysed using GraphPad Prism 8.0 (San Diego, CA, USA; www.graphpad.com/ scientific-software/prism/, accessed 1 December 2019). A pvalue <0.05 was considered significant.

Results
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 [35] and purified beta cells [36]. 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 highfat-diet-fed DBA2J mouse model of diabetes following 30 and 90 days of the diet vs RC-fed mice (ESM Fig. 3a,b) [37]. No evidence of C2CD4A or C2CD4B upregulation was observed in the islets of individuals with type 2 diabetes vs normoglycaemic control participants [38].
Examination of the human VPS13C/C2CD4A/C2CD4B locus revealed multiple regulatory elements ( Fig. 1), consistent with recent findings [9]. A SNP, rs7163757, has recently been shown by fine mapping as the likely causal variant for type 2 diabetes risk at this locus [9]. Correspondingly, CRISPR activation (CRISPRa) or CRISPR interference (CRISPRi) at the rs7163757 site has the most marked effects on neighbouring genes [39], 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 [9] 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 Cor the T variants, with 10% (8/80 or 10/100, respectively) of embryos showing fluorescence in the pancreas.
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 Ca 2+ 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 Ca 2+ 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 C2cd4bnull 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) [39] 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 Ca 2+ dynamics (ESM Fig. 11a,b) or in beta cell-beta cell coupling (ESM Fig. 11c-f). Correspondingly, no changes in voltage-activated Ca 2+ 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 genderspecific differences in glucose homeostasis. E2 and testosterone levels were unaltered in both male and female C2cd4bnull 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 C2cd4bnull 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, C2cd4anull 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 C2cd4anull mice (ESM Fig. 13j).
C2CD4A but not C2CD4B C2 domains support Ca 2+ -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 [11], consistent with a reported localisation in the nucleus in COS7 cells [10]. 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 Ca 2+ 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 [21] expressing either a control construct, in which Syt1 (bearing five C2 domains) was fused to GFP [40] or equivalent C2CD4A or C2CD4B constructs (N-terminal linkage). In response to an increase in intracellular free Ca 2+ , provoked by 50 μmol/l extracellular Ca 2+ 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 Ca 2+ 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 Ca 2+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 Ca  These protein tyrosine phosphatase-like transmembrane proteins are granule-resident and implicated in granule trafficking and exocytosis [42].

Discussion
The overall aim of this study was to examine the biological roles of C2cd4b and C2cd4a in vivo, focusing on the pancreatic beta cell and pituitary gland. These questions have been pursued using mouse and fish knockout models and relevant cell lines. In contrast to earlier findings [15], we observed that global C2cd4a deletion in the mouse exerted no effects on insulin secretion in vitro or in vivo. These findings are consistent with the considerably (tenfold) lower expression of C2cd4a than C2cd4b in mouse islets and mouse-derived beta cell lines (ESM Table 1), though we should emphasise that no attempt was made here to quantify protein (rather than mRNA) levels. The reasons for the differences between the present study and that of Kuo et al [15] with respect to C2cd4a are presently unclear.
Deletion of C2cd4b leads to weight gain in males, but defective insulin secretion in female mice Our findings reveal a strikingly sexually dimorphic phenotype of C2cd4b-null mice, which is also strongly dependent upon diet. In contrast to females, male C2cd4b null mice displayed no evident metabolic abnormalities at most ages when maintained on an RC diet. Although mutant male mice exhibited a substantial increase in body weight on the HFD, fasting glucose and glucose tolerance were impaired only in older (>20 weeks) animals. In stark contrast, female mutant mice displayed abnormal glucose tolerance from as early as 12 weeks when fed an RC diet and 8 weeks when fed an HFD (with a tendency towards abnormal glucose tolerance [p > 0.05] being observed at 8 weeks in RC-fed mice), despite unaltered body weight. Non-significant changes in lean mass were observed in male C2cd4b-null mice on HFD, which are reminiscent of findings from the IMPC (www.mousephenotype.org/data/ genes/MGI:1922947, accessed November 2020) when this line was fed an RC diet. The mechanisms underlying these changes remain unclear but are unlikely to involve elevated testosterone levels in knockout animals, since these were not observed in our study.
Although earlier human studies were not stratified by sex, we note that a statistically significant (p = 0.004) impact of the human type 2 diabetes-associated variant rs7172432 (which is in perfect linkage disequilibrium with rs7163757: R 2 = 1.0) [15] at the VPS13C/C2CD4A/C2CD4B locus on waist circumference has previously been reported [43], and may be related to the alterations in body weight we observed here in male C2cd4b-knockout mice.
What factors underlie these sex-dependent differences in insulin secretion between WT and C2cd4b-null mice? Our findings suggest that beta cell-extrinsic mechanisms, possibly involving circulating factors, contribute to (and may even be the drivers of) altered insulin secretion in the living animal. Potential contributors are changes in FSH levels [44,45] in female C2cd4b-null mice, reflecting altered expression of the gene in the pituitary. Decreased FSH production is expected, in turn, to decrease circulating oestrogen levels. We note, however, that measurements of oestrogen are complicated in fertile mice due to fluctuations in the oestrous cycle [46]. Positive actions of oestrogen on beta cell insulin content (via oestrogen receptor α) and secretion (via oestrogen receptor β) are well known [44,47] and may thus underlie the weaker insulin secretion in C2cd4b-knockout mice. Consistent with a requirement for sexual maturity, no differences were apparent in beta cell function between genotypes in the living fish embryo (ESM Fig.4) at stages where differences in circulating sex hormones are not anticipated, and fasting blood glucose levels did not differ between WT and knockout fish at adult stages. These findings are in contrast to earlier studies in zebrafish larvae using oligonucleotidemediated C2cd4a gene knockdown [13], which reported alterations in beta cell mass.
Such sexual dimorphism on the impact of variants at this locus, however, has not been reported in human GWAS data [48,49]. One possible explanation is that dimorphism in mice reflects well-known differences in the response of males and females to HFD [50]. More likely, in our view, is that in rodent islets (and pituitary; see below), C2cd4b expression predominates over C2cd4a, while in humans, C2CD4A and C2CD4B expression are comparable. Consequently, in humans, changes in both C2CD4A and C2CD4B may mediate the effect of the GWAS signal for type 2 diabetes, dampening a sexually dimorphic effect of changes in C2CD4B expression.
Might altered genetic risk for type 2 diabetes in humans result from changes in C2CD4A or C2CD4B expression in the brain? Importantly, high levels of expression of C2CD4A and C2CD4B in the pituitary (GTEx data, The Human Protein Atlas; www.proteinatlas.org/ENSG00000198535-C2CD4A/ tissue and www.proteinatlas.org/ENSG00000205502-C2CD4B/tissue; accessed November 2020) are consistent with this view. Correspondingly, an eQTL for C2CD4A is reported for rs7163757 in the pituitary (www.gtexportal.org/ home/snp/rs7163757#sqtl-block, accessed November 2020), with a (non-significant) tendency towards decreased expression of C2CD4A and, more strongly, VPS13C in carriers of C (risk allele) vs T alleles.

Intracellular signalling by C2cd4a and C2cd4b
Our observation that neither C2CD4A nor C2CD4B are localised exclusively in the nucleus, as previously reported in COS7 cells [10], islets and MIN6 cells [15], was unexpected, but implies a more dynamic role for both C2CD4A, and possibly C2CD4B, in intracellular signalling. We note that the present study explored the subcellular distribution of the H. sapiens protein, rather than the M. musculus homologue examined previously [15], providing a potential explanation for these differences. Interestingly, predictions from the primary structure [11] indicate that neither C2CD4A nor C2CD4B (human or mouse) possesses a C2 domain with a bona fide Ca 2+ -binding site [51,52]. Nevertheless, we provide a direct demonstration of Ca 2+ -dependent recruitment of C2CD4A to the plasma membrane (Fig. 6), whereas C2CD4B would appear to exert its function independently of Ca 2+ binding. What signalling mechanism(s) may lie downstream of C2CD4A (or C2CD4B)? The interacting protein PTPRN2 (also known as phogrin and IA-2β) is of particular interest [42]. Inactivation of PTPRN2/phogrin, a secretory granulelocalised protein [53], leads to defective insulin secretion in mice [54,55]. Importantly, double knockout of PTPRN2/ phogrin and the closely-related Ptprn (IA-2) gene leads to defective FSH and LH production and female infertility [54], implying an important role in the anterior pituitary. An interaction between C2CD4A or C2CD4B and PTPRN2 in the pituitary (and possibly the beta cell) may, therefore, contribute to the effects of altered expression on diabetes risk.
Limitations of the study The present work was undertaken using animals on a mixed C57BL6N:C57BL6J background and future studies are required to confirm the findings here in pure-bred strains [56]. The impact of C2cd4b deletion on fertility, and on the role of C2cd4a and C2cd4b in the pituitary, remain to be explored, as does the contribution of changes in food intake or energy expenditure to differences in body weight. More work will also be necessary to determine whether c2cd4a-null zebrafish display beta cell deficiency.
Restrictions imposed by the coronavirus disease-2019 (COVID-19) pandemic prevented further explorations of these points during the present study.

Supplementary Information
The online version contains peer-reviewed but unedited supplementary material available at https://doi.org/10.1007/ s00125-020-05350-x. Authors' relationships and activities GAR has received grant funding and consultancy fees from Sun Pharmaceuticals and Les Laboratoires Servier. All other authors declare that there are no relationships or activities that might bias, or be perceived to bias, their work.  The ribbon diagram reproduced from [57] under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons. org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium. (b) Localisation of GFP-Syt1 in cells incubated in KREBH buffer with 3 mmol/l glucose, shown before and after an increase in intracellular Ca 2+ levels (achieved by addition of 50 ng/ml ionomycin), obtained by simultaneous wide-field and TIRF image acquisition (positive control). (c) Localisation of GFP-tagged C2CD4A and -C2CD4B before and after addition of ionomycin, obtained by simultaneous wide-field and TIRF image acquisition. (d, f) Time courses for the translocation of the GFP-tagged proteins, Syt1 (d), C2CD4A (e) or C2CD4B (f) obtained by TIRF imaging before and after addition of ionomycin (iono). (d-f) Solid black lines represent mean; SEM shown by grey bars. (g) Assessment of fluorescence intensity change (%) reveals an increase in C2CD4A intensity at the plasma membrane after the imposed increase in intracellular Ca 2+ levels, which was similar to that observed for Syt1 protein translocation, whilst C2CD4B translocation was significantly lower than Syt1. For each condition, n=3 independent experiments were performed; n=44 cells were tracked in Syt1, n=29 cells were tracked in the case of C2CD4A, and n=30 cells were tracked in the case of C2CD4B. Scale bars = 10 μm. **p<0.01, one-way ANOVA Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.