A TRPV1‐to‐secretagogin regulatory axis controls pancreatic β‐cell survival by modulating protein turnover

Secretagogin is a molecular predictor of type 2 diabetes

Secretagogin is a Ca²⁺‐sensor protein, and when silenced in vitro negatively couples to insulin secretion (Wagner et al, 2000; Hasegawa et al, 2013; Yang et al, 2016). Secretagogin mRNA and protein expression are preferentially associated with endocrine cell identities, including hormone or neuropeptide‐producing cells in brain, pituitary, gastrointestinal tract and pancreas in both human and mouse (Fig EV1A and B; reviewed in: Alpar et al, 2012). More specifically, secretagogin immunoreactivity was seen in glucagon⁺/α‐cells, insulin⁺/β‐cells, somatostatin⁺/δ‐cells and pancreatic polypeptide⁺/PP‐cells in the islets of Langerhans in mouse (Fig EV1C). Thus, secretagogin is a ubiquitous cellular mark in the endocrine pancreas. Nevertheless, this study only focuses on dissecting secretagogin function in differentiated β‐cells.

Despite in vitro probing of its function in exocytosis earlier, the effects of secretagogin deficiency on glucose homoeostasis—particularly in humans—remain unknown. Therefore, we established whether secretagogin could partake in the development of a diabetic phenotype by analysing gene expression data from pancreatic islets of n = 32 subjects diagnosed with type 2 diabetes and n = 170 healthy controls (Fadista et al, 2014). We found a significant negative correlation between secretagogin mRNA expression and the incidence of diabetes (log₂FC = −0.42, P = 0.0066; Fig 1A). Higher secretagogin mRNA expression also predicted lower glycated haemoglobin in blood (HbA1c; R ² = 0.059, P = 0.0097; Fig 1A₁). Moreover, a significant positive correlation between secretagogin and insulin mRNA expression (R ² = 0.2, P = 2.7e⁻¹¹; Fig 1A₂) suggested a role for secretagogin in the hormonal control of glucose metabolism in vivo.

TRP channels regulate secretagogin mRNA expression

Ca²⁺‐sensors are regulated by ionic conductances through Ca²⁺ permeable channels, thus allowing their efficient translation of environmental stimuli into adequately‐tuned physiological cellular responses (Celio, 1990; Schwaller, 2010). Considering that ion binding of any sensor protein is not restricted to a single cation, we first tested secretagogin's metal ion specificity (the basis for being a “Ca²⁺‐sensor”) using recombinant secretagogin (Fig EV2). We demonstrate seceretagogin's preference for bivalent members of the Beryllium group (Be, Ca, Mg, Sr) for protein‐fold stabilization using the Thermofluor technique (Ericsson et al, 2006), which particularly favours Ca²⁺ over Mg²⁺ under quasi‐physiological ion concentrations considering its ~100‐fold higher affinity for Ca²⁺ (Rogstam et al, 2007).

We hypothesized that Ca²⁺ entry could serve as an essential feedback for the expression of this Ca²⁺‐sensor protein. Therefore, we searched for Ca²⁺ permeable channels that (i) can work independently or in tandem with glucose‐regulated voltage‐gated Ca²⁺ channels, which are lowly expressed in β‐cells (Rorsman & Trube, 1986), (ii) their phasic activity can fine‐tune β‐cell Ca²⁺ signalling, (iii) their ligand levels are changing in type 2 diabetes and (iv) can affect cell proliferation (tissue size) during pancreas development or in pancreatic tumours since secretagogin is re‐expressed upon tumorigenic transformation (Wagner et al, 2000; Birkenkamp‐Demtroder et al, 2005; Lai et al, 2006). An appealing candidate satisfying the above criteria is the transient receptor potential (TRP) family of channels, which are evolutionarily conserved in the endocrine lineage, Ca²⁺ permeable, and their genetic ablation is implicated in tissue size control (Zhao et al, 2013; Riera et al, 2014; Malenczyk et al, 2015). When screening single‐cell RNA‐seq data in 270 β‐cells from healthy and type 2 diabetic donors (Segerstolpe et al, 2016), we find that 17 TRP family members are expressed (> 0.001 RPKM; Fig 1B) with TRPC1 and TRPC4AP levels significantly increased in diabetes. Expression of most other channels was retained in β‐cells from diabetic donors. We validated mRNA expression for Trpc1, Trpm3, Trpm5, Trpm7, Trpv1 and Trpv3 as representatives of the different TRP subfamilies (Clapham et al, 2001) as well as for Grin1 (Marquard et al, 2015) in INS‐1E cells (Fig 1C). Next, we determined the effect of agonist stimulation of NMDA receptor, TRPM3, TRPV1 and TRPV3 with exogenous ligands (NMDA (20 μM), CIM 0126 (1 μM; Held et al, 2015), capsaicin (300 nM) and 2‐APB (25 μM; Hu et al, 2009), respectively) on secretagogin mRNA expression. We applied 30 mM KCl to induce depolarization‐dependent Ca²⁺ influx as control. After 30 min, CIM 0126 (P < 0.05), 2‐APB (P < 0.01) and capsaicin (P < 0.001) increased secretagogin mRNA levels with capsaicin producing the most pronounced change (Fig 1D). In turn, KCl‐induced passive Ca²⁺ overload reduced secretagogin expression, which was modulated by NMDA co‐application (Fig 1D). These data show that temporally confined, receptor‐mediated signalling events produce meaningful signals to induce secretagogin transcription.

TRPV1‐induced secretagogin accumulation in β‐cells is Ca²⁺ dependent

TRPV1s are particularly interesting since they are rapidly activated by anandamide, other endovanilloids and, for example, capsaicin (Di Marzo & De Petrocellis, 2012). In pancreas, both α‐ and β‐cells express rate‐limiting enzymes of anandamide metabolism and TRPV1s (Akiba et al, 2004; Malenczyk et al, 2013, 2015), which have been functionally implicated in the control of the intracellular Ca²⁺ level, insulin secretion and β‐cell turnover (Akiba et al, 2004; De Petrocellis et al, 2007; Riera et al, 2014; Malenczyk et al, 2015). Anandamide is present in circulating blood, and its levels significantly increase in obesity and type 2 diabetes (Matias et al, 2006). In juvenile mouse plasma, we measured anandamide concentrations to be 20.4 ± 8.2 nM, superseding pK_i for anandamide to activate TRPV1s (5.78 nM; Ross et al, 2001).

We verified TRPV1 involvement in capsaicin‐induced secretagogin mRNA expression (30 and 120 min time‐points are shown; P < 0.05; Fig 1E) by demonstrating its sensitivity to capsazepine (TRPV1 antagonist, 10 μM) pre‐treatment. We could also see the stimulatory effect of capsaicin on protein translation in INS‐1E cells (2–12 h; Fig 1F). Finally, we confirmed the role of Ca²⁺ in the transcriptional regulation of secretagogin expression by showing that removal of extracellular Ca²⁺ precluded the TRPV1‐dependent transcription of secretagogin (Fig 1G). Thus, a Ca²⁺‐dependent mechanism might operate downstream from TRPV1 channels to control secretagogin function in β‐cells.

TRPV1s recruit Sp1 transcription factors to regulate secretagogin transcription

We then analysed the mouse and human secretagogin promoter in silico to predict binding sites for transcription factors potentially involved in its expressional control (Tables EV1 and EV2). Since Ca²⁺ influx seems indispensable to increase secretagogin mRNA content in INS‐1E cells, we focused on the Ca²⁺‐dependent transcription factor Sp1 because its consensus binding sites were detected in the putative secretagogin promoter (−1,400 to +1) in both species (Fig 2A and A₁). Notably, Sp1 activity has previously been linked to transmembrane channels permeable for bivalent cations, including TRPV1 (Moon et al, 2012).

Acute capsaicin stimulation (300 nM; 30 min) of TRPV1 receptors in INS‐1E cells induced the nuclear translocation of Sp1. This response was capsazepine sensitive, confirming TRPV1 involvement (P < 0.05 vs. control; Figs 2B–B₂ and EV3A–A₃). Similarly, activation of TRPM3 and TRPV3 led to Sp1 translocation (Fig EV3C–C₆) at magnitudes lesser than for TRPV1, corroborating our observations on secretagogin mRNA expression (Fig 1C). Likewise, TRPV1 stimulation under Ca²⁺‐free conditions occluded Sp1 translocation (Figs 2B₃ and EV3B–B₃).

To define the requirement of Sp1 binding to the enhancement of secretagogin transcription, we generated a construct with the full mouse secretagogin promoter (Scgn‐full, its 1,400 bps 5′ flanking region) driving firefly luciferase in comparison to a construct in which predicted Sp1 binding sites were deleted (Scgn‐ΔSp1; Fig 2C). Capsaicin (300 nM; 30 min) significantly increased luciferase‐induced chemiluminescence in the Scgn‐full construct (P < 0.01 vs. control; Fig 2C₁), measured as the ratio of firefly to Renilla luciferase activity. In contrast, Scgn‐ΔSp1 resulted in a substantial decrease in both basal and capsaicin‐stimulated luciferase expression and activity (P < 0.01 vs. Scgn‐full; Fig 2C₁), demonstrating the critical contribution of Sp1 transcription factors to regulating secretagogin expression.

Secretagogin links TRPV1 activity to regulated insulin exocytosis

We used Trpv1 ^−/− mice to confirm Sp1 function downstream from TRPV1s in β‐cells in vivo. We show significantly reduced total and nuclear Sp1 immunoreactivity (P < 0.01 and P < 0.01, respectively; Fig 3A–A₃) in Trpv1 ^−/− mice relative to wild‐type littermates. Coincidently, secretagogin protein (P < 0.05; Fig 3B–B₂) and mRNA levels (P < 0.05; Fig 3C) were significantly reduced on a Trpv1 ^−/− background. Cumulatively, the above data show that endovanilloid signalling in β‐cells promoting Ca²⁺ influx via TRPV1s can modulate secretagogin expression through an Sp1‐dependent transcriptional mechanism.

Considering that circulating levels of anandamide, the chief endogenous TRPV1 agonist, are increased in type 2 diabetes (Matias et al, 2006; Starowicz et al, 2008) and lead to TRPV1 desensitization (Zsombok et al, 2011; Sanz‐Salvador et al, 2012; Gao et al, 2012), we asked whether reduced secretagogin levels in Trpv1 ^−/− mice represent an adaptive response or instead the loss of secretagogin itself can per se precipitate diabetes and modulate TRPV1 signalling. In cultured INS‐1E cells, we found that acute siRNA silencing of secretagogin increased TRPV1 protein levels (P < 0.05; Figs 4A and EV4A and B). Nevertheless, TRPV1 phosphorylation at its Ser⁸⁰⁰ residue did not change (as the ratio of total TRPV1; Fig 4A) suggesting that excess TRPV1s brought about by secretagogin knock‐down do not confer gain‐of‐function. Next, we generated secretagogin^−/− mice (for histological validation of protein loss see: Fig EV4C–E). Notably, secretagogin^−/− mice present > sixfold increase in Trpv1 mRNA expression (P < 0.05; Fig 4B), as well as elevated TRPV1 protein levels in pancreatic islets (P < 0.05), particularly in β‐cells (Fig 4C–C₂). Therefore, we used Ca²⁺ imaging to establish whether surplus TRPV1s are functionally competent or, alike our in vitro observations (Fig 4A), are signal deficient. Ca²⁺ imaging in dissociated pancreatic islets showed no difference in either glucose (16 mM) or capsaicin (300 nM)‐stimulated Ca²⁺ influx between secretagogin^−/− and wild‐type β‐cells (Fig 4D and D₁). Furthermore, and despite the upregulation of TRPV1s, capsaicin, (300 nM) stimulation of pancreatic islets isolated from secretagogin^−/− mice was insufficient to potentiate basal (2.75 mM glucose) or rescue glucose‐induced insulin release (16.5 mM glucose; both P < 0.01 vs. wild type; Fig 4E). Cumulatively, our data suggest that secretagogin acting downstream of Ca²⁺ influx is an essential effector linking TRPV1 activity and glucose metabolism to regulated insulin exocytosis in β‐cells, and whose loss can stem insulin release even in the presence of excess TRPV1s.

Secretagogin^−/− mice develop progressing glucose intolerance

We went on to test whether secretagogin loss‐of‐function may be causal to a diabetic phenotype in vivo. Indeed, secretagogin^−/− mice at both 6 weeks and 6 months of age were glucose intolerant as shown by their attenuated blood glucose clearance (Fig 5A). Moreover, secretagogin^−/− mice at 6 months of age presented significantly elevated fasting blood glucose levels (P < 0.01; Fig 5A). Notwithstanding, secretagogin^−/− mice presented no change in their body weight relative to their wild‐type littermates (17.6 ± 1.5 g [secretagogin^−/−] vs. 17.3 ± 1.7 g [wild type] at 6 weeks and 28.7 ± 3.5 g [secretagogin^−/−] vs. 28.5 ± 3.7 g [wild type] at 6 months). We then generated β‐cell‐specific secretagogin^−/− mice (Ins1‐Cre^ERT::secretagogin^{floxed/floxed}; “β‐Scgn ^CKO”) to confirm that the detrimental effect of secretagogin loss on glucose metabolism relies specifically on β‐cell dysfunction. β‐Scgn ^CKO mice lacked detectable immunoreactivity for secretagogin in > 80% of β‐cells 1 week after daily tamoxifen administration (P < 0.001; Fig EV4F–H) and presented both elevated fasting blood glucose levels (P < 0.01; Fig 5B) and significantly attenuated blood glucose clearance (Fig 5B). Next, we found that pancreatic islets from secretagogin^−/− mice, unlike those of their wild‐type littermates, were insensitive to glucose (P < 0.05 vs. wild type) or KCl stimulation (P < 0.01 vs. wild type) ex vivo (Fig EV5A). These data were supported by decreased insulin mRNA expression (P < 0.01; Fig EV5B), decreased intensity of insulin immunoreactivity in individual β‐cells (at both 6 weeks and 6 months, P < 0.05; Fig EV5C) and reduced total insulin content in pancreatic islets (P < 0.05; Fig EV5D) in secretagogin^−/− mice. Disrupted insulin production in secretagogin^−/− mice was also suggested by the reduced number of insulin granules in β‐cells (P < 0.01; Fig EV5E–E₂) with a particularly notable effect on the number of docked vesicles (in direct contact with the plasma membrane of β‐cells; P < 0.001; Fig EV5E₃). These data together with previous in vitro interaction analyses of secretagogin and cytoskeletal proteins (tubulin and molecular motors of axonal traffic; Maj et al, 2010; Bauer et al, 2011), as well as components of the exocytosis machinery (Rogstam et al, 2007; Romanov et al, 2015) unequivocally place secretagogin within the SNARE complex.

When assaying the number of cell types within pancreatic islets of both 6‐week and 6‐month‐old secretagogin^−/− mice, we found an age‐dependent β‐cell loss, which was accompanied with α‐cell hyperplasia (Fig 5C and D). The number of δ‐ and PP‐cells remained unchanged at either time point. Since α‐to‐β‐cell imbalance is a principal cause of progressive glucose intolerance and diabetes (Deng et al, 2004), a phenotype secretagogin^−/− mice exhibit by 6 months of age (P < 0.01 vs. wild type at 6 months of age, Fig 5C–D₂), we tested whether β‐cell loss is a surrogate of secretagogin loss‐of‐function or instead, secretagogin directly controls β‐cell survival and turnover.

Secretagogin modulates protein folding and degradation through protein–protein interactions

Considering the abundance of secretagogin on, for example, ER membranes (Romanov et al, 2015; Hanics et al, 2017), we screened possible protein–protein interactions unrelated to either the cytoskeleton or SNARE assembly. By using His₆‐tagged recombinant secretagogin in the presence of Ca²⁺ (100 μM), we precipitated putative interacting partners from INS‐1E cells and identified their amino acid sequences by mass spectrometry. We retained canonical interacting proteins participating in vesicle‐mediated transport, exocytosis and cytoskeletal organization as positive controls (Fig 6A, Table EV3). Here, we focused on captured proteins controlling protein folding and (de‐)ubiquitination (Fig 6B) since disrupted protein maturation is particularly detrimental for secretory cells and could lead to ER stress and eventual cell death (Fig 6C). When annotating secretagogin's interacting partners functionally grouped as those with “protein folding” involvement, we find members of the chaperonin‐containing complex (CCT; Fig 6B, Table EV3), known to assist cytoplasmic proteins including actin, tubulin, p53 and c‐Myc to acquire their native and functionally active conformations (Grantham et al, 2002; Koch et al, 2007; Lin et al, 2012; Trinidad et al, 2013). As such, we then probed T‐complex protein 1 subunit 8 (CCT8) in pancreatic islets of secretagogin^−/− mice and verified its levels to be significantly reduced (Fig 6D).

Additionally, mRNA expression of Atf4, a typical marker of ER stress (Harding et al, 2003), was significantly increased in islets from secretagogin^−/− mice relative to wild‐type littermate controls (P < 0.05; Fig 6E). We validated the above data as being predictive, rather than auxiliary to progressive diabetes in our secretagogin^−/− model, by acute siRNA silencing of secretagogin expression in INS‐1E cells (Fig EV4A and B). These experiments (48 h survival after siRNA exposure) confirmed significantly increased Atf4 RNA expression (P < 0.001; Fig 6E₁). If reduced secretagogin expression induces ER stress then one can hypothesize a negative correlation between secretagogin and ATF4 mRNA levels in human diabetic subjects. Indeed, we found a negative correlation in human pancreatic islets (R ² = −0.35, P < 0.001; Fig 6F) emphasizing a likely mechanistic link between secretagogin loss‐of‐function and ensuing ER stress.

Unresolved ER stress and the decreased stability of proteins affecting the cell cycle may lead to apoptosis (Suzuki et al, 2001; Yan et al, 2014). Since the number of β‐cells in pancreatic islets of secretagogin^−/− mice is significantly reduced when compared to wild‐type littermates (Fig 5D–D₂), we analysed the effect of secretagogin deficiency on the level of CHOP (Fig 6C), which mediates apoptosis (Matsumoto et al, 1996). Chop mRNA expression was significantly upregulated in both INS‐1E cells upon secretagogin knock‐down (P < 0.05; Fig 6G) and in secretagogin^−/− islets (P < 0.05; Fig 6G₁) relative to their respective controls. These data were extended by showing significantly increased CHOP immunoreactivity in pancreatic islets from secretagogin^−/− mice (P < 0.01; Fig 6H–H₂) to suggest that ER stress‐related cell death is the likely cause of β‐cell loss in secretagogin^−/− mice.

Secretagogin interacts with USP9X in β‐cells to control cell survival

Impaired protein folding, the primary cause of ER stress in β‐cells, is resolved by increased protein degradation through ubiquitination (Allen et al, 2004). Acute silencing of secretagogin expression increased the level of ubiquitinated proteins (total) in INS‐1E cells by ~50%, equivalent to that of the proteasome inhibitor MG132 (10 μM, 6 h; Fig 7A). However, and contrasting the effect of MG132, secretagogin knock‐down did not increase the total protein content of INS‐1E cells (Fig 7A₁), suggesting either a reduced rate of protein translation or accelerated proteasomal degradation. Our interactome profiling (Fig 6A and B) suggested that Ca²⁺‐bound secretagogin could selectively interact with the ubiquitin carboxyl‐terminal hydrolase USP9X, and possibly USP7. Secretagogin's interaction with both USP9X and USP7 was first validated by pull‐down with His₆‐tagged secretagogin followed by immunoblot detection (Fig 7B). Secondly, we found a significant correlation between secretagogin and USP9X mRNA expression in our human cohort (R ² = 0.21, P < 0.001; Fig 7C) lending further support to functional interactions. Subsequently, we hypothesized that secretagogin's interaction with ubiquitin carboxyl‐terminal hydrolases could prime their deubiquitinating/protein stabilizing activity (with enzymatic activity being reduced on null backgrounds). We tested this concept by detecting known USP9X and USP7 substrates involved in cell cycle control: p53, Mdm2, Mcl‐1 (Li et al, 2002; Schwickart et al, 2010). As such, secretagogin knock‐down produced a pronounced decrease in p53 levels in INS‐1E cells (Figs 7D and EV6A), validating that secretagogin could interact with USP9X, and thus modulate its activity.

To test the direct output of these changes on cell survival, we assayed Ki67 and cleaved caspase 3, markers of cell proliferation and apoptotic cell death, respectively, in INS‐1E cells after secretagogin knock‐down. Secretagogin silencing significantly decreased the number of Ki67⁺ INS‐1E cells (P < 0.01; Fig EV6B–B₂). Simultaneously, the number of cleaved caspase 3⁺ INS‐1E cells significantly increased (P < 0.05; Fig EV6C–C₂). Next, we asked whether inhibition of protein degradation could rescue the detrimental effect of secretagogin knock‐down on INS‐1E cell survival. We found that lactacystin (5 μM, 6 h), a proteasome inhibitor, significantly decreased the number of apoptotic INS‐1E cells which first underwent secretagogin silencing (P < 0.01 vs. siRNA without lactacystin; Fig 7E–E₄), which confirmed secretagogin's role in the USP9X‐mediated modulation of protein turnover to impact β‐cell survival (Fig 7F).

Cell lines

INS‐1E cells (Merglen et al, 2004) were cultured at 37°C in RPMI‐1640 medium supplemented with glutamine (2 mM), glucose (11 mM), HEPES (10 mM), heat‐inactivated foetal bovine serum (FBS; 5%), sodium pyruvate (1 mM), β‐mercaptoethanol (50 μM), penicillin (50 μg/ml) and streptomycin (100 μg/ml). Cells were routinely sub‐cultured in 24‐well plates up to passage 120, and allowed to reach ~80% confluence. Knock‐down of secretagogin expression was achieved by the co‐application of secretagogin siRNA (50 nM; Santa Cruz) and jetPRIME transfection reagent (polyplus‐transfection SA) for 48 h. Scrambled siRNA (50 nM, Santa Cruz) was used as control. Ca²⁺‐free conditions were established by addition of EGTA (1 mM) to modified Krebs‐Ringer bicarbonate HEPES buffer containing (in mM): 135 NaCl, 3.6 KCl, 0.5 MgCl₂, 1.5 NaH₂PO₄, 5 NaHCO₃, 10 HEPES at pH 7.4.

Generation of secretagogin^−/− and β‐Scgn ^CKO mice

Secretagogin^−/− (Scgn ^−/−) mice (Hanics et al, 2017) were custom‐generated at MMRRC (University of California, Davis, Mouse Biology Program) using the “two‐in‐one” targeting strategy (Skarnes et al, 2011), which generates full knock‐outs by expressing a termination signal after exon 3 of the secretagogin gene. The ensuing truncated protein terminates before the first EF‐hand domain, excluding Ca²⁺‐binding activity. To generate β‐cell‐specific secretagogin knock‐out mice (β‐Scgn ^CKO), Scgn ^{floxed/floxed} mice were bred with Ins1‐Cre^ERT mice (Jackson #024709). β‐Scgn ^CKO and wild‐type mice (4‐week old) were intraperitoneally injected with tamoxifen daily over a period of 5 days (200 mg/kg b.w. freshly dissolved in corn oil, both from Sigma; Tamarina et al, 2014). Secretagogin deletion was confirmed a week after the last tamoxifen injection (Fig EV4F–H). Animals (including Trpv1 ^−/− mice, Jackson #3770) were kept under standard housing conditions with a 12‐h/12‐h light/dark cycle and food and water available ad libitum. Experiments on live animals conformed to the European Communities Council Directive (86/609/EEC) were approved by regional ethical committees and regulated by applicable local laws (Tierversuchsgesetz 2012, BGBI, Nr. 114/2012, Austria). Particular effort was directed towards minimizing the number of animals used and their suffering during experiments.

Isolation of pancreatic islets

Islets of wild‐type and secretagogin^−/− mice were obtained after perfusion of their pancreata with Hank's balanced salt solution (HBSS; Invitrogen/Gibco) containing collagenase (type I, 0.33 mg/ml; Sigma) and HEPES (25 mM) followed by purification on Histopaque 1077 gradients (Sigma). After repeated washes in HBSS containing 10% FBS, isolated islets were maintained in RPMI‐1640 medium supplemented as above in humidified atmosphere (5% CO₂) at 37°C overnight prior to the commencement of hormone release measurements, assessment of gene expression or protein levels.

Calcium imaging

Single cells were prepared by isolation of pancreatic islets using the papain dissociation system (Worthington) with some modifications: 300 islets were incubated in 500 μl papain solution in EBSS supplemented with 0.005% DNase at 37°C for 20 min. Then, single cells were plated on 12‐mm coverslips coated with matrigel (Corning) in complete RPMI‐1640 (1:2 v/v ratio) and subsequently cultured in RPMI‐1640 medium supplemented as above in humidified atmosphere (5% CO₂) at 37°C. Twenty‐four hours later, Ca²⁺ responses from isolated pancreatic cells were recorded using the Ca²⁺ indicator Fura‐2AM (Invitrogen). Ratiometric imaging was performed by using a VisiChrome monochromator (Visitron Systems) and CoolSnap HQ² back‐cooled camera (Photometrics) mounted onto a Zeiss Axiovert microscope equipped with a water‐immersion 40×/differential interference contrast objective (Plan‐Apochromat/N.A. 1.0). Ca²⁺ responses were recorded in the VisiView software (Visitron Systems) and statistically analysed in SigmaPlot 13.0 (Systat Software Inc.). Krebs‐Ringer solution with low glucose was chosen as basic external solution and contained (in mM): 119 NaCl, 2.5 KCl, 1 NaH₂PO₄, 1.5 CaCl₂ and 1.5 MgCl₂, 20 HEPES, 3 glucose (pH 7.4). Experiments were performed at 25°C.

Molecular pharmacology

The effect of the following drugs, alone or in combination, on the subcellular localization and/or protein levels of Sp1, secretagogin, ubiquitinated proteins, apoptotic (e.g. cleaved caspase 3), proliferation (e.g. Ki67) and cell cycle (e.g. p53) markers as well as insulin secretion in INS‐1E cells or pancreatic islets was assessed: 2‐aminoethoxydiphenylborane (2‐APB; non‐selective TRPV3 agonist, 25 μM), (E)‐N‐[(4‐hydroxy‐3‐methoxyphenyl)methyl]‐8‐methyl‐6‐nonenamide (capsaicin; TRPV1 agonist, 300 nM), N‐[2‐(4‐chlorophenyl)ethyl]‐1,3,4,5‐tetrahydro‐7,8‐dihydroxy‐2H‐2‐benzazepine‐2‐carbothioamide capsazepine; TRPV1 antagonist, 10 μM), 3,4‐dihydro‐N‐(5‐methyl‐3‐isoxazolyl)‐α‐phenyl‐1(2H)‐quinolineacetamide (CIM 0126; TRPM3 agonist, 1 μM), (2R,3S,4R)‐3‐hydroxy‐2‐[(1S)‐1‐hydroxy‐2‐methylpropyl]‐4‐methyl‐5‐oxo‐2‐pyrrolidinecarboxy‐N‐acetyl‐L‐cysteine thioester (lactacystin; proteasome inhibitor, 5 μM), N‐[(phenylmethoxy)carbonyl]‐L‐leucyl‐N‐[(1S)‐1‐formyl‐3‐methylbutyl]‐L‐leucinamide (MG132; proteasome inhibitor, 5 μM) and N‐methyl‐D‐aspartate (NMDA; NMDA receptor agonist 20 μM).

Production and purification of His₆‐human secretagogin

Expression constructs for human secretagogin (UniProt ID: O76038) included amino acids 1–276. The coding sequence was amplified by PCR using appropriate primers, Pfu‐Turbo polymerase (Stratagene) and cDNA as template and cloned by a ligation independent method into a pNIC28Bsa4 expression vector (GenBank accession no. EF198106) providing a cleavable N‐terminal His₆‐tag (H₂N‐MHHHHHHSSGVDLGTENLYFQ*S). Protein was expressed in E. coli BL21 (DE3) at 21°C in Luria–Bertani broth (LB) supplemented with 30 μg/ml kanamycin. Protein expression was induced by the addition of 0.1 mM isopropyl β‐D‐1‐thiogalactopyranoside at OD₆₀₀ of 0.5–0.6. Cells were harvested 20 h post‐induction, washed in 0.3 M NaCl and 10 mM Tris–HCl (pH 8.0) and disrupted by lysozyme (40 μg/ml) and DNaseI (2 μg/ml) treatment followed by sonication. Recombinant His₆‐hSGCN was purified by affinity chromatography using His‐Pur™ Ni‐NTA resin (Thermo Scientific) and eluted with an increasing imidazole gradient. Using a desalting column (HiPrep‐26/10; Pharmacia Biotech), the elution buffer was exchanged to 0.15 M NaCl and 25 mM Tris–HCl (pH 8.0). Size exclusion chromatography was then performed (Fig EV2A and C) on a Superdex‐200 column (Pharmacia Biotech) equilibrated with 0.15 M NaCl and 25 mM Tris–HCl (pH 8.0). Pure protein samples were concentrated using a Centriprep YM‐10 centrifugal device (Millipore) to 21.2 mg/ml. Protein concentration was determined according to Bradford using Pierce‐Thermo reagents and bovine serum albumin (BSA) as standard. Samples were analysed by SDS–polyacrylamide gel electrophoresis, ESI‐MS mass spectrometry and their folding integrity was confirmed by circular dichroism spectroscopy (Fig EV2B). Protein samples were aliquoted in a buffer containing 0.15 M NaCl and 25 mM Tris–HCl (pH 8.0), flash‐frozen in liquid N₂ and stored at −80°C.

In silico transcription factor binding site prediction and firefly luciferase assay

A 1,400‐bp fragment (−1,400 to +1) of the 5′ flanking region of both the murine (NC_000079.6) and human (NC_000006.12) secretagogin genes was analysed for potential transcription factor binding sites using the open‐label PROMO software (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_6.4). For activity determination of the murine secretagogin promoter, its 1,400‐bp 5′ flanking region (−1,400 to +1) referred to as full Scgn promoter or a deletion mutant lacking predicted Sp1 binding sites at bps −548 to −542, −485 to −480 and −11 to −3 (Fig 2C) were custom‐synthesized (Eurofins Genomics) by cloning between XhoI and NcoI restriction sites into a pGL3‐Basic vector (Promega) also containing a firefly luciferase cassette. Subsequently, INS‐1E cells cultured in 96‐well plates (50,000/well) were co‐transfected with either construct and a pCMV‐green Renilla luciferase vector (Thermo Fisher; as control) using jetPRIME transfection reagent (Polyplus‐transfection SA). Eighteen h later, cells were stimulated with capsaicin (300 nM) for 30 min. After an additional 3.5 h, firefly and Renilla luciferase lumiensence was measured using the Dual‐Luciferase^® Reporter Assay System (Promega) on a GloMax^® luminometer (Promega). Data were analysed as a ratio of firefly to Renilla luciferase lumiensence intensity and normalized to values from unstimulated control cells.

RNA isolation and gene expression analysis

Total RNA from INS‐1E cells and pancreatic islets was isolated using the RNeasy Mini Kit (Qiagen) followed by DNase digestion and verifying RNA integrity on 2% agarose gels (500 ng RNA). cDNA was prepared by reverse transcription with random primers using the High‐capacity cDNA Reverse Transcription Kit (Applied Biosystems). PCR was performed using 2× Master mix (Promega) and specific primer pairs (Table EV4). PCR products were resolved on 2% agarose gels and imaged on a ChemiDoc XRS⁺ system (Bio‐Rad). Quantitative PCR analysis (CFX Connect, Bio‐Rad) was performed using 50 ng of cDNA template, iTaq Universal SYBR Green Supermix (Bio‐Rad) and specific primer pairs (Table EV4).

Western blotting

Proteins were extracted from INS‐1E cells or isolated pancreatic islets using a radioimmunoprecipitation assay buffer containing (in mM): 50 Tris (pH 7.4), 150 NaCl, 10 NaF, 5 EDTA (pH 8.0), 1 Na₃VO₄, 1 PMSF, 1% TX‐100, pepstatin A (5 μg/μl), leupeptin (10 μg/μl) and aprotinine (2 μg/μl), optionally supplemented with MG132 (Tocris; 10 μM; to inhibit protein deubiquitination). Protein lysates were centrifuged at 12,000 g at 4°C for 10 min. For the analysis of total protein levels, lysates were pre‐incubated with carbocyanine (Cy) 5 labelling buffer (GE Healthcare) for 10 min. Samples were denatured in loading buffer (GE Healthcare) and boiled at 95°C for 5 min. Proteins were probed by loading 20 μg aliquots under denaturing conditions on SDS–PAGE (13.5% or 8–18% gradient gels), followed by wet transfer onto PVDF membranes. Primary antibodies were listed in Table EV5. After exposure to Cy3‐ or Cy5‐conjugated secondary antibodies (GE Healthcare, 1:2,500; 1 h), target proteins were visualized using an automated Amersham WB system (GE Healthcare).

Tissue preparation, immunocytochemistry and immunohistochemistry

For immunocytochemistry, cells were plated on 12‐mm coverslips coated with poly‐D‐lysine (0.001%), washed in 0.1 M phosphate buffer (PB, pH 7.4) and immersion fixed in 4% paraformaldehyde (PFA) for 20 min. Mice were transcardially perfused with ice‐cold 0.1 M PB, followed by 4% PFA in PB (100 ml at 3 ml/min flow speed). Pancreata were rapidly dissected and post‐fixed in 4% PFA overnight. After equilibrating in 30% sucrose for 48–72 h, tissues were cryosectioned at a thickness of 14 μm (Leica CM1850) and thaw‐mounted onto SuperFrost⁺ glass slides. After rinsing in 0.1 M PB, specimens were exposed to a blocking solution composed of 0.1 M PB, 10% normal donkey serum, 5% BSA and 0.3% TX‐100 for 3 h followed by overnight incubation with select combinations of primary antibodies (Table EV5). Cy2, 3 or 5‐conjugated secondary antibodies (1:300, Jackson) were applied in 0.1 M PB supplemented with 2% BSA (20–22°C, 2 h). Nuclei were routinely counterstained with Hoechst 33342 (1:20,000; Sigma). Cells and pancreas sections were imaged on a Zeiss LSM880 laser‐scanning microscope equipped with a Plan‐Apochromat 40×/1.3 oil immersion objective at 1.0–2.5× optical zoom. Images were acquired in the ZEN2010 software package. Quantitative analysis was performed using ImageJ 1.45 with appropriate plug‐ins. Multipanel images were assembled in CorelDraw X7 (Corel Corp.).

Electron microscopy

Male mice (n = 3/group) were fasted overnight (~15 h) with water available ad libitum. Thirty min after intraperitoneal glucose challenge (2 g/kg body weight), mice were euthanized by cervical dislocation and transcardially perfused with 0.9% NaCl, followed by fixative (4% PFA, 0.1% glutaraldehyde, 15% picric acid) in PB (3 ml/min flow speed). Pancreata were then dissected out and post‐fixed overnight in the same fixative. After vigorous washing in ice‐cold PB, vibratome sections were cut (40–60 μm) containing pancreatic islets. Sections were repeatedly washed in PB, cryoprotected in 30% sucrose and subsequently freeze–thawed three times in liquid N₂. After extensive washing in PB, slices were incubated with H₂O₂ (1%, 20–22°C, 20 min) to block endogenous peroxidase activity. After washing again with PB, sections were incubated with rabbit anti‐insulin antibody (1:400; 4°C, 48 h). Sections were extensively washed, incubated with biotinylated horse anti‐mouse secondary antibody (1:250; Vector Labs., 20–22°C, 2 h), rinsed again and then exposed to pre‐formed avidin biotin‐peroxidase complexes (1:200; Vector, 2 h) and developed with 3,3′‐diaminobenzidine (5 mg/ml, 0.03% H₂O₂ as substrate). Sections were then osmificated (1% OsO₄ in PB, 15 min), dehydrated in increasing ethanol concentrations followed by flat embedding in Durcupan (FLUKA, ACM). Ultrathin sections (60 nm) were cut on a Leica ultramicrotome, collected on Formvar‐coated single‐slot grids, and analysed with a Tecnai 10 electron microscope (FEI). Investigators were blinded to the experimental groups during the entire procedure. Quantitative analysis (n = 30 cells/group) was performed using ImageJ 1.45 with appropriate plug‐ins. The total number of insulin‐containing secretory vesicles was normalized to the cytoplasm surface. Insulin granules directly contacting the plasmalemma (Fig EV5E′ and E₁′) were considered as docked, and their number was normalized to the length of the plasma membrane (expressed as vesicle/μm).

Pull‐down assay

Sixty micro gram of His₆‐tagged secretagogin and INS‐1E cell lysate (400 μg of protein) was mixed in binding buffer (pH 7.4) containing (in mM): 25 Tris–HCl, 300 NaCl, 0.1 CaCl₂, 10 imidazole to obtain a 100‐μl total sample volume, and incubated together with 40 μl of pre‐equilibrated Ni‐NTA beads in Spin‐X Centrifuge Tube Filter 0.22 μm cellulose acetate columns (Costar) at 20–22°C for 30 min. In control experiments, His₆‐tagged secretagogin was replaced with binding buffer alone to obtain 100 μl of total sample. Flow‐through fractions were removed after spinning at 1,500 g for 30 s. Beads were washed (3×) with 100 μl buffer (in mM): 25 Tris–HCl, 300 NaCl, 0.1 CaCl₂, 50 imidazole. Proteins were eluted with 70 μl elution buffer composed of (in mM): 25 Tris–HCl, 300 NaCl, 0.1 CaCl₂ and 300 imidazole. Samples were stored at −80°C until processing for mass spectrometry.

LC‐MS/MS analysis of protein complexes including statistics

Eluted proteins (17.5 μg) of His₆‐secretagogin (n = 3) and control pull‐downs (n = 3) were digested with trypsin via filter‐aided sample preparation (Wisniewski et al, 2009) with minor modifications (Aradska et al, 2015). Tryptic digests were desalted and concentrated on customized reversed‐phase C18 stage tips (Rappsilber et al, 2007). Purified peptides were reconstituted in 5% formic acid (FA), and 2 × 1 μg of each sample (technical replicates) was analysed by liquid chromatography‐tandem mass spectrometry (LC‐MS/MS). Peptides were separated on an Ultimate 3000 nanoRSLC system (Thermo Fisher) equipped with a PM100‐C18 pre‐column (3 μm, 75 μm × 20 mm) and a PepMap‐C18 analytical column (2 μm, 75 μm × 500 mm). Ten μl of each sample was loaded onto the trap column for 5 min using 0.1% FA in water at a flow rate of 5 μl/min. Afterwards, peptides were forward‐flushed to and separated on the analytical column at a flow rate of 300 nl/min using a 55‐min gradient ranging from 5 to 37.5% solvent B, followed by a 5‐min gradient from 37.5 to 50% solvent B and finally, to 90% solvent B for 5 min before re‐equilibration to 5% solvent B (solvent A: 0.1% FA in water; solvent B: 0.1% FA in 80% acetonitrile). Eluted peptides were in‐line analysed on a Thermo Q‐Exactive Plus mass spectrometer (Thermo Fisher) in positive ion mode. MS scans were performed in a range of m/z 380–1,800 at a resolution of 70,000. In a data‐dependent acquisition mode, the 20 most intense precursor ions were selected and fragmented via higher energy collisional dissociation at 27% normalized collision energy with a fixed first mass of 100 m/z. Fragment ions were detected at a resolution of 17,500. Dynamic exclusion of selected peptides was enabled for 60 s, and maximal accumulation times were set to 100 and 50 ms in MS and MSⁿ modes, respectively. To describe the secretagogin‐specific interactome, all MS raw data files were analysed by the open‐source software MaxQuant 1.5.3.30 (Cox & Mann, 2008) as previously described (Smidak et al, 2016). Proteins were identified against the Swiss‐Prot rat (Rattus norvegicus) reference proteome database (as of November 2015; 31,457 entries) permitting a mass tolerance of 5 ppm and 20 ppm for MS and MS² spectra, respectively. Maximum two missed cleavages were allowed and minimum two peptide identifications per protein were required. Carbamidomethylation of cysteines was set as fixed modification while methionine oxidation and N‐terminal protein acetylation were chosen as variable modifications. For both peptide and protein identification, the false discovery rate (FDR) was set to 0.01. MaxQuant results were further processed using the Perseus statistical package (version 1.5.4.1) to identify interactors from label‐free data (LFQ intensities). Proteins were filtered for reversed sequences, contaminants, and if they were only identified by site. Intensity values were log₂‐transformed and proteins were further filtered for having LFQ values in minimum three samples at least in one group (either in the target or in the control). Zero intensities were input and replaced by normal distribution, and secretagogin pull‐downs were compared to control samples via two‐sided Student's t‐test with P < 0.05 (applying permutation‐based FDR for truncation). Proteins with a minimum twofold increase of their LFQ intensities in the target pull‐down were considered as specific for secretagogin if they were identified by at least two unique peptides.

LC‐MS/MS analysis of endocannabinoid contents

Extraction, purification and quantification of anandamide from the blood of embryos at embryonic day 18.5 followed published protocols (De Marchi et al, 2003; Matias et al, 2006). After lipid extraction and pre‐purification on silica gel columns, anandamide levels from three animals were analysed by isotope dilution using liquid chromatography‐atmospheric pressure chemical ionization‐mass spectrometry (Shimadzu LCMS‐2020).

Insulin secretion and total insulin content ELISA assay

Pancreatic islets were incubated in drug‐free modified Krebs‐Ringer bicarbonate HEPES buffer (KRBHB) containing (in mM): 135 NaCl, 3.6 KCl, 1.5 CaCl₂, 0.5 MgCl₂, 1.5 NaH₂PO₄, 5 NaHCO₃, 10 HEPES (pH 7.4) and BSA (0.1%) supplemented with 2.75 mM glucose for 1 h. Subsequently, samples were transferred to microcentrifuge tubes at a density of 10 islets/tube. Islets were incubated in modified KRBHB supplemented with 2.75 mM glucose with or without 30 mM KCl or 16.5 mM glucose for 30 min. Alternatively, after overnight culturing, islets (30 islets/group) were homogenized in water using 27G needles. Homogenates were mixed with 0.18 M HCl in 96% EtOH (1:3 v/v ratio), incubated at 4°C for 16 h and centrifuged at 1,500 g at 4°C for 15 min. Supernatants were stored at −80°C until processing. Insulin levels were measured by a rat/mouse insulin ELISA Kit (Millipore) with insulin concentrations normalized to either the number of islets or β‐cells.

Glucose tolerance test

Male mice (n = 6/group) were fasted overnight (~15 h) with water available ad libitum. Blood glucose level was measured from tail blood using a FreeStyle Lite glucose meter (Abbott Diabetes Care) before (baseline, 0 min) as well as 15, 30, 60, 90 and 120 min after intraperitoneal glucose challenge (2 g/kg body weight). Data were expressed as mg/dl blood glucose.

RNA‐seq and histochemistry‐based body‐wide mapping of secretagogin distribution

Secretagogin expression in nervous and peripheral tissue in both human and mouse was determined based on (i) open‐source RNA‐seq data and (ii) cap analysis of gene expression (CAGE). RNA‐seq data in human was directly imported from the Genotype‐Tissue Expression Portal (GTEx; GTEx Consortium et al, 2013) while mouse RNA‐seq data were obtained from the ENCODE project repository (Kundaje et al, 2015; Fig EV1A, Table EV6). Secretagogin transcription start sides were quantified based on FANTOM CAGE data (Carninci et al, 2005) using the ZENBU genome browser (http://fantom.gsc.riken.jp/zenbu/). Antibody‐based protein distribution maps in human and mouse tissues were extracted from the Human Protein Atlas project (www.proteinatlas.org) based on antibodies raised against the C‐terminal domain of secretagogin complemented by immunohistochemistry of particular tissues (Fig EV1B, Table EV6) using the same antibody (Mulder et al, 2010).

Gene expression in human subjects

Islets from 202 cadaver donors of European ancestry were provided by the Nordic Islet Transplantation Programme (http://www.nordicislets.org). All procedures were approved by the ethics committee at Lund University and performed as previously described (Fadista et al, 2014). Briefly, purity of islets was assessed by dithizone followed by islets culture in CMRL 1066 (ICN Biomedicals) supplemented with 10 mM HEPES, 2 mM l‐glutamine, 50 μg/ml gentamicin, 0.25 μg/ml Fungizone (GIBCO), 20 μg/ml ciprofloxacin (Bayer Healthcare) and 10 mM nicotinamide at 37°C (5% CO₂) for 1–9 days prior to RNA preparation. Total RNA was isolated with the AllPrep DNA/RNA Mini Kit following the manufacturer's instructions (Qiagen). RNA quality and concentration were measured using an Agilent 2100 bioanalyzer (Bio‐Rad) and a Nanodrop ND‐1000 (NanoDrop Technologies). Samples were stratified based upon glucose tolerance estimated from HbA1c, that is donors with normal glucose tolerance (HbA1c < 6%, n = 123), impaired glucose tolerance (IGT, 6% ≤ HbA1c < 6.5%, n = 47), and type 2 diabetes (HbA1c ≥ 6.5%, n = 32). A linear model adjusting for age and sex as implemented in the R Matrix eQTL package was used to determine the expression of genes associated with glucose tolerance status. Raw data were base called and de‐multiplexed using CASAVA 1.8.2 (Illumina) before alignment to hg19 with STAR. To count the number of reads aligned to specific transcripts feature Counts (v 1.4.4, http://bioinf.wehi.edu.au/featureCounts/) was used. Raw data were normalized using trimmed mean of M‐values (TMM) and transformed into log₂ counts per million (log₂CPM) using voom (limma R, Bioconductor) before linear modelling. After TMM‐normalization, batch effects were removed using the Combat function (sva package).

Single‐cell RNA‐seq of pancreatic islets

Analysis was performed based on a publicly available expression matrix (Segerstolpe et al, 2016) from 270 human β‐cells from six healthy (n = 171 cells) and four type 2 diabetic donors (n = 99 cells). Briefly, human tissue and primary islets were purchased from Prodo Laboratories Inc., providing islets isolated from donor pancreata obtained from deceased individuals with research consent from Organ Procurement Organizations. Human islet samples (85–95% pure) were cultured for 4 days in complete Prodo Islet Media Standard PIM(S) after arrival. Islets were dissociated and distributed by FACS into 384‐well plates. Single‐cell RNA‐seq libraries were produced with the Smart‐seq2 protocol (Segerstolpe et al, 2016). Sequencing was carried out on an Illumina HiSeq 2000 generating 43‐bp single‐end reads. Sequence reads were aligned towards the human genome (hg19 assembly) using STAR (v2.3.0e), and uniquely aligned reads within RefSeq gene annotations were used to quantify gene expression as RPKMs using the “rpkmforgenes” routine (sandberg.cmb.ki.se/media/data/rnaseq/).

Statistics

All experiments were performed in triplicate unless stated otherwise. Data from pharmacology experiments were analysed using one‐way ANOVA followed by pairwise comparisons where appropriate. Student's t‐test (independent group design) was used to statistically evaluate data on pancreatic islet morphology, INS‐1E proliferation and apoptosis and gene expression. Data from Ca²⁺‐imaging and RNA‐seq were analysed using the Mann–Whitney rank‐sum test. Data are expressed as means ± s.d. unless specified. Fold changes represent the percentage change from the untreated (control) value in individual experiments. A P < 0.05 value was considered statistically significant.

Accession numbers

Raw data on single‐cell RNA‐seq of pancreatic beta cells and human islets sequenced in bulk were deposited in ArrayExpress (E‐MTAB‐5061) and Gene Expression Omnibus (GSE50398), respectively. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD006544.