The transcription factor STAT6 plays a critical role in promoting beta cell viability and is depleted in islets of individuals with type 1 diabetes

Human formalin-fixed paraffin-embedded (FFPE) pancreas samples were obtained from the Exeter Archival Diabetes Biobank (http://​foulis.​vub.​ac.​be/), an archival collection of post-mortem pancreas samples [28]. These consisted of six samples taken from individuals with recent-onset type 1 diabetes plus six control samples from individuals of similar age and sex (ESM Table 1). All samples were studied with full ethical approval (15/W/0258).

Where a single antigen was examined, sections were studied using immunoperoxidase staining. In these experiments, samples were dewaxed, rehydrated and heated in a citrate antigen retrieval buffer (pH 7.4) for 20 min using a microwave (800 W). Sections were blocked in 5% (vol./vol.) normal goat serum and then probed with primary and relevant secondary antibodies (ESM Table 2). Haematoxylin was used as a counterstain prior to mounting of the section in distyrene–xylene-based mountant. Alternatively, where two or more antigens were examined, sections were sequentially stained with primary antibodies and species-appropriate fluorescently labelled secondary antisera prior to mounting. Antisera were validated either using relevant control tissues or were used in accordance with manufacturers’ instructions.

Images were captured using an AF6000 fluorescent microscope (Leica Microsystems, Milton Keynes, UK). For quantification studies, randomly selected insulin-containing islets from individuals with and without diabetes were imaged using identical microscope and camera settings. Regions of interest were drawn around the islet periphery and the mean fluorescence intensity (MFI) of each antigen was assessed using Image J version 1.50b (https://​imagej.​nih.​gov/​ij/​) with Java 1.8.0_77 https://​www.​oracle.​com/​technetwork/​java/​javase/​8u77-relnotes-2944725.​html). Alternatively, the MFI of STAT6 immunostaining was calculated only in insulin-positive regions using a custom MATLAB script (version R2015b) (www.​mathworks.​com/​products/​new_​products/​release2015b.​html), VIOLA, developed in house (University of Exeter). The VIOLA script identifies insulin-positive regions within each image by thresholding against insulin-negative regions. It then reports the MFI of a second antigen (in this case STAT6) only in areas where the two are co-localised.

Cultured INS-1E cells (a gift from C. Wollheim, University of Geneva, Switzerland) were used to perform most in vitro experiments. Cell culture was achieved in RPMI-1640 at 11 mmol/l glucose (Lonza, Basel, Switzerland) supplemented with 10% FBS, 2 mmol/l l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin and 50 mmol/l 2-mercaptoethanol (all from ThermoFisher, Boston, MA, USA) and treatments were performed at 2 × 10⁵ cell/ml seeding density [29]. Cells were treated with 20 ng/ml of each cytokine (IL-4, IL-13, IL-1β, TNF-α, IFN-γ and IL-6; all from R&D systems, Abingdon, UK), 250 μmol/l palmitic acid complexed to 1% bovine serum albumin or after serum withdrawal for the desired duration. These concentrations were chosen since they had been shown to be effective in our previous studies [20, 30]. Human EndoC-βH1 beta cells were sourced and cultured as described in [31]. Cells were routinely tested for mycoplasma contamination and were negative.

Knockdown of target transcripts was achieved using small interference RNAs (siRNAs) for rat Stat6 and Sirpα (also known as Sirpa) (ThermoFisher). Cells were transfected with target or scrambled siRNA. Scrambled siRNA was generated randomly from the Stat6 sequence (GAAUUAAUCGUCGUCUU), and tested against the NCBI database to confirm the lack of off-target effects. Commercial siRNA sequences are proprietary. Optimem (ThermoFisher) and lipofectamine RNAi Max (Invitrogen, Boston, MA, USA) were used as transfection reagents and successful knockdown was confirmed by western blotting and/or quantitative reverse transcription PCR (qRT-PCR).

Cellular proteins were extracted and used for western blotting as previously described [20]. Primary antibodies (ESM Table 2) were added at 4°C in blocking solution unless stated otherwise. After overnight incubation, membranes were washed for 15 min in tris-buffered saline–Tween (TBST) and probed with appropriate alkaline phosphatase-conjugated secondary antibodies (Merck, Darmstadt, Germany) for 1 h at room temperature. Bands were detected with CDP-star chemiluminescent substrate (Merck) or by Licor Odyssey detection system (Licor, Cambridge, UK) when fluorescent secondary antibodies were used. Densitometric analysis was performed using Image Studio version 5.2 (https://​www.​licor.​com/​bio/​products/​software/​image_​studio/​) after normalising for expression of β-actin or glyceraldehyde 3-phosphate dehydrogenase (GAPDH).

RNA was extracted from cells using an RNeasy Mini kit (Qiagen, Hilden, Germany) and its quantity and quality were estimated by NanoDrop measurement (ThermoFisher). RNA (500 ng) was used for cDNA synthesis (Qiagen) and gene expression was monitored by qRT-PCR with SYBR Green master mix using commercially available RT2 Profiler PCR Array and primers for genes of interest (Qiagen). Amplicons were generated on the QuantStudio Flex 12K (Applied Biosystems, Boston, MA, USA) and gene expression was calculated using the comparative threshold cycle method (\( {2}^{-\Delta \Delta {\mathrm{C}}_{\mathrm{t}}} \)) after normalising with transcripts encoding Hprt1 and Yy1 [32].

Viability was estimated using either Trypan Blue (0.4% wt/vol. in PBS) or propidium iodide (PI) (Merck) as previously described [26]. Routinely, each experimental condition was replicated six times and individual experiments were repeated on at least three separate occasions.

A single time point cell cycle analysis was performed by PI staining as described [33].

All statistical analyses were performed on Graphpad Prism version 7.0 (https://​www.​graphpad.​com/​scientific-software/​prism/​) and data are presented as mean values ± SEM. Unpaired Student’s t test or ANOVA (with post hoc Tukey’s test) were used to assess statistical significance between mean values. Data were considered statistically significant when p < 0.05.

The cytoprotective actions of IL-13 were investigated in rodent INS-1E and human EndoC-βH1 cells. As expected, withdrawal of serum or treatment of INS-1E cells with a proinflammatory cytokine cocktail caused a dramatic loss of viability, which was attenuated by IL-13 (Fig. 1b–c). IL-4 also protected beta cells against serum withdrawal (serum withdrawal 53.5 ± 1.5% cell death; serum withdrawal + IL-4 44.4 ± 1.1%; p < 0.01). Importantly, an equivalent cytoprotective response to IL-13 was also observed in EndoC-βH1 cells treated with proinflammatory cytokines (Fig. 1d), confirming that the response is not restricted only to rodent cells. IL-13 treatment of INS-1E cells led to increased tyrosine phosphorylation of STAT6 (Fig. 1a) within 15 min and, as seen previously [20], three separate immunoreactive pSTAT6 bands could be detected by western blotting, all of which were enhanced in response to IL-13.

To understand the importance of STAT6 in mediating the cytoprotective response to IL-13 in beta cells, siRNA molecules targeting Stat6 selectively were employed. Transfection of Stat6 siRNA into INS-1E cells caused an approximately 75% reduction in STAT6 protein levels relative to the scrambled siRNA-treated control cells, within 48 h (Fig. 2a). STAT6 knockdown was stable for at least 4 days (Fig. 2b) but returned to pre-treatment levels within 6 days of transfection (not shown).

To examine the effect of depletion of STAT6 on the ability of IL-13 to promote cytoprotection, siRNA directed against Stat6 was transfected into cells for a period of 24 h prior to treatment with IL-13. The cells were then exposed either to a period of serum withdrawal or were treated with a cocktail of proinflammatory cytokines or 250 μmol/l palmitate. As expected, IL-13 improved the viability of cells incubated under each of these conditions (Fig. 2c, e). By contrast, when STAT6 was knocked down prior to IL-13 treatment, the cytoprotective responses were abrogated. The protective action of IL-4 was similarly sensitive to STAT6 knockdown (Fig. 2d). Interestingly, however, knockdown of STAT6 did not itself lead to any loss of beta cell viability in the absence of a specific cytotoxic stimulus.

Given that the cytoprotective response to IL-13, mediated by STAT6, was most evident following a period of pre-treatment in INS-1E cells, we next examined whether it was accompanied by changes in gene expression. To our knowledge, the transcriptional response to STAT6 activation has not been characterised extensively in beta cells. Therefore, RNA was extracted from INS-1E cells treated with IL-13 for 48 h and cDNA was synthesised for analysis using a targeted PCR array. A range of genes were upregulated robustly under these conditions, with the largest increases observed in the levels of Sirpα, Mcl1, Bcl2l1, Epor, Smad1, Ptpn1, Socs1 and Sh2b1 (Fig. 3a). qRT-PCR analysis confirmed that Sirpα and Socs1 were significantly upregulated by IL-13 (Fig. 3b, c) whereas the increases in Bcl2l1 and Mcl1 did not achieve statistical significance by this method (Fig. 3d,e). Importantly, however, western blot analysis confirmed the upregulation of signal regulatory protein α (SIRPα), myeloid cell leukaemia-1 (MCL1) and B cell lymphoma-extra large (BCLXL), the gene product of Bcl2l1, following exposure of INS-1E cells to IL-13 (Fig. 4a, ESM Fig. 1). Exposure of cells to IL-4 increased Socs1, Sirpα, Bcl2l1, and Epor mRNA levels (Fig. 3b–d, f) and, as in the case of IL-13, this was confirmed at the protein level in cell extracts (Fig. 4a, c, ESM Fig. 1). Neither IL-13 nor IL-4 altered Ifngr1 expression as judged by qRT-PCR analysis, despite the large increase in expression observed in the PCR array (Fig. 3g). Knockdown of STAT6 abrogated the stimulatory effects of IL-13 and IL-4 on Sirpα, Socs1and Bcl2l1 mRNA and their corresponding gene products (Figs 3b–d, f, g and 4c). Equivalent IL-13-induced changes in SIRPα were observed in human EndoC-βH1cells (Fig. 4b). MCL-1 protein expression was elevated in response to IL-13 in human EndoC-βH1cells, although the effects on BCLXL were less marked in these cells (Fig. 4b).

Among the gene products found to be upregulated by IL-13 and IL-4, SIRPα was the most unexpected. This protein is normally expressed at high levels on myeloid cells [34]. It has been studied only rarely in pancreatic beta cells, although we have noted a report implicating SIRPα (also known as Src homology 2 [SH2] domain-containing protein tyrosine phosphatase substrate 1 [SHPS-1]) in the control of insulin secretion [35]. We therefore used primary human islets and were able to verify that treatment with IL-13 resulted in a marked increase in SIRPα expression (Fig. 5a). To study the effects of SIRPα in more detail, we then used siRNA constructs to silence the protein in INS-1E cells. SIRPα expression was robustly knocked down over a period of at least 4 days by the targeted siRNA treatment, as confirmed by qRT-PCR (Fig. 5b) and western blot analysis (Fig. 5c). Importantly, knockdown of SIRPα did not alter the level of STAT6 in INS-1E cells (Fig. 5c). Surprisingly, however, under basal conditions, knockdown of SIRPα significantly enhanced beta cell death compared with cells treated with scrambled control siRNA (Fig. 5d). Analysis of the sub-G1 peak (corresponding to fragmented DNA) by flow cytometry confirmed the increase in cell death associated with depletion of SIRPα (Fig. 5e). Moreover, cell death induced by the withdrawal of serum from cells was exacerbated by the loss of SIRPα (Fig. 5d). It was then assessed whether increasing the expression of SIRPα in beta cells could induce the opposite effect, and promote cell viability. As expected, in these experiments, SIRPα over-expression significantly improved the viability of INS-1E cells cultured in the absence of serum (Fig. 5f). Taken together, these data implicate SIRPα as a novel regulator of beta cell viability whose levels are controlled in a STAT6-dependent manner.

Immunohistochemical staining of STAT6 in human pancreas tissue from non-diabetic donors revealed a robust expression of the protein in a subset of cells in pancreatic islets, with much lower levels observed in the surrounding exocrine tissue (Fig. 6a). Co-immunofluorescence staining revealed that STAT6 is strongly co-localised with insulin, but not with glucagon, suggesting its preferential expression in beta cells (ESM Fig. 2).

To determine whether STAT6 expression is altered in diabetes, a series of six pancreas samples from individuals with recent-onset type 1 diabetes and from six healthy control donors with similar age and sex profiles were used. As expected, STAT6 was expressed robustly within the islets of control individuals but it was noticeably reduced in the insulin-containing islets (ICIs) of people with type 1 diabetes (Fig. 6a–c). To quantify these changes in expression, four or five ICIs from each individual were imaged using identical microscope and camera settings. Measurements of the MFI of the anti-STAT6 immunolabelling confirmed the downregulation of STAT6 in ICIs in type 1 diabetes (Fig. 6c, d). As previously reported, HLA class I (HLAI) expression was significantly elevated in ICIs with depleted STAT6 expression from subjects with type 1 diabetes (Fig. 6e). Since the EADB cohort is primarily an archival collection of post-mortem samples, we verified these data with samples from the Network for Pancreatic Organ Donors with Diabetes (nPOD) collection of organ donor pancreases. Immunofluorescence analysis revealed a similar pattern, with STAT6 robustly expressed in ICIs from healthy individuals but diminished in the islets of individuals with type 1 diabetes (ESM Fig. 3).

To investigate this phenomenon further, we considered the possibility that changes in the islet milieu associated with the development of diabetes might be involved in promoting the loss of STAT6. Accordingly, we used INS-1E cells which, like human islets, express high levels of STAT6 when cultured under control conditions. However, western blot analysis revealed that when these cells were treated with proinflammatory cytokines (IL1β, IFNγ, TNFα and IL-6) for 48 h or were exposed to either 250 μmol/l palmitate or conditions of serum withdrawal, STAT6 was significantly depleted (Fig. 7a, b). Interestingly, pre-treatment of the cells with IL-13 for 48 h led to an increase in STAT6 expression relative to control conditions (Fig. 7c), suggesting a feed-forward upregulation of STAT6 in cells exposed to the anti-inflammatory cytokine. Moreover, addition of IL-13 resulted in partial preservation of STAT6 expression in cells treated with the proinflammatory cytokine cocktail (Fig. 7c).