To investigate whether WFS1 dysfunction leads to sterile inflammation via terminal ER stress in pancreatic β-cells, we first examined apoptosis and pro-inflammatory cytokine gene expression levels in Wfs1 knockdown INS-1E cells. In this in vitro Wolfram syndrome model, cleaved caspase-3 levels were increased, suggesting activation of apoptosis ( Figures 1A, B ). Furthermore, the gene expression levels of pro-inflammatory cytokines [Il-1β, Il-6, and Cxcl1(Il-8)] and Chop were significantly upregulated ( Figure 1C ). These results indicate that WFS1 plays a role in regulating ER stress-mediated cell death and pro-inflammatory cytokine gene expression.

Diabetes, caused by ER stress-mediated pancreatic β-cell apoptosis, is one of the major and early manifestations of Wolfram syndrome (1, 19), suggesting that pancreatic β-cells of Wolfram syndrome patients are exposed to hyperglycemia for extended time periods. Therefore, we investigated how WFS1-deficient pancreatic β-cells produce pro-inflammatory cytokines under a chronic high-glucose environment. First, we examined the influence of chronic high-glucose conditions on cell death and ER stress using Wfs1-knockout INS-1 832/13 cells (Wfs1-KO INS-1 cells) ( Figure 2A ). These Wfs1-KO INS-1 cells possess the characteristics of pancreatic β-cells in Wolfram syndrome: thapsigargin intolerance and mitochondrial dysfunction (20) ( Figure S1 ). Wfs1-KO INS-1 cells treated with high glucose showed elevated proinsulin expression, increased caspase-3/7 activity, decreased cell viability, and increased gene expression of ER stress markers such as Chop, sXbp1, Bip, and Txnip compared to wild-type INS-1 832/13 cells (Wfs1-WT INS-1 cells) ( Figures 2B–D ). In this condition, the gene expression levels of pro-inflammatory cytokines (Il-1β, Il-6), chemokine (Ccl2), and vascular endothelial growth factor A (VegfA) were increased in Wfs1-KO INS-1 cells ( Figure 2E ). The chronic high-glucose condition enhanced not only the cell apoptosis but also the expression of pro-inflammatory cytokine genes in Wfs1-KO INS-1 cells.

Pro-inflammatory cytokines secreted from pancreatic β-cells are known to act in a paracrine or autocrine manner in type 2 diabetes models (21). Therefore, we hypothesized that the locally secreted cytokines might influence the characteristics of pancreatic β-cells in Wolfram syndrome. To test this idea, we evaluated the cell death and ER stress markers in Wfs1-KO INS-1 cells treated with cytokines. Wfs1-KO INS-1 cells treated with IFN-γ and IL-1β showed enhanced ER stress-mediated cell death ( Figure 3A ). Furthermore, the gene expression levels of Il-1β, Il-6, and Ccl2 were higher in Wfs1-KO INS-1 cells compared to Wfs1-WT INS-1 cells ( Figure 3B ). These results indicate that cytokine treatment enhances ER stress-mediated cell death and upregulates the pro-inflammatory cytokine gene expression in Wfs1-deficient β-cells.

Next, we investigated the mechanisms underlying the upregulation of pro-inflammatory cytokine gene expressions in Wfs1-KO INS-1 cells. We first examined the protein expression level of nuclear-localized NF-κB, which is known to regulate the expression levels of pro-inflammatory cytokines genes (13). We observed enhanced NF-κB nuclear translocation in Wfs1-KO INS-1 cells following high-glucose treatment ( Figure 4A ).

We next examined how high-glucose treatment activates the NF-κB pathway. As described above, the downstream pathways of terminal UPR were activated in high-glucose treated Wfs1-KO INS-1 cells, enhancing the expression level of ER stress markers such as Txnip. TXNIP is known to activate the NLRP3 inflammasome, which leads to caspase-1 activation and IL-1β maturation, and its gene expression is regulated by PERK and IRE1 (22–24). Therefore, to determine whether the terminal UPR pathway enhances the NF-κB signaling pathways, we downregulated PERK and IRE1α in Wfs1-KO INS-1 cells. We found that the PERK was activated in Wfs1-KO INS-1 cells by high-glucose treatment ( Figure 4B ). Unlike knocking down IRE1α, knocking down PERK suppressed the enhanced expression of pro-inflammatory cytokine genes in high-glucose treated Wfs1-KO INS-1 cells ( Figures 4C, D ). These results suggest that the induction of pro-inflammatory cytokine gene expressions in high-glucose treated Wfs1-KO INS-1 cells was mainly mediated by the PERK pathway and subsequent NF-κB nuclear translocation. The increased Txnip expression in high-glucose treated Wfs1-KO INS-1 cells was suppressed by simultaneously knocking down PERK and IRE1α ( Figure 4E ). These results suggest that high-glucose treatment of Wfs1-KO-INS-1 cells enhances NF-κB-mediated pro-inflammatory cytokine genes expression via the PERK pathway and enhances TXNIP-mediated IL-1β processing through the PERK and IRE1α pathways.

To confirm the findings above, we next investigated whether inflammation occurs in a mouse model of Wolfram syndrome. Although several Wolfram syndrome mouse models have been developed, we used the Wfs1 whole-body knockout 129S6 mice (Wfs1-KO mice) in this study (18, 25), which recapitulate human Wolfram syndrome phenotypes and progressively develop glucose intolerance between 4.5 and 6.5 weeks old that persists through at least 10 months of age (18) ( Figures S2A, B ). It is known to be challenging to isolate a large number of islets from these mice because of their genetic background and small size. Thus, we had to pool islets from several mice or to use imaging mass cytometry for the following experiments. First, we evaluated gene expression levels of pro-inflammatory cytokines and ER stress markers in primary islets from Wfs1-wild-type (WT) and Wfs1-KO mice. The primary islets isolated from 3-week-old to 12-week-old Wfs1-WT or -KO male mice were pooled and cultured in 30 mM glucose for 3 days. The expression levels of pro-inflammatory cytokine genes (Il-1β, Ccl2, Il-6) and ER stress markers (Bip, sXbp1, Txnip, Chop) tended to be higher in Wfs1-KO primary islets than in control islets. However, there was a large variation among the samples, and the difference was not statistically significant (data not shown). To further investigate the relationship between WFS1 and islet-localized inflammation, we collected the pancreatic tissues from mature-adult (16-week-old) and middle-aged (10-month-old) Wfs1-WT and Wfs1-KO mice and examined whether macrophages infiltrate the islets. Pancreatic tissue sections from each age were stained with Iba1, a marker for M1/M2 macrophage. Although there were no significant differences in 16-week-old mice, higher Iba1 density was observed in Wfs1-KO mice islets at 10 months age-old ( Figures 5A, B ). To further evaluate the character of these islet-resident macrophages, we performed imaging mass cytometry (IMC) on the Wfs1-KO mice pancreatic sections. The intra-islet macrophages in Wfs1-KO mice pancreatic islets expressed CD68 and IL-1β, indicating inflammatory M1-like characteristics ( Figure 5C ). There was also strong fibrosis in the islets of Wfs1-KO mice ( Figures 5D, E ), previously described as the hallmark of chronic inflammation in the diabetic islets (26). Together, these findings indicate the presence of inflammation in the Wfs1-KO mice islets.

To examine whether the inflammation was also present in other tissues, we assessed cytokine levels in serum and bone marrow-derived macrophages of Wfs1-KO mice. Serum cytokine levels, notably interferon gamma-induced protein 10 (IP-10, CXCL10), was elevated in Wfs1-KO mice ( Figure S3 ). These results suggest that the inflammation in these Wfs1-KO mice was not limited to the pancreatic islets. Although the high-glucose condition did not alter the gene expression levels of pro-inflammatory cytokines in bone marrow-derived macrophages ( Figures S4A, B ), advanced glycation end products (AGE) treatment enhanced Ccl2 gene expression in bone marrow-derived macrophages of Wfs1-KO mice ( Figures S4C, D ).

In Wfs1-KO INS-1 cells, we observed that high-glucose treatment enhanced VegfA gene expression ( Figure 2E ). VEGFA, an angiogenic factor, is expressed in the mouse islet and regulates vascular growth and permeability (27–29). Therefore, we tested if increased VEGFA expression may cause abnormal angiogenesis in the pancreatic islets of Wfs1-KO mice. We evaluated the endothelial cell marker (CD31) in the pancreas of Wfs1-WT and Wfs1-KO mice. A higher CD31 signal density was observed in Wfs1-KO mice islets compared to Wfs1-WT mice ( Figures 6A, B ). These results indicate the existence of hypervascularization in the islets of Wfs1-KO mice.

Wfs1-KO INS-1 832/13 cells were generated in collaboration with Genome Engineering and Induced Pluripotent Stem Cell Center (GEiC) at Washington University in St. Louis as described before (20). INS-1 832/13 cells were cultured in 11 mM glucose RPMI 1640 (Thermo Fisher Scientific, Cat# 11875) supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, Cat# 10099141), 1 mM sodium pyruvate (Corning, Cat# 25-000-CI), 90 µM β-mercaptoethanol (MilliporeSigma, Cat# M3148) and 100 U/mL penicillin-streptomycin (Thermo Fisher Scientific, Cat# 15140122). INS-1E cells were cultured in 11 mM RPMI 1640 supplemented with 10% FBS, 1 mM sodium pyruvate, 50 µM β-mercaptoethanol, 2 mM GlutaMAX (Thermo Fisher Scientific, Cat# 35050061), NEAA (Thermo Fisher Scientific, Cat# 11140050), 10 mM HEPES (Corning, Cat# 25-060-CI), and 100 U/mL penicillin-streptomycin. Rat IFN-γ, rat IL-1β (R&D Systems, Cat# 585-IF-100, 501-RL-010), and thapsigargin (MilliporeSigma, Cat# T9033) were used for the treatments.

129S6 whole body Wfs1-knockout mice were a kind gift from Dr. Sulev Kõks (University of Tartu) (25). Amino acids 360-890 of WFS1 protein were replaced with an in‐frame NLSLacZNeo cassette. Mouse genotypes were determined using multiplex PCR performed by Transnetyx (Cordova, TN). All animal experiments were performed according to procedures approved by the Institutional Animal Care and Use Committee at the Washington University School of Medicine (Protocol # 20-0334). All mice were housed in a pathogen-free animal facility, and food and water were provided ad libitum throughout the study. For the pancreatic section slides, 129S6 whole body Wfs1-knockout mice were euthanized in a carbon dioxide chamber followed by cervical dislocation. The mice were perfused with PBS and 4% paraformaldehyde (PFA) via the left ventricle. After perfusion, the pancreas was removed and fixed with 4% PFA for 48 h at 4°C. The fixed pancreas was gradually replaced by ethanol and finally kept in 70% ethanol. Paraffin embedding and sectioning were performed by Histology and Morphology Core at Musculoskeletal Research Center at Washington University in St. Louis. Slides were sectioned at 5 µm thickness. Microscopy imaging was performed using a Leica DM6B (Leica Microsystems, Germany). All quantitative analyses were conducted using Fiji ImageJ in a double-blinded fashion.

Paraffin-embedded pancreatic section slides were deparaffinized, and antigen retrieval was performed by submerging the slides in 10 mM sodium citrate solution (pH 6.0) for 30 min at 95°C. The tissues were permeabilized with phosphate buffer saline (PBS) containing 0.1% Triton X-100 for 30 min at room temperature. For the DAB stained samples, endogenous peroxidase was quenched using BLOXALL Blocking solution (VECTOR, Cat# SP-6000). Blocking was performed in 2% bovine serum albumin (BSA) for 1 h. The following primary antibodies were diluted in 0.2% BSA and incubated overnight at 4°C: anti-WFS1 (Proteintech, Cat# 1158-1-AP, 1:100), anti-Iba1 (Novus, Cat# NB100-1028, 1:50), anti-CD31 (abcam, Cat# ab124432, 1:100), and Alexa-Fluor 488 conjugated anti-insulin (Invitrogen, Cat# 53-9769-82, 1:100). After washing the primary antibodies 3 times with PBS, the tissues were incubated for 30 min at room temperature with ImmPRESS polymer reagent (VECTOR, Cat# MP-7401-50) for DAB staining, or AlexaFluor 594 donkey anti-rabbit IgG (Invitrogen, Cat# A21207) for immunofluorescent staining, and washed for 2 times with PBS. The color development for DAB staining was performed using peroxidase substrate solution (VECTOR, Cat# SK-4100) following the manufacturer’s protocol, and then the nucleus was stained by Vector Hematoxylin QS (VECTOR, Cat# H-3404). Tissues were mounted in Vectamount permanent mounting medium (VECTOR, Cat# H5000-60).

Deparaffinized and rehydrated slides were immersed in Bouin’s solution at 56°C for 1 h. Subsequently, the slides were washed with tap water for 5 min. The washed sections were stained in Weigert’s Hematoxylin solution for 10 min and then washed again with tap water for 5 min. Next, the slides were stained in Trichrome solution for 15 min and rinsed in 1% acetic acid for 1 min. Slides were dehydrated in alcohol twice for 1 min each, and sections were cleared in xylene twice for 1 min each. All the Masson’s trichrome staining procedures were performed by the Musculoskeletal Research Center at Washington University in St. Louis.

The antibody staining on formalin-fixed paraffin-embedded (FFPE) mouse pancreatic tissue was performed based on Fluidigm immunohistochemistry protocol. Briefly, the FFPE slides were deparaffinized followed by antigen retrieval using 10 mM sodium citrate (pH 6.0) for 30 min at 96°C. Blocking was performed using metal-free 3% BSA for 45 min prior to the addition of the primary antibody solution. The primary antibodies were anti-F4/80(BM8) (Fluidigm, Cat# 3146008B), anti-CD68(FA-11) (Biolegend, Cat# 137001), anti-CD163(EPR19518) (Abcam, Cat# ab182422), and anti-IL-1β/IL-1F2(Novus Biologicals, Cat# NBP1-42767). Metal conjugation was performed using MAXPAR^® X8 Multimetal Labeling Kit (Fluidigm, Cat# 201300), and all the antibodies were used in a 1:25 dilution. After overnight incubation of the antibody mixture at 4°C in a humidity chamber, the slide was washed with 0.2% Triton X-100 in PBS and then stained with 0.0025% Ruthenium Red for contrast staining. After the contrast staining, a DNA intercalator for nuclear identification was added. The slides were washed in distilled water and then air-dried before imaging. Antibody conjugation and CyTOF2/Helios imaging were performed by the Immunomonitoring Laboratory (IML) in Bursky Center for Human Immunology & Immunotherapy Programs (CHiiPs) at Washington University in St. Louis. Optimization of the multiplex panel involves assessing dual signal spillover into +1, +2, and +16 channels and lack of signal (false negative). Assessment of signal in other channels was performed using MCD Viewer software and generating a thumbnail image file for each metal channel. The imaging data were converted to TIFF images using HistoCAT++ (ver 3.0.0), and CellProfiler (ver 4.1.3) was used for the cell masking. Image visualization and data analysis were performed using HistoCAT (ver 1.761).

INS-1 832/13 cells were plated on 96-well flat clear bottom white polystyrene plates (Corning, Cat# 3610) and treated with or without the indicated reagents. After the treatment, CellTiter-Fluor Cell Viability Assay (Promega, Cat# G6080) reagent was added directly to cells and incubated for 30 min in the dark. After the fluorescence measurement, Caspase-Glo 3/7 Assay (Promega, Cat# G8090) reagent was added to the cells and incubated for 30 min. Fluorescence for cell viability and luminescence for caspase-3/7 activity was measured using Infinite M1000 plate reader (Tecan). Caspase-3/7 activity was normalized to cell viability according to the manufacturer’s protocol.

Cells were washed in cold PBS and immediately lysed on ice in Mammalian Protein Extraction Reagent (Thermo Fisher Scientific, Cat# 78501) supplemented with 1X cOmplete ^™ protease inhibitor cocktail (MilliporeSigma, Cat# 11873580001) and 1X PhosSTOP ^™ phosphatase inhibitor (MilliporeSigma, Cat# 4906845001) before centrifugation at 15,000 rpm for 10 min at 4°C. Nuclear and cytoplasmic extraction was performed using NE-PER™ (Thermo Fisher Scientific, Cat# 78833) according to the manufacturer’s protocol. Protein lysates were prepared using 4x Laemmli sample buffer (Bio-Rad Laboratories, Cat# 1610747) heated at 60°C for 15 min. The protein was resolved by SDS-PAGE and transferred to Immobilon-P PVDF membrane [MilliporeSigma, 0.2 µm pore size (Cat# ISEQ20200) for proinsulin and 0.45 µm pore size for others (Cat# IPVH00010)]. For detecting proinsulin, additional fixation and antigen retrieval steps were performed as reported previously (47). The antibodies used for immunoblotting were anti-WFS1 (Proteintech, Cat# 1158-1-AP), anti-LaminB1 (Proteintech, Cat# 12987-1-AP), and those purchased from Cell Signaling Technology: anti-cleaved caspase-3 (Cat# 9664), anti-insulin (Cat# 8138), anti-pPERK (Cat# 3179), anti-PERK (Cat# 3192), anti-IRE1α (Cat# 32945), anti-NF-κB (Cat# 8242), anti-GAPDH (Cat# 2118), anti-α-tubulin (Cat# 2125), and the secondary antibodies conjugated to horseradish peroxidase. Bands were detected by ECL Select (MilliporeSigma, Cat# RPN2235) and Bio-Rad ChemiDoc MP. Quantitative analyses were conducted using Fiji ImageJ.

Total RNA was extracted from INS-1E, INS-1 832/13, or isolated primary mouse macrophages using RNeasy Mini Kit (Qiagen, Cat# 74106) and reverse-transcribed using High-Capacity cDNA Reverse Transcription Kits (Thermo Fisher Scientific, Cat# 4368814). The qPCR was performed in 3-8 replicates for each sample. All the qPCR primer sequences used in this study are listed in Table S1 .

INS-1E or INS-1 832/13 cells were seeded in a 12-well plate and transfected with siRNA using RNAiMAX (Thermo Fisher Scientific, Cat# 13778150). ON-TARGET plus SMARTpool siRNA was used for knocking down rat Wfs1 (Dharmacon, Cat# L-087932-02). The siRNA against rat Perk, Ire1α, and non-targeting siRNA were purchased from MilliporeSigma (rat Perk, SASI_Rn01_00064452; rat Ire1α, SASI_Rn02_00372535; SIC001 for non-targeting). Each siRNA was used in 40 nM concentration, and RNAiMAX was used 1.0 µL against 1 pmol of siRNA. The cells were collected for the analysis 48 h after the siRNA transfection.

All statistics were performed using GraphPad Prism 9. Statistical analysis was performed by unpaired two-tailed Student’s t-test or one-way ANOVA followed by Tukey’s multiple comparisons test. P <0.05 was considered statistically significant.