Frozen stocks of EndoC-βH5 cells were purchased from Human Cell Design (Toulouse, France). The EndoC-βH5 cells originate from fetal pancreatic rudiments and were generated from differentiation and immortalization by transduction with the SV40 Large T-antigen (SV40LT) and human telomerase reverse transcriptase (hTERT) transgenes (9, 16). The EndoC-βH5 cells have furthermore undergone de-immortalization, including removal of the SV40L and hTERT transgenes, as well as maturation steps to generate functional and mature human beta cells (8, 9). Upon thawing, cells had a viability between 74-91% (n=24). The EndoC-βH5 cells were seeded in appropriate plates coated with β-coat (Human Cell Design) and maintained in Ultiβ1 cell culture medium (Human Cell Design) containing 100 U/mL penicillin and 100 µg/mL streptomycin (Life Technologies). Based on the experimental setup, cells were seeded in 24-well (200,000 cells/well) or 96-well (50,000 cells/well) plates in duplicate and incubated in a humidified incubator at 37°C with 5% CO₂. Media change was carried out every 2-3 days. Cells were incubated for up to 7 days in the presence or absence of recombinant human IL-1β (R&D systems, Minneapolis, MN, USA), recombinant human IFNγ (PeproTech, Rocky Hill, NJ, USA) and/or recombinant human TNFα (R&D Systems). For long-term treatments (4 and 7 days) with cytokines, the culture medium was replaced with fresh medium and cytokines every 2-3 days.

The ApoTox-Glo Triplex Assay (Promega, Madison, WI, USA) was used to measure live-cell protease activity, dead-cell protease activity, and caspase-3/7 activity as markers of viability, cytotoxicity, and apoptosis, respectively. The assay was used according to the manufacturer’s protocol. Luminescence and fluorescence were measured on an Infinite M200 PRO plate reader (Tecan, Männedorf, Switzerland).

Preparation of protein lysates, measurements of protein concentrations, and immunoblotting were performed as previously described (17). Antibodies used were anti-cleaved caspase-7 (#9491S, Cell Signaling, 1:500), anti-major histocompatibility complex (MHC)-I (the human leukocyte antigen (HLA)-A/B/C molecules (#ALX-805-711-C100, Enzo, Farmingdale, NY, USA, 1:2000), anti-phosho-c-Jun N-terminal kinase (P-JNK) (#9252, Cell Signaling, 1:1000), anti-nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor α (IκBα) (#J1512, Santa Cruz Biotechnology, Dallas, TX, USA, 1:500), anti-phospho-signal transducer and activator of transcription 1 (P-STAT1) (7649S, Cell Signaling, 1:1000), anti-Tubulin (T8203, Sigma-Aldrich, St. Louis, MO, USA, 1:2000), anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (#9482, Abcam, Cambridge, UK, 1:5000) and secondary HRP-conjugated anti-mouse (#7076, Cell Signaling, 1:1000) or anti-rabbit (#7074, Cell Signaling, 1:2000) IgG antibodies. Tubulin or GAPDH was used as internal controls for normalization.

Cells were seeded in black 96-well plates with clear bottom. After treatment for 48 hours, cells were fixed in 4% paraformaldehyde (Thermo Fisher) for 15 min at RT and permeabilized in 0.25% Triton X-100 (Sigma-Aldrich) for 20 min at RT. Blocking was done using 5% goat serum (Invitrogen) for 2 hours at RT. Staining of MHC-I (HLA-A/B/C molecules) was carried out using anti-MHC-I antibody (W6/32) (#ALX-805-711-C100, Enzo, 1:200 O/N at 4°C) and Alexa Fluor 594-conjugated anti-mouse IgG antibody (#A11005, Invitrogen, 1:2000 for 2 hours RT). Furthermore, apoptosis was assessed by the detection of fragmented DNA using the Click-iT-Plus TUNEL Assay (Invitrogen, Waltham, Massachusetts, USA) according to the manufactures protocol. Nuclei staining was done with 0.5 µg/mL DAPI (Sigma-Aldrich) or 5 µg/mL Hoechst 33342 (Invitrogen) in PBS. Fluorescence was measured using the ImageXpress PICO Automated Cell Imaging System (Molecular Devices (UK) Limited, Berkshire, UK) and analyzed with the acquisition software (CellReporterXpress 2.9.3). Fluorescence was visualized using a Nikon ECLIPSE Ti2 microscope (Nikon, Tokyo, Japan). Images were acquired at x20 and x60 magnification and prepared using FIJI software (version 1.49).

Gene expression levels of mRNAs and long non-coding RNAs (lncRNAs) shown previously to be regulators of beta-cell death and cellular impairment (13, 18, 19) were investigated by real-time quantitative PCR (qPCR). RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and cDNA was synthesized using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Real-time qPCR was carried out with TaqMan Assays and TaqMan Gene Expression Master Mix (Applied Biosystems, Waltham, MA, USA) using a CFX384 C1000 Thermal cycler (Bio-Rad). Relative expression levels were analyzed using the 2^-ΔΔCt method (20) with normalization to GAPDH. GAPDH was chosen as this was the most stable housekeeping gene measured in EndoC-βH5 cells by real-time qPCR among GAPDH, Peptidylprolyl isomerase A (PPIA), hypoxanthine Phosphoribosyltransferase 1 (HPRT1), beta-actin (ACTB) and 18S ribosomal RNA (data not shown).

The chemokines IFN-gamma-inducible protein 10 (IP-10)/C–X–C motif chemokine 10 (CXCL10), Thymus and activation-regulated chemokine (TARC)/CCL17, Eotaxin/CCL11, Eotaxin-3/CCL26, Macrophage Inflammatory Proteins (MIP)-1β/CCL3, MIP-1α/CCL4, IL-8/CXCL8, Monocyte chemoattractant protein (MCP)-1/CCL2, MCP-4/CCL13, and Macrophage-derived chemokine (MDC)/CCL22 were measured in the cell culture medium by the V-PLEX Chemokine Panel 1 Human Kit (Meso Scale Diagnostics LLC., Rockville, Maryland) using a MESO Quickplex SQ 120 instrument (Meso Scale Diagnostics).

Cells were exposed to cytokines for 48 hours prior to glucose-stimulated insulin secretion (GSIS), including 24 hours in standard Ultiβ1 cell culture medium and 24 hours in Ulti-ST starvation medium (Human Cell Design) with and without IL-1β, IFNγ and TNFα. Cells were washed twice and incubated for 60 min in βKrebs (Human Cell Design). Cells were then incubated for 40 min in βKrebs supplemented with 0 or 20 mM glucose. Supernatants were collected and analyzed for insulin. The cells were lysed in Cell Death Detection Lysis Buffer (Roche, Basel, Switzerland) for analysis of insulin content. Insulin was measured by ELISA (Mercodia, Uppsala, Sweden) and read on an Infinite M200 PRO plate reader (Tecan). Insulin was normalized to DNA content using CyQuant (Thermo Fisher Scientific) according to the manufacturer’s protocol.

Cells were seeded in pre-coated XFe96 Seahorse plates (40,000 cells/well). After cytokine exposure, cells were washed twice and transferred into 180 µl Krebs-Ringer-HEPES (KRH) buffer (135 mM NaCl, 3.6 mM KCl, 10 mM HEPES, 0.5 mM MgCl2, 1.5 mM CaCl2, 0.5 mM NaH2PO4, 2 mM glutamine, 0,1% (w/v) fatty acid free BSA (Sigma-Aldrich), pH adjusted to 7.4 at 37°C with NaOH). Cells were incubated for 1 hour at 37°C in a CO₂-free incubator prior to the experiment. After basal measurements, 11 mM glucose was injected to record respiratory response. Thereafter, 2 µM oligomycin (Sigma-Aldrich) was injected to inhibit ATP synthase and determine respiratory efficiency. Finally, 2 µM rotenone (Sigma-Aldrich), 2 µM antimycin A (Sigma-Aldrich), and 2.5 µM Hoechst 33342 was injected to inhibit mitochondrial respiration and stain for cell nuclei. Cell plates were immediately scanned using an ImageXpress PICO for cell count and quantified using the CellReporterXpress software. All data was normalized to cell count and corrected for non-mitochondrial respiration.

RNA-Seq libraries were sequenced on a DNBseq platform to produce 100 bp long paired-end reads. An average of ~40 million reads per replicate (EndoC-βH5 n=4) were obtained after pre-processing and QC (removal of low-quality reads and adapter sequences) using SOAPnuke software (21). Reads were aligned using TopHat (version 2.1.1) (22) to GChr38 genome with default parameters. Afterwards, reads were assigned to Ensembl version 107 gene annotation using htseq-count (version 0.9.1) (23) with default parameters. Minimum gene expression was defined based on a CPM cut-off of 0.5 in at least 4 samples. Differential gene expression analysis was performed on the two groups using the QL F-test in edgeR. Differentially expressed mRNAs and lncRNAs were identified using a cut-off of abs(log2FC) >0.5 and an FDR-adjusted p-value <0.05. We extracted the gene expression of pinpointed candidate genes from a total of 152 loci recently shown to be associated with type 1 diabetes (24). For cytokine receptor expression evaluation, TPM (transcripts per million) values were calculated for within-sample comparison of gene expression levels. All statistical analyses were conducted using Bioconductor packages in R.

Core analysis was performed using the Ingenuity Pathway Analysis (IPA) software (Qiagen, Valencia, CA, USA) on the differentially expressed genes to predict mechanistic and functional gene relationships, and identify top canonical pathways and molecular and cellular functions (24). The summarized upstream regulators, diseases, and mechanisms for the differentially expressed genes were visualized as a network and grouped using the IPA Path Designer Module.

Data are presented as means ± SEM. Graphs were constructed using GraphPad Prism 9. Statistical analyses were performed using one- or two-way ANOVA and/or one- or two-tailed paired t-test where appropriate. P-values <0.05 were considered statistically significant. Multiple testing was corrected using the Benjamini-Hochberg's correction.

To determine the sensitivity profile of the EndoC-βH5 cells, we initially examined the effects of cytokine exposure for 24, 48 and 96 hours, and after 7 days, using cytokine concentrations previously used by us and others on isolated human islets and EndoC-βH1 cells (50 U/mL of IL-1β, 1000 U/mL IFNγ and 1000 U/mL TNFα) (25, 26). Between 48 and 96 hours of cytokine exposure, caspase-3/7 activity and cytotoxicity were elevated and reached a ~2-fold increase after 4 days of cytokine exposure ( Figures 1A, B ). After 7 days of cytokine exposure, the caspase-3/7 activity and cytotoxicity were not further increased compared to 96 hours of cytokine treatment ( Figures 1A, B ). However, the viability of the cells declined with prolonged exposure to cytokines ( Figure 1C ). Microscopic inspection of cell morphology showed severe changes and detachment of cells after 7 days of cytokine exposure indicative of extensive cell death ( Supplementary Figure S1 ).

Next, we studied the responses to various doses of individual and combinations of cytokines after 96 hours of exposure. As seen in Figures 1D, E , caspase-3/7 activity and cytotoxicity were dose-dependently induced by IFNγ alone. In contrast, IL-1β alone had no effects. When combined with IFNγ, neither IL-1β nor TNFα had synergistic effects on IFNγ-induced cell death. The reduction in cell viability by cytokine exposure was modest at this time point and did not reach statistical significance in any of the conditions ( Supplementary Figure S2 ). These results indicate that only IFNγ exerts death-promoting and toxic effects on EndoC-βH5 cells.

End-stage apoptosis was evaluated by DNA fragmentation labelling with TUNEL. After 96 hours of cytokine exposure, end-stage apoptosis was elevated by ~1.6-fold in cells treated with either IFNγ alone or a combination of IL-1β, IFNγ, and TNFα ( Figure 1F ). We observed around 20% TUNEL-positive cells in the untreated control condition indicative of a relatively high basal apoptosis rate ( Figure 1F ). These results further support that IFNγ is the main cytotoxic cytokine among the tested ones in EndoC-βH5 cells.

To study cytokine signal transduction in EndoC-βH5 cells, we investigated key signaling pathways of each cytokine. We first confirmed that cytokine exposure leads to apoptotic caspase activation shown by immunoblotting of cleaved, activated caspase 7 ( Figure 2A ). We then looked at proximal cytokine signal transduction by immunoblotting of the cellular content of activated, phosphorylated signal transducer of activated transcription 1 (P-STAT1) as a marker of IFNγ signaling, and P-JNK, and inhibitory κBα (IκBα) as markers of IL-1β and TNFα signaling (27). Cytokine treatment for 30 min strongly enhanced P-STAT1, whereas no effects were seen on P-JNK and IκBα ( Figure 2B ). INS-1E cell lysates were used as positive control and showed degradation of IκBα, elevated P-STAT1 and P-JNK after 30 min of cytokine exposure ( Figure 2B ).

A distal signaling event in cytokine-exposed beta cells is the induction of endoplasmic reticulum (ER) stress (13). Individual cytokine exposure showed that only IFNγ induced ER stress as revealed by increased expression of the ER stress markers DNA damage-inducible transcript 3 (DDIT3)/C/EBP homologous protein (CHOP) (28) and Harakiri (HRK)/Death Protein 5 (DP5) after 48 hours of cytokine exposure (29) ( Figures 2C, D ). Furthermore, increased expression of CXCL10, known as a key event in IFNγ signaling in beta-cell dysfunction and apoptosis (30), was greatly induced by IFNγ exposure ( Figure 2E ). These results suggest that at the time point examined, only IFNγ is capable of inducing intracellular signaling in EndoC-βH5 cells.

We next investigated MHC class I presence on EndoC-βH5 cells, a hallmark of immune-mediated beta-cell destruction in type 1 diabetes (18, 27, 31), in the EndoC-βH5 cells in response to cytokines. By immunoblotting, we found that MHC-I protein was induced after 96 hours of cytokine treatment ( Figure 3A ). Consistent with this, cytokines upregulated the expression of HLA-A/B/C which encode MHC-I ( Figures 3B–D ). Again, these effects were driven exclusively by IFNγ ( Figures 3E–G ). Induction of MHC-I was confirmed by immunofluorescence staining ( Figure 3H ). These findings demonstrate that IFNγ upregulate MHC-I in EndoC-βH5 cells.

We examined if cytokines increase the secretion of chemokines which play important roles in the immune cell – beta-cell crosstalk in type 1 diabetes (32). The accumulated levels of 10 chemokines were measured in the cell culture medium after 48 hours of exposure to IL-1β, IFNγ, and TNFα in combination or alone. All chemokines measured were increased in response to the treatment with the cytokine mixture, and for treatment with the individual cytokines, IFNγ was the sole cytokine capable of increasing chemokine secretion ( Figures 4A–L and Supplementary Figures S3A–H ). These data show that EndoC-βH5 cells secrete chemokines upon treatment with IFNγ.

To investigate the impact of cytokines on the insulin secretory capacity of EndoC-βH5 cells, GSIS experiments were performed following exposure to IL-1β, IFNγ and TNFα for 48 hours. In untreated control cells, there was a significant increase in GSIS corresponding to a fold change of >6 ( Figure 4M ). This induction was diminished by ~30% by cytokine exposure ( Figure 4M ). The effect correlated with tendencies towards reduced cellular insulin content ( Figure 4N ) and decreased expression of INS and PDX1 ( Supplementary Figures S4A, B ). These results demonstrate that cytokines cause functional impairment of EndoC-βH5 cells.

To investigate the effects of cytokines on mitochondrial function of EndoC-βH5 cells, we measured the oxygen consumption rate (OCR) using Seahorse extracellular flux technology following cytokine treatments ( Figure 4O ). Basal respiration decreased after cytokine exposure in all conditions except IL-1β alone ( Figure 4P ). Glucose oxidation, proton leak and ATP production were decreased in response to treatment with a combination of the cytokines. The glucose response was enhanced in cells exposed to the combined cytokines, as compared to control ( Figure 4Q ). No effects were observed for coupling efficiency ( Figure 4R ). The results suggest that cytokines, and particularly IFNγ, alter mitochondrial respiration in EndoC-βH5 cells.

We next studied the impact of cytokines on the global gene expression profile of EndoC-βH5 cells. Cells were treated with or without IL-1β, IFNγ and TNFα for 48 hours followed by stranded RNA sequencing. On average, library sizes of ~40 million reads per sample were obtained ( Supplementary Figure S5A ). Multidimensional scaling (MDS) analysis showed a high separation between control and cytokine-treated samples ( Supplementary Figure S5B ). A total of 16,525 genes were detected of which more than one-third (6,000 genes) were differentially expressed in response to cytokines ( Figure 5A ; cut off of abs(log2FC) >0.5 and FDR adjusted p-value <0.05). The top 10 up- and downregulated coding (mRNA) and lncRNA genes are shown in Table 1 . As lncRNAs are emerging as important regulators of cellular functions and disease, we wished to validate the expression of selected lncRNAs with roles in diabetes, e.g. taurine-upregulated gene 1 (TUG1), LNC78/TCL1 upstream neural differentiation-associated RNA (TUNAR), maternally expressed gene 3 (MEG3), growth arrest-specific transcript 5 (GAS5), and LINC25/LINC01370, known for their regulation of transcription factors and/or effects on beta-cell function and apoptosis (33–37). EndoC-βH5 cells exposed to cytokines for 48 hours showed significantly decreased expression of the lncRNAs TUG1 and TUNAR ( Supplementary Figures S6A, B ). The lncRNAs MEG3 and GAS5 were not significantly affected by cytokine exposure ( Supplementary Figures S6C, D ). LINC-25 was not expressed (data not shown).

The fact that we failed to observe any functional and signaling effects of IL-1β and TNFα in EndoC-βH5 cells, prompted us to retrieve information about the basal expression of the receptors for the cytokines used from the RNA-Seq data. Retrieval of the TPM values for interleukin 1 receptor, type 1 (IL1R1) and 2 (IL1R2), interleukin 1 receptor accessory protein (IL1RAP), and TNF receptor superfamily member 1A (TNFRSF1A) and 1B (TNFRSF1B) revealed very low expression levels with TPM values <1 ( Supplementary Table S1 ). In contrast, the IFNγ receptors IFNGR1 and IFNGR2 were expressed at a higher level with TPM values of ~3-10. TPM values of key beta-cell identity genes (INS, PDX1, MAFA, and NKX6-1) are included in the Supplementary Table S1 for comparison. A low expression level of IL1R1 compared to IFNGR1 was confirmed by qPCR, which revealed Ct values of 35-38 for IL1R1 and Ct values of 31-33 for IFNGR1 (data not shown).

Using the IPA tool, we were able to predict the top relevant disease-related and molecular interactions of the differentially expressed genes after exposure to cytokines. The Core analysis revealed that the top-affected canonical pathway was ‘Senescence Pathway’, and the top-affected molecular and cellular function was ‘Cell Death and Survival’ ( Table 2 ). Most significant upstream regulators, diseases, and mechanisms of the differentially expressed genes are represented in a graphical summary network ( Figure 5B ). Activated pathways in this network included ‘Diabetes mellitus’, ‘Senescence Pathway’, ‘Endocrine pancreatic dysfunction’, ‘Interferon Signaling’, and ‘Organismal death’.

Finally, we extracted gene expression data of pinpointed candidate genes from 152 type 1 diabetes-associated loci and identified 42 candidate genes that were differentially expressed upon cytokine exposure. Twenty of these were significantly upregulated and 22 significantly downregulated ( Figure 5C ). The data obtained demonstrate that cytokine treatment causes profound changes to the transcriptome of EndoC-βH5 cells.