The human EndoC-βH1 β-cell line [research resource identifier (RRID): CVCL_L909, Univercell-Biosolutions, France] was cultured in Matrigel/fibronectin-coated plates as previously described (27). Cells were cultured in DMEM containing 5.6 mmol/L glucose, 10 mmol/L nicotinamide, 5.5 μg/mL transferrin, 50 μmol/L 2-mercaptoethanol, 6.7 ng/mL selenite, 2% BSA fatty acid free, 100 U/mL penicillin, and 100 μg/mL streptomycin. We confirmed that cells were mycoplasma-free using the MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland).

Proinflammatory cytokine concentrations were selected according to previously established experiments in human β-cells (11, 28): recombinant human IFNα (PeproTech Inc., Rocky Hill, NJ) at 1000 U/mL; recombinant human IFNγ (PeproTech Inc., Rocky Hill, NJ) at 1000 U/mL; and recombinant human IL-1β (R&D Systems, Abingdon, UK) at 50 U/mL. Ruxolitinib, baricitinib, or deucravacitinib (Selleckchem, Planegg, Germany) were prepared in DMSO (used as vehicle) and cells were treated as indicated in the figures. Ruxolitinib and baricitinib concentrations were selected based on previous dose-response experiments (unpublished data). For treatments involving cytokines, 2% FBS was added to the culture medium.

The percentage of apoptosis was measured by fluorescence microscopy upon staining with the DNA-binding dyes Hoechst 33342 and propidium iodide (Sigma-Aldrich, Saint Louis, MO, USA) as described (29). At least 600 cells were counted for each experimental condition. Viability was assessed by two independent researchers, one of whom was unaware of sample identity, with >90% agreement between results.

Caspase 3/7 activity was determined using the Caspase-Glo^® 3/7 assay (Promega, Madison, WI, USA) following the manufacturer’s instructions. Briefly, upon incubation in 100 µL culture medium, cells were incubated with 100 µL Caspase-Glo^® 3/7 reagent at room temperature for 1 h before recording luminescence with a POLASTAR plate reader (BMG Labtech, Ortenberg, Germany).

The release of C-X-C motif chemokine ligand 10 (CXCL10) to the culture medium was detected using Human ProcartaPlex immunoassays (Invitrogen, Vienna, Austria) following the manufacturer’s recommendations. Reactions were read with a MagPix system (Luminex, Austin, TX, USA).

Cells were transfected using Lipofectamine 2000 (Invitrogen) with pRL-CMV encoding Renilla luciferase (Promega) and luciferase reporter constructs for either gamma-interferon activation site (GAS) (Panomics, Fremont, CA, USA) or IFN-stimulated regulatory element (ISRE) (kindly provided by Dr Izortze Santin, University of the Basque Country, Spain). After recovery, cells were treated with either IFNα for 2 h or IFNγ for 24 h (30). Luciferase activity was measured in a POLASTAR plate reader (BMG Labtech) using the Dual-Luciferase Reporter Assay System (Promega) and corrected for the luciferase activity of the internal control plasmid, i.e., pRL-CMV.

Poly(A)⁺ mRNA was extracted using Dynabeads mRNA DIRECT kit (Invitrogen) and cDNA synthesis was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time PCR was performed on the CFX96 Real Time System (Bio-Rad) as described (31) and the housekeeping gene β-actin was used to correct expression values. Of note, β-actin expression was not altered by the experimental conditions used herein. All primers used here are listed in Supplementary Table 1 .

Western blotting analysis was performed as described (32). Briefly, cells were washed with cold PBS and lysed in Laemmli buffer. Immunoblotting was performed using antibodies against phospho-STAT1 (P-STAT1), phospho-STAT2 (P-STAT2), STAT1, STAT2 (all at 1:1000 dilution), and α-tubulin (1:5000). Peroxidase-conjugated antibodies (1:5000) were used as secondary antibodies. SuperSignal West Femto chemiluminescent substrate (Thermo Scientific, Rockford, IL, USA) and ChemiDoc XRS+ (Bio-Rad Laboratories, Hercules, CA, USA) were used to detect bands.

Immunofluorescence was carried out as described (21, 33). First, cells were washed with cold PBS and fixed with 4% paraformaldehyde. Afterwards, cells were permeabilised and incubated with the mouse anti-MHC Class I (W6/32) antibody (1:1000). The Alexa Fluor 568 polyclonal goat anti-mouse IgG was used as secondary antibody and Hoechst 33342 for counterstaining. Coverslips were mounted with fluorescent mounting medium (Dako, Carpintera, CA, USA) and images were taken on a Zeiss LSM900 microscope with Airyscan 2 (Zeiss-Vision, Munich, Germany) and a x40 objective. Quantification was performed using ZEN (version 3.3; Zeiss-Vision) and open-source FIJI (version 2.0; https://fiji.sc) softwares.

All antibodies used here are listed in Supplementary Table 2 .

All the original, uncropped images representing immunoblots and microscopic photos are provided in the Supplementary Material.

After preincubation in modified Krebs-Ringer for 1 h, cells were sequentially stimulated with low (0 mmol/L) and high glucose (20 mmol/L) for 1 h (each stimulation) as previously described (34). Insulin secreted and insulin content from lysed cells were measured using a human insulin ELISA kit (Mercodia, Uppsala, Sweden) following the manufacturer’s instructions. The amount of secreted insulin as % of total insulin was calculated as previously described (35) and data were normalized to insulin secretion at 20 mmol/L glucose in vehicle-treated cells without IFNα (considered as 100%). See Supplementary Material for further details.

The GraphPad Prism 7.0 software (GraphPad Software, La Jolla, CA, USA) was used for statistical analyses. Data are shown as mean ± SEM of independent experiments (i.e. considering EndoC-βH1 cells from different passages as n = 1). The statistical significance of differences between groups was evaluated using one-way ANOVA followed by Dunnett’s test or two-way ANOVA followed by Sidak’s test or Dunnett’s test, as appropriate. Differences were considered statistically significant when p ≤ 0.05.

IFNα-mediated TYK2 activation leads to STAT1 and STAT2 phosphorylation, which will eventually upregulate several ISGs, including HLA-ABC, CXCL10, and MX1 ( Supplementary Figure 1A ). Pre-treatment with deucravacitinib inhibited IFNα-induced STAT1 and STAT2 phosphorylation in a dose-dependent manner, where deucravacitinib showed greater potency against IFNα-stimulated STAT1 phosphorylation ( Figures 1A, B ). We then selected two doses, 10 and 1000 nmol/L, for the follow-up experiments. Next, we examined how deucravacitinib affects the kinetics of IFNα-induced STAT activation. IFNα increased P-STAT1 and P-STAT2 levels, with a maximum effect at 1-4 h post-treatment and a return to baseline by 24 h ( Figures 1C, D ; Supplementary Figure 1B ). Although STAT1 and STAT2 protein levels were already upregulated by 8 h, STAT2 expression reached peak level at 16 h, while STAT1 expression was still increasing by 24 h ( Supplementary Figures 1C, D ). Exposure to 1000 nmol/L deucravacitinib abrogated the IFNα-stimulated STAT1 and STAT2 phosphorylation and protein expression, whereas 10 nmol/L deucravacitinib had only a minor effect ( Figures 1C, D and Supplementary Figures 1B–D ). Furthermore, IFNα-induced MHC class I protein overexpression was blocked by 1000 nmol/L deucravacitinib ( Figures 1E, F ). Finally, deucravacitinib did not affect β-cell viability nor changed glucose-stimulated insulin secretion and insulin content in the absence or presence of IFNα ( Supplementary Figures 1E–G ).

We compared deucravacitinib with ruxolitinib and baricitinib, two JAK1/JAK2 inhibitors previously tested in β-cells (13, 14). First, we measured the levels of P-STAT1 and P-STAT2 upon stimulation with IFNα or IFNγ ( Figures 2A–C ; Supplementary Figure 2 ). Ruxolitinib, baricitinib, and deucravacitinib prevented IFNα-stimulated increase in P-STAT1 and P-STAT2 levels ( Figure 2B ; Supplementary Figures 2A, B ). Nevertheless, deucravacitinib did not change IFNγ-induced STAT1 phosphorylation, whereas ruxolitinib and baricitinib blocked it ( Figure 2C ; Supplementary Figure 2C ). We next assessed ISRE and GAS reporter activities upon stimulation with IFNα or IFNγ ( Figures 2D, E ). While all three inhibitors abrogated IFNα-stimulated ISRE reporter activity ( Figure 2D ), IFNγ-induced GAS activation was barely affected by deucravacitinib ( Figure 2E ). As TYK2 is not involved in the IFNγ-triggered signaling pathway, the lack of deucravacitinib effect in IFNγ-treated cells is expected.

Assessement of the expression of some ISGs and ER stress markers showed that all three inhibitors prevented IFNα-induced upregulation of HLA-ABC, CXCL10, and MX1 in a dose-dependent manner ( Figures 2F–K ). Although ruxolitinib and baricitinib inhibited the mRNA expression of the ER stress markers C/EBP homologous protein (CHOP) and spliced isoform of XBP1 X-box binding protein 1 (XBP1s), only 10 nmol/L deucravacitinib reduced CHOP expression ( Figures 2I, J ). None of these inhibitors changed the expression of activating transcription factor 3 (ATF3) ( Figure 2K ).

Previous studies showed that a combination of IFNα + IL-1β, two cytokines that might be present in the islet milieu at early stages of insulitis, induces β-cell apoptosis, inflammation, and ER stress (11, 14, 23). Thus, we investigated whether deucravacitinib protects β-cells after IFNα + IL-1β exposure ( Figure 3A ). We observed that deucravacitinib completely prevented IFNα + IL-1β-induced apoptosis ( Figures 3B, C ). Moreover, deucravacitinib-treated cells showed reduced levels of P-STAT1 and STAT1 ( Figure 3D ; Supplementary Figures 3A, B ) as well as HLA-ABC, MX1, CHOP, and CXCL10 mRNA expression ( Figures 3E–G, J ). MHC class I protein expression and CXCL10 secretion were also decreased by TYK2 inhibition ( Figures 3H, I, K ).

We next evaluated whether deucravacitinib protects against cytokines that, as compared with IFNα, probably appear later in the progression of islet inflammation: IFNγ and IL-1β (36). After treatment for 24 h ( Figure 3A ), deucravacitinib inhibited IFNγ + IL-1β-induced apoptosis in a dose-dependent manner (60% and 92% protection at 10 and 1000 nmol/L, respectively) ( Figure 3L ). These results were confirmed by the caspase 3/7 activity ( Figure 3M ). Deucravacitinib did not affect IFNγ + IL-1β-induced STAT1 phosphorylation and protein expression ( Figure 3N ; Supplementary Figures 3C, D ) or HLA-ABC mRNA expression ( Figure 3O ); in fact, 1000 nmol/L deucravacitinib increased P-STAT1 levels ( Supplementary Figure 3C ). Conversely, deucravacitinib diminished MX1 and CXCL10 mRNA expression, whereas CHOP was reduced only at 10 nmol/L deucravacitinib ( Figures 3P–R ).

So far, we investigated whether pre-treatment with deucravacitinib prevents the effects of different cytokines in β-cells. Here, we assessed if deucravacitinib could abrogate these damaging effects. EndoC-βH1 cells were pre-treated with either IFNα + IL-1β or IFNγ + IL-1β for 24 h. Afterwards, 1000 nmol/L deucravacitinib was added for an additional 24 h still in the presence of cytokines ( Figure 4A ). Deucravacitinib partially decreased IFNα + IL-1β-induced apoptosis (60% decrease) ( Figure 4B ). IFNα + IL-1β-stimulated HLA-ABC mRNA expression remained unchanged in deucravacitinib-treated cells ( Figure 4D ), which agrees with previous data showing an IFNα-triggered long-lasting expression of HLA-ABC (13). STAT1 protein levels, CXCL10 secretion, and CHOP mRNA expression were reduced by 26-42% ( Figures 4C, F, H ; Supplementary Figure 3E ), while the expression of MX1 and CXCL10 was completely inhibited by deucravacitinib ( Figures 4E, G ).

Similarly to IFNα + IL-1β, deucravacitinib diminished IFNγ + IL-1β-induced apoptosis (64% decrease) but did not modify HLA-ABC mRNA expression ( Figures 4I, K ). Protein levels of STAT1 and CXCL10, however, were not altered by TYK2 inhibition, whereas a slight, non-significant 30% reduction was seen in CHOP expression ( Figures 4J, M, O ; Supplementary Figure 3F ). Expression of MX1 and CXCL10 was only partially affected by deucravacitinib under IFNγ + IL-1β conditions ( Figures 4L, N ).