Peptidylarginine Deiminase Inhibition Prevents Diabetes Development in NOD Mice

Emerging evidence demonstrates a role for posttranslational modifications in the pathogenesis of type 1 diabetes (1–10). One of these posttranslational modifications is citrullination, the conversion of arginine into citrulline, mediated by peptidylarginine deiminases (PADs), of which five isozymes have been described (11). The loss of a positively charged arginine in peptides enhances their binding affinity to type 1 diabetes predisposing HLA-DR4 molecules (12,13) and may thereby elicit autoreactive T-cell responses. Indeed, it has been shown that autoreactive CD4⁺ T cells of patients with type 1 diabetes recognize citrullinated GAD65 (4). Moreover, our group showed that citrullinated glucose-regulated protein 78 (GRP78) is an autoantigen in NOD mice and in human type 1 diabetes (5,6,10). We also provided direct evidence for cytokine-induced citrullination of GRP78 in INS-1E β-cells (5) and human islets (10), indicating that GRP78 can be citrullinated in β-cells in the absence of immune cells. In addition, we showed that inflammatory cytokines induce translocation of GRP78 from the endoplasmic reticulum to the β-cell membrane and its subsequent secretion (5,14), providing the ideal environment to become citrullinated in the extracellular space (15). Although it remains unknown which cells are responsible for protein citrullination in the pancreas, Padi2 is highly expressed in NOD islets relative to C57BL/6 islets (5,16). Also, increased levels of PAD4 in neutrophils of individuals with type 1 diabetes have been reported (17,18). Interestingly, PAD4 is crucial for the formation of neutrophil extracellular traps (NETs) (19), a process associated with the pathogenesis of type 1 diabetes (20,21). Neutrophils are the first immune cells to infiltrate the islets of NOD mice (20) and have been found in the human pancreas before disease onset (21), with a substantial fraction forming NETs (21). Moreover, various studies targeting neutrophil activity or NETs showed a marked protection against diabetes development in NOD mice (20,22–24).

Citrullination does not occur exclusively in type 1 diabetes. In other autoimmune diseases, citrullinated proteins are present in inflamed target tissues and are associated with the break in immune tolerance (25). In rheumatoid arthritis, for example, citrullinated autoantigens are known to be causative (25). Several studies have reported improved disease outcome without any signs of in vivo toxicity (26–30) using pan-PAD inhibitors, such as Cl-amidine or BB-Cl-amidine, in animal models for rheumatoid arthritis (26,31,32), multiple sclerosis (33), systemic lupus erythematosus (30,34), and ulcerative colitis (27–29). Disease improvement, through direct inhibition of citrullination, was shown to be linked to decreased NET formation (NETosis) (34–37), modulation of dendritic cell function (37), and a shift in Th1/Th2 profiles (31). On the basis of these results, PAD inhibition is gaining interest as a strategy to treat or prevent autoimmune diseases that are associated with abnormal PAD activity.

The mechanism by which BB-Cl-amidine and its mother compound Cl-amidine inactivates PAD has been well described (38). Both compounds irreversibly inactivate PAD enzymes through covalent modification of a conserved cysteine in the active site of the PAD enzymes, thereby having an effect on all PAD enzymes. When PAD enzymes become activated, they undergo a calcium-dependent conformational change that moves a nucleophilic cysteine residue into the active site, and only then, this cysteine is available for the PAD inhibitor (38). Compared with Cl-amidine, BB-Cl-amidine has a longer in vivo half-life (1.75 h vs. ∼15 min), a higher cellular potency (half-maximal effective concentration of 8.8 ± 0.6 μmol/L vs. >200 μmol/L for Cl-amidine), and a comparable selectivity for the different PAD enzymes (34). With no reports on the potential efficacy of PAD inhibition in type 1 diabetes models, we here evaluated the effect of BB-Cl-amidine in NOD mice. We observed full protection against diabetes development associated with a decrease in citrullination and in autoantibody titers against citrullinated GRP78. Effects on innate and adaptive immune responses were also observed, with decreased NETosis of bone marrow–derived neutrophils, increased serum Th2 cytokines, and regulatory T cells (T_reg) in peripheral tissues and a decrease in effector memory CD4⁺ T cells (T_EM) and interferon-γ (IFN-γ)–producing CD4⁺ and CD8⁺ T cells in the pancreas. These findings point to a role for dysregulated PAD activity and citrullination in autoimmune diabetes and provide initial insight into PAD inhibition as a potential therapeutic strategy.

NOD mice were inbred and housed under semibarrier conditions in our animal facility. Eight-week-old female NOD mice were used. Treatments involved subcutaneous injections with BB-Cl-amidine (1 μg/g body weight) or vehicle (25% DMSO in PBS) six times per week until 25 weeks of age for diabetes incidence or until 13 weeks of age for mechanistic studies (29,31,34,39). Experiments were approved by the institutional animal ethics committee of KU Leuven.

Islets were isolated from 10-week-old NOD mice using collagenase (40). After overnight resting, the islets were dispersed into single-cell suspensions by incubation for 90 s at 37°C in 0.025% trypsin-EDTA (Thermo Fisher Scientific). Cells were surface stained for cell sorting by flow cytometry as previously described (40). A detailed protocol is provided in the Supplementary Material.

RNA isolation and cDNA generation were performed as previously described (40). SYBR green quantitative PCR was performed on a QuantStudio 3 (Thermo Fisher Scientific) per manufacturer’s instructions. Gene expression levels of insulin 1 (Ins1), glucagon (Glc), CD45, Padi2, and Padi4 were normalized to the geometric mean of three housekeeping genes (actin, RPL27, and HPRT) and analyzed by relative quantification using the 2^−ΔCt method.

Five-micrometer sections of paraffin-embedded pancreata were collected at 50-μm intervals. Heat-mediated antigen retrieval was performed on deparaffinized sections using citrate buffer (pH 6.1, 20 min; Dako). To detect citrullination, endogenous peroxidase activity was blocked with 0.5% H₂O₂ in methanol for 20 min. Sections were incubated with rabbit anticitrulline (1:50; Abcam) for 1 h and with anti-rabbit horseradish peroxidase (Envision; Dako) for 30 min. Sections were incubated with 3,3′-diaminobenzidine (Dako), counterstained with hematoxylin (2 min), dehydrated, and mounted with DEPEX. For insulin detection, sections were blocked with 10% normal goat serum (Dako) for 20 min and incubated with guinea pig anti-insulin (1:150; Dako) for 1 h and Alexa Fluor 633–conjugated goat anti-guinea pig (1:500; Life Sciences) for 30 min. Sections were counterstained with Hoechst (1:2,000; Thermo Fisher Scientific) and mounted with Mowiol (Polysciences). All incubations were done at room temperature (RT). Imaging was performed using an Olympus BX41 or Zeiss LSM 780 confocal nonlinear optical microscope.

Mice were tested for clinical signs of diabetes by evaluating glucose levels in the urine with Diastix Reagent Strips (Bayer). As soon as glucosuria was measured, further follow-up was done by measuring blood glucose levels. Mice were considered diabetic when glucosuria was present and blood glucose levels were >200 mg/dL for 2 consecutive days.

At least 25 islets per mouse were scored for lymphocytic infiltration in a blinded manner using hematoxylin-eosin–stained sections.

Pancreata were homogenized in acidic ethanol (91% ethanol, 9% 1 mol/L H₃PO₄) at 4°C overnight and sonicated. Insulin content was determined in the supernatant by ELISA (Mercodia) and normalized to the weight of the pancreas.

Protein samples (50 μg) were separated on 4–12% Bis-Tris gels (Invitrogen) and blotted onto polyvinylidene fluoride membranes (Hybond-ECL; GE Healthcare). To detect citrullination, the Anti-Citrulline (Modified) Detection Kit (Millipore) was used. Membranes were incubated with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and revealed using the ImageQuant LAS 500 system (GE Healthcare). Blots were normalized by reprobing for GAPDH (Thermo Fisher Scientific).

DMSO- (n = 4) and BB-Cl-amidine–treated (n = 4) mouse pancreata were pooled in buffer (50 mmol/L HEPES, pH 7.6, 0.5% NP40, and 1 mmol/L phenylmethylsulfonyl fluoride) and homogenized using a tissue homogenizer (maximum speed 3 × 20 s at 10-min intervals). The lysed tissue samples were centrifuged (12,000 rpm, 30 min) at 4°C. The supernatants were isolated, and the protein concentration was determined using a detergent compatible assay. Equal protein concentrations from DMSO- and BB-Cl-amidine–treated samples were labeled with biotin-phenylglyoxal (PG) in quadruplicate as previously described (41). A detailed protocol is provided in the Supplementary Material.

A total of eight peptide digests were obtained from quadruplicate biotin-PG labeling of the DMSO- and BB-Cl-amidine–treated samples. The samples were labeled using a tandem mass tag labeling system and prepared for liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis on a nanoACQUITY UPLC (Waters Corporation) coupled to an Orbitrap Fusion Lumos Tribrid (Thermo Fisher Scientific) mass spectrometer. The raw data were processed using Proteome Discoverer version 2.1.1.21 (Thermo Fisher Scientific) and searched against the SwissProt murine (downloaded July 2019) database using Mascot version 2.6.2 (Matrix Science). A detailed protocol is provided in the Supplementary Material.

Mouse serum autoantibody levels against citrullinated GRP78 were determined by an in-house–developed Meso Scale Discovery (MSD) ELISA. In vitro citrullinated GRP78 was generated as previously described (5). MULTI-ARRAY 96 plates (MSD) were coated (overnight, 4°C) with 5 μg/mL in vitro citrullinated GRP78 in PBS (pH 7) or with 3% skimmed milk in PBS (uncoated control). Blocking was performed with 3% skimmed milk in PBS with 0.5% Tween 20 (2 h at RT). Mouse serum was diluted 1:5 in blocking solution and incubated for 2 h at RT. Wells were incubated for 2 h at RT with SULFO-TAG–labeled goat anti-mouse IgG. Read buffer was added, and detection was done on an MSD reader. To exclude nonspecific signals, each sample had its background signal (serum binding in uncoated well) subtracted from the reading.

Bone marrow–derived cells were harvested from the humeri, tibias, and femurs, and neutrophils were isolated using the Neutrophil Isolation Kit (Miltenyi Biotec). Neutrophils were cultured on coverslips precoated with 0.01% Poly L-lysine (Sigma-Aldrich) for 3 h at 37°C in 5% CO₂. Paraformaldehyde-fixed neutrophils were stained with Hoechst (Thermo Fisher Scientific), and NETosis was quantified blindly using a Nikon Eclipse TI microscope. Percentage of NET-releasing cells was calculated by normalizing to the total number of cells. Images were taken on a Zeiss LSM 880 confocal Airyscan microscope.

Serum levels of IFN-γ, interleukin (IL)-4, and IL-10 were determined by a single measurement using a V-plex assay from MSD. All concentrations were above the lower limit of detection.

Blood was incubated three times with NH₄Cl to lyse red blood cells. Single-cell suspensions from spleen were obtained by mechanical disruption, while the pancreas was enzymatically digested for 15 min at 37°C in RPMI medium (72400; Thermo Fisher Scientific) supplemented with 5% bovine serum, 0.05 mmol/L β-mercaptoethanol, 100 units/mL penicillin, 100 units/mL streptomycin, 1 mg/mL collagenase VII, and 0.02 mg/mL DNase I. For phenotypic characterization of T_reg, T_EM, central memory T-cell (T_CM), and naive T-cell (T_naive) subsets, cell suspensions were stained using a Zombie Yellow Fixable Viability Kit (BioLegend) and with antibodies against CD4 (BD Biosciences), CD3, CD8, CD44, CD62L, CD25, and Foxp3 (eBioscience). Intracellular staining was performed with Foxp3/Transcription Factor Staining Buffer Set (eBioscience). To evaluate the effect of BB-Cl-amidine treatment on cytokine production, 5 × 10⁵ cells from the spleen and pancreas were stimulated in vitro with 0.025 μg/mL phorbol 12-myristate 13-acetate (PMA) (Sigma-Aldrich) and 0.25 μg/mL ionomycin (Sigma-Aldrich) in the presence of GolgiStop (1:500; BD Biosciences) for 4 h at 37°C in 5% CO₂. Next, cells were incubated with antibodies against CD4 and CD8, treated with Cytofix/Cytoperm (BD Biosciences), and stained with anti-IFN-γ antibody (eBioscience). Samples were acquired on a BD FACS Canto II and analyzed with FlowJo software (TreeStar). The percentage of cytokine-producing cells was determined, and the expression level of the cytokine was assessed using geometric mean fluorescence intensity (MFI). For each experiment, the geometric MFI of each sample was normalized to the average geometric MFI of the DMSO group (MFI fold change).

Data were analyzed using GraphPad Prism 8 or Scaffold software. Diabetes incidence was evaluated by Kaplan-Meier survival analysis with Mantel-Cox log-rank test. Data were expressed as mean ± SEM, and after exclusion of outliers, data were analyzed by one-way ANOVA, unpaired t test with Welch correction, or Mann-Whitney U test, as indicated. P < 0.05 was considered statistically significant.

All data sets generated during the current study are available from the corresponding author upon request.

Although we have previously shown that islets of 3- and 10-week-old NOD mice have elevated PAD activity levels and enhanced Padi2 mRNA expression compared with C57BL/6 islets (5), it remains to be elucidated which cell population within the islets is responsible for this high PAD activity. To investigate this, we performed flow cytometry sorting on 10-week-old NOD islet cells, making use of an antibody panel consisting of CD71, CD24, CD49f, and CD45 to enable isolation of highly purified α- and β-cells (40) as well as immune cells. The purity of immune cell fraction was confirmed by marginal expression of Ins1 and Glc. Purity of α- and β-cell fractions was confirmed by complete absence of leukocyte marker CD45 in both fractions and marginal expression levels of Glc (in β-cells) and Ins1 (in α-cells) (Fig. 1A–C). In immune as well as in α- and β-cells, Padi2 was clearly expressed (Fig. 1D), with a trend to higher levels in immune and β-cells compared with α-cells. Of note, Padi4 mRNA levels were below the detection limit (data not shown). In line with this, immunohistochemical analysis of citrullination in pancreas sections of NOD mice demonstrated intense staining in infiltrated immune cells, while endocrine islet cells stained weakly positive. Parallel staining in age-matched C57BL/6 also showed citrullination in the endocrine islet cells (Fig. 1E and Supplementary Fig. 1). These results indicate that citrullination occurs in immune cells infiltrated in the islets and in endocrine α- and β-cells, even in nonpathological conditions.

To examine the role of citrullination on diabetes development, NOD mice were treated with BB-Cl-amidine. Treatment was started at 8 weeks of age, a time point at which insulitis is already ongoing in NOD mice. BB-Cl-amidine treatment fully prevented diabetes development, with all mice free of diabetes until 25 weeks of age, against 44% diabetes free in DMSO-treated mice (P < 0.001) (Fig. 2A). Therapy was well tolerated as confirmed by normal weight curves (Supplementary Fig. 2). Evaluation of pancreatic insulin content at 13 weeks of age, before the onset of diabetes, showed a fourfold increase of insulin in the pancreata of BB-Cl-amidine–treated mice compared with the control (P < 0.01) (Fig. 2B and C), pointing to preserved insulin production in BB-Cl-amidine–treated mice.

To evaluate the direct effect of BB-Cl-amidine on inhibition of citrullination, we performed immunohistochemical staining of citrullinated protein residues on pancreatic sections, showing a reduced staining intensity in BB-Cl-amidine–treated compared with DMSO-treated mice (Fig. 3A and Supplementary Fig. 3). Quantification of the overall level of citrullinated proteins by Western blotting in the pancreata of NOD mice treated with BB-Cl-amidine and vehicle demonstrated that PAD inhibition reduced the total level of protein citrullination by ∼50% (P < 0.01) (Fig. 3B and C). This finding was further confirmed and extended by more quantitative LC-MS/MS analysis. A total of 1,003 citrullinated proteins were identified in pancreata from mice treated with BB-Cl-amidine and vehicle, of which 765 were significantly decreased in abundance by the treatment regimen (Fig. 3D and Supplementary Tables 1 and 2). Interestingly, citrullinated GRP78, reported by our group as an autoantigen in murine and human type 1 diabetes (5,10), was the most significantly reduced citrullinated protein, with a decrease of 2.66-fold (P < 0.0001). We have previously reported the generation of autoantibodies against citrullinated GRP78 in NOD mice (5). In line with this, we observed here that BB-Cl-amidine treatment decreased autoantibody titers against citrullinated GRP78, both at 13 (P < 0.05) (Fig. 3E) and at 25 weeks of age (P < 0.05) (Fig. 3F).

Neutrophils have been suggested to play an important role in the pathogenesis of autoimmune diabetes because they are the first immune cells to infiltrate the islets of NOD mice (20). Also, in patients with type 1 diabetes, neutrophil infiltration is observed in the pancreas, even before disease onset, with some degree of NETosis (21). Since citrullination is a prerequisite for the initiation of NETosis, we questioned here whether BB-Cl-amidine treatment could have an impact on the propensity of neutrophils to form NETs. To this end, spontaneous NETosis was assessed in bone marrow–derived neutrophils isolated from mice treated with BB-Cl-amidine or vehicle. NETosis was significantly reduced in neutrophils from mice treated with BB-Cl-amidine (P < 0.05) (Fig. 4). This finding shows that systemic administration of BB-Cl-amidine has an effect on the capacity of neutrophils to form NETs.

The finding of preserved insulin production by BB-Cl-amidine led us to investigate the degree and phenotype of immune infiltration. To this end, immunohistochemical evaluation of islet immune infiltration at 13 weeks of age showed only a moderate overall decrease of insulitis in the BB-Cl-amidine group (Fig. 5A), suggesting a less destructive form of insulitis. To evaluate this further, we measured in vitro cytokine production of the immune cells isolated from the pancreas. Although the overall frequency of IFN-γ–producing T cells was low in both groups, mice treated with BB-Cl-amidine presented a significantly lower percentage of IFN-γ–producing CD4⁺ and CD8⁺ T cells (P < 0.01 and P < 0.05, respectively) (Fig. 5B and C). Evaluation of the phenotype of immune cells in the pancreas (representative gating strategy in Supplementary Fig. 4A) revealed a significant decrease in the percentage of CD4⁺ T_EM cells (P < 0.05) (Fig. 5E) in BB-Cl-amidine–treated mice, whereas no changes were observed in the frequency of CD8⁺ T_EM and T_reg cells (Fig. 5D). No differences were observed in CD4⁺ and CD8⁺ T_naive and T_CM subsets (Supplementary Fig. 5). Collectively, these data indicate that although there was only a minor change in the overall level of insulitis, BB-Cl-amidine alters the phenotype of infiltrating immune cells as demonstrated by the decrease in Th1 cytokine production and reduction of T_EM locally in the pancreas.

Next, we investigated the effect of BB-Cl-amidine treatment on Th1 and Th2 profiles and immune cell phenotypes in the periphery because BB-Cl-amidine has been shown to act through this pathway in other disease models (31). First, we measured serum levels of Th1 and Th2 cytokines in NOD mice treated with BB-Cl-amidine or vehicle. Whereas no significant changes were measured in circulating IFN-γ titers (Fig. 6A), IL-4 and IL-10 titers were significantly increased in the BB-Cl-amidine–treated group (P < 0.01 for both) (Fig. 6B and C). This resulted in an overall reduction in the IFN-γ/IL-10 ratio (P < 0.05) (42) (Fig. 6D). In the spleens of BB-Cl-amidine–treated mice, there was a decrease in IFN-γ production indicated by the reduced frequency of IFN-γ–producing CD4⁺ T cells (P < 0.01) (Fig. 6E) as well as the lower expression of IFN-γ in IFN-γ–producing CD4⁺ and CD8⁺ T-cell populations (P < 0.0001 for both) (Fig. 6F and H). To specifically investigate the effect on the cellular phenotype of immune cell subsets, we used flow cytometry to study T-cell subsets in the blood and spleen (representative gating strategy in Supplementary Fig. 4B and C). This revealed that BB-Cl-amidine induced a significant increase in the percentage of T_reg in blood (P < 0.05) (Fig. 7A) and spleen (P < 0.01) (Fig. 7D). With regard to T_EM (CD4⁺ and CD8⁺) in the periphery, there was a significant decrease in the percentage of CD4⁺ T_EM cells in the blood and of splenic CD8⁺ T_EM cells (P < 0.05 and P < 0.001, respectively) (Fig. 7B and F). No changes were observed in CD4⁺ and CD8⁺ T_CM populations in the blood and spleen, whereas CD4⁺ and CD8⁺ T_naive in blood (both P < 0.05) (Supplementary Fig. 6A and C) and CD8⁺ T_naive in spleen (P < 0.01) (Supplementary Fig. 6G) were increased in BB-Cl-amidine–treated mice. These findings suggest that in the periphery, BB-Cl-amidine induces a shift toward Th2 cytokine production, a decrease in T_EM, and an induction of T_reg.

This study describes the effect of the pan-PAD inhibitor BB-Cl-amidine, showing a complete prevention of diabetes development in NOD mice. This finding highlights the involvement of inflammation-induced citrullination in diabetes development and adds type 1 diabetes to the list of autoimmune diseases in which citrullination participates in disease pathogenesis, a list that includes rheumatoid arthritis (26,31,32), multiple sclerosis (33), lupus (30,34), and ulcerative colitis (27–29). This observation may have significant implications for the development of new therapeutic strategies to mitigate the progression and/or development of type 1 diabetes.

We here provide evidence that the previously reported high expression of Padi2 in islets of NOD mice (5,16) originates both from the endocrine cells, including α- and β-cells, and from the infiltrating immune cells, thereby conclusively demonstrating that β-cells themselves have the capacity to citrullinate. This finding formed the basis for evaluating the effect of PAD inhibition in NOD mice. In the current treatment regimen, we started treatment at 8 weeks of age, a time point at which insulitis is already ongoing. The observation that inhibition of PAD activity, and thus citrullination, protected NOD mice against diabetes development suggests that citrullination may play a role in the amplification of the disease rather than being an initial trigger in breaking of immune tolerance. This strengthens our earlier findings of minor autoreactivity against citrullinated GRP78 in prediabetic NOD mice (5) and the increased occurrence of autoantibodies and autoreactive T-cell responses in patients with long-standing type 1 diabetes versus those with new-onset diabetes (10). This view also fits well with the knowledge that citrullination is mostly associated with inflammation, thereby playing a role in several autoimmune diseases where citrullination takes place in the inflamed target tissues (26–34). The low levels of citrullination in islets of C57BL/6 mice further demonstrate that citrullination is present already under physiological conditions; however, the degree of citrullination is markedly increased in NOD islets associated with immune infiltration.

BB-Cl-amidine treatment is effective in decreasing overall levels of citrullination in the pancreas, as shown by Western blotting and quantitative LC-MS/MS. Of interest, the most significant reduction was measured for citrullinated GRP78. In line with this, the level of autoantibodies against citrullinated GRP78 is decreased in the circulation of BB-Cl-amidine–treated NOD mice. Although autoantibodies are not believed to be pathogenic in type 1 diabetes, this finding underlines the important effect of BB-Cl-amidine on reducing the antigenicity of β-cell proteins in the inflamed islet environment.

Pan-PAD inhibitors are known to alter NET release (34–37). Neutrophils and NETosis are believed to play an important role in the initiation and perpetuation of type 1 diabetes (20,21,43). Here, we show that BB-Cl-amidine reduces spontaneous NET release ex vivo from bone marrow–derived neutrophils of NOD mice, a finding that may be linked to a decrease in citrullinated histone H3 levels. Indeed, citrullination of histone H3, mediated by activation of PADs in neutrophils, is an initiating factor for the formation of NETs (44) and has been shown to be reduced by PAD inhibition (36). Our results are in line with studies in lupus (30,34) and collagen-induced arthritis (37), demonstrating that reduced NETosis in response to PAD inhibition is associated with disease amelioration.

Looking at the islet level, BB-Cl-amidine treatment resulted in only a marginal reduction of insulitis, whereas pancreatic insulin levels were preserved. However, we did observe a significant reduction in the percentage of CD4⁺ T_EM and in the frequency of IFN-γ–producing CD4⁺ and CD8⁺ T cells in the pancreata of BB-Cl-amidine–treated mice. The observed reduction in IFN-γ production may reflect an impaired effector function of the infiltrating immune cells (45). These results suggest that the immune cell infiltrate in BB-Cl-amidine–treated mice is less aggressive (46).

Our results on increased frequency of T_reg in the blood and spleen demonstrate a peripheral effect of BB-Cl-amidine. Additionally, splenic IFN-γ production of BB-Cl-amidine–treated mice was significantly reduced. These results suggest a more regulatory environment that favors immune tolerance and prevents intensification of the immune assault. Also, with Th1/Th2 polarization believed to play a role in type 1 diabetes pathogenesis, the significant increase in IL-4 and decrease in the IFN-γ/IL-10 ratio in the circulation of BB-Cl-amidine–treated mice supports this hypothesis. Of note, whether Th1/Th2 polarization is a cause rather than a consequence of disease development is not fully established. In line with these findings, an induction of Th2 cells by BB-Cl-amidine has been shown in a mouse model of rheumatoid arthritis (31), and a direct effect of PAD inhibition on enhancing the differentiation of Th2 subsets through inhibition of GATA3 citrullination was demonstrated mechanistically by Sun et al. (47).

For the first time, a pan-PAD inhibitor was used in a very long treatment regimen (17 weeks), with no detrimental side effects observed. This, together with reported disease amelioration with the use of PAD inhibitors in other autoimmune models, opens the road to potential clinical therapeutic applications. However, since PADs also participate in physiological processes (11), such as gene regulation and nerve myelination, such therapeutic treatment regimens will need careful consideration. The recent development of specific PAD2 (48) and PAD4 (49) inhibitors, as well as reversible PAD inhibitors (50), may be a solution, but their efficacy to treat autoimmune diseases remains to be confirmed. Results from ongoing studies may further invite to consider PAD inhibition as a therapeutic strategy not only for type 1 diabetes but also for other autoimmune diseases.

In conclusion, treatment of NOD mice with BB-Cl-amidine resulted in complete diabetes prevention associated with decreased protein citrullination, anticitrulline autoantibody formation, NETosis, and in situ IFN-γ responses. We postulate that these effects, combined with the direct effect of PAD inhibition on T cells, might lead to a less aggressive inflammatory environment in the islets and in the periphery. Our findings align with previous reports that citrullination plays a pathogenic role in type 1 diabetes and demonstrate that PAD inhibition is an attractive target for clinical therapeutic applications in autoimmunity.

Acknowledgments. The technical assistance of Eline Desager, Jos Laureys, Marijke Viaene, and Martine Gilis (Laboratory of Clinical and Experimental Endocrinology, KU Leuven) is greatly appreciated. The authors thank the KU Leuven Flow Cytometry Core Facility for assistance with flow cytometry. Fluorescent images were recorded on a Zeiss LSM 780–SP Mai Tai HP DeepSee (Cell and Tissue Imaging Cluster supported by Hercules AKUL/11/37 and FWO G.0929.15 to Pieter Vanden Berghe, University of Leuven) or on a Zeiss LSM 880–Airyscan (Cell and Tissue Imaging Cluster supported by Hercules AKUL/15/37_GOH1816N and FWO G.0929.15 to Pieter Vanden Berghe, University of Leuven). The authors thank the University of Massachusetts Medical School proteomics core facility director Dr. Scott Shaffer for helping with proteomics analysis and Dr. Roshanak Aslebagh for running proteomics samples.

Funding. This work was supported by IMI2-JU under grant agreements 115797 (INNODIA) and 945268 (INNODIA HARVEST). This joint undertaking receives support from the European Union’s Horizon 2020 research and innovation program and European Federation of Pharmaceutical Industries and Associations, JDRF, and The Leona M. and Harry B. Helmsley Charitable Trust. Further support came from the KU Leuven (C16/18/006), Fonds voor Wetenschappelijk Onderzoek Vlaanderen (a predoctoral fellowship for S.B. [11A0220N], D.P.C. [11Y6716N], and A.C. [1189518N] and a postdoctoral fellowship for M.B. [12R0719N]), and the National Institutes of Health (R35-GM-118112 to P.R.T.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.M.C.S. contributed to the study design and conduct and analysis and interpretation of the data and wrote and edited the manuscript. S.B. helped with the design, conduct, and analysis of the NETosis data. Y.B. contributed by injecting mice and performing experiments. R.T. and P.R.T. conducted and analyzed the proteomic analysis of the biotin-PG–labeled samples and the LC-MS/MS. D.P.C. helped with mice treatment and contributed to the design and analysis of the FACS data. C.B. and R.S. performed and analyzed the islet isolation, sorting, and quantitative PCR. S.M. and P.R.T. provided the BB-Cl-amidine. A.C. helped with conducting experiments. S.Y. and R.M. contributed to the design and interpretation of islet experiments. C.M., M.B., and L.O. designed the research, interpreted data, and wrote and edited the manuscript. M.B. also conducted and analyzed experiments. All authors edited the manuscript and gave their final approval of the version to be published. C.M. and L.O. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.