The effect of interleukin-22 treatment on autoimmune diabetes in the NOD mouse

In an attempt to prevent or reverse type 1 diabetes, basic and clinical research has focused intensely on immunotherapies, beta cell replacement and preservation of beta cell function and mass [1]. The complexity of type 1 diabetes pathogenesis [2] and the human beta cell [3] has shifted the focus to better understanding disease heterogeneity. Most recently, post-translational modifications of beta cell proteins have been identified as ‘neo-antigens’ in the NOD mouse [4] and individuals with type 1 diabetes (reviewed in [5]), further contributing to the complexity of this disorder.

The mechanism of why and how neo-antigens are produced remain unknown. It has been hypothesised that the normal physiological role of beta cells as highly secretory cells, when combined with various environmental stimuli, significantly increases oxidative and endoplasmic reticulum (ER) stress. This leads to MHC presentation of secretory pathway proteins and neo-antigens, precipitating disease in individuals with a genetic predisposition to autoimmunity [5]. Thus, therapies with the ability to reduce beta cell oxidative stress, protein misfolding and ER stress may limit neo-antigen production, reducing the likelihood of activation of potentially pathogenic neo-antigen-specific T cells that are subject to normal tolerance mechanisms [5]. Furthermore, once autoimmunity develops, beta cell damage is largely driven by T cell cytokines that drive oxidative stress and consequent beta cell apoptosis, providing another potential benefit of therapies inhibiting oxidative stress.

We previously demonstrated that the cytokine IL-22 directly reduces pancreatic beta cell oxidative and ER stress caused by cytokines and glucolipotoxicity in murine and human islets that highly express the IL-22 receptor [6]. When IL-22 therapy was administered in the short term to obese hyperglycaemic mice, we and others observed a reduction in hyperglycaemia [6, 7], which was associated with suppression of islet ER stress and inflammation, promotion of high-quality insulin and restoration of glucose homeostasis [6]. To address whether IL-22 therapy could reduce beta cell stress and prevent the development of autoimmune diabetes, or treat established disease, we used the NOD mouse experimental model of type 1 diabetes. We investigated whether, using this dosing regimen, prophylactic IL-22 therapy would prevent diabetes, and whether therapeutic IL-22 administration would improve beta cell function or gut integrity in the NOD mouse.

Prophylactic therapy with rmIL-22 for 32 weeks failed to delay or prevent the development of overt diabetes in NOD mice (rmIL-22, 8 out of 10; PBS, 12 out of 15; p = 0.82, logrank test; Fig. 1a). Median diabetes-free survival in mice treated with rmIL-22 was 175 days (95% CI 115, 265 days) compared with 162 days (95% CI 143, 265 days) in the vehicle control-treated mice (p = 0.90, Mann–Whitney U test; Fig. 1a). rmIL-22 therapy for 32 weeks did not result in an alteration in body weight (p _slope = 0.095, linear regression; Fig. 1b) or non-fasted blood glucose (p _slope = 0.057, linear regression; Fig. 1c), compared with vehicle-treated mice.

After overt diabetes had been confirmed in animals treated in the prophylactic control arm, mice were given suboptimal insulin therapy (LinBit insulin pellets) and reallocated to receive therapeutic doses of either rmIL-22 or vehicle to determine whether therapy could prolong the ‘honeymoon phase’ of diabetes progression. At the time of randomisation, the mice did not significantly differ in age (PBS, 173 days, 95% CI 110, 235 days; rmIL-22, 166 days, 95% CI 98, 235 days). In the therapeutic intervention arm, rmIL-22 failed to significantly prolong median time to diabetes progression (39 days, 95% CI 6, 54 days; n = 5; Fig. 1d) in comparison with the vehicle-treated animals (35 days, 95% CI 6, 47 days; n = 6; p = 0.46, logrank test; Fig. 1d). Throughout this therapeutic phase of the study, rmIL-22 intervention did not significantly affect body weight (p _slope = 0.65, linear regression; Fig. 1e) or non-fasted blood glucose (p _slope = 0.43, linear regression; Fig. 1f) compared with the animals that were randomised to PBS treatment.

To determine whether therapeutic administration of rmIL-22 induced changes in the pancreas, pancreases were collected at the time when diabetes progression was confirmed in the therapeutic experiment. At the time of diabetes progression and hyperglycaemia, islet area and total islet area including infiltrate did not increase upon rmIL-22 treatment (p = 0.07 and p = 0.35, respectively, Mann–Whitney U test; Fig. 2a, b). While varying degrees of immune cell infiltrate were observed in islets from both treatment groups (p < 0.0001, χ ²-test; Fig. 2c), rmIL-22 therapy accounted for less than 0.1% of the total variance in insulitis grading (Bonferroni’s multiple comparisons test; Fig. 2c).

No significant differences in the insulitis index were observed (p = 0.69; Fig. 2d). Residual beta cell function, as determined by circulating C-peptide, reflected a similar variation, but no significant improvement, after therapeutic administration of rmIL-22 in comparison with vehicle-treated mice (p = 0.27; Fig. 2e). Despite the variation observed with fasting C-peptide levels, no correlation was observed with islet area or insulitis index (data not shown). However, C-peptide was positively correlated with higher proportions of non-infiltrated islets (p = 0.03, Spearman’s rank order correlation; Fig. 2f) and negatively correlated with lower proportions of peri-insulitis and 50% or less insulitis (p = 0.003 and 0.033, respectively, Spearman’s rank order correlation; Fig. 2f). Therapeutic administration of rmIL-22 did not affect intestinal mucosal barrier function as determined by circulating LPS, a surrogate marker for impaired mucosal integrity (p = 0.15; Fig. 2g). LPS levels did not correlate with circulating C-peptide or islet pathology (data not shown). No differences were observed in macrophage or neutrophil infiltration, and beta cell staining with the ER stress marker GRP78 was not altered by rmIL-22 treatment (Fig. 2h, i).

Here we used rmIL-22 therapy to determine whether diabetes onset could be delayed or whether hyperglycaemia could be controlled in the NOD mouse. In this small study, we found that neither prophylactic treatment starting at 6 weeks of age, nor therapeutic intervention of rmIL-22 once overt diabetes was confirmed, delayed the onset of diabetes or improved glycaemic control in comparison with vehicle-treated mice.

In contrast to previous findings from our group [6] and others [7] in mouse models of obesity and type 2 diabetes, we observed little change in beta cell function after IL-22 therapy in the NOD mouse. The amount of IL-22 required to overcome beta cell ER stress to limit self-antigen presentation and beta cell loss in autoimmune diabetes may differ substantially from that required to contend with the low-level chronic inflammation in obesity and type 2 diabetes. Greater IL-22 exposure could be achieved by more frequent administration or the use of more metabolically stable hybrid proteins, such as Fc-based fusion proteins, which have been shown to extend the half-life of IL-22 by approximately 3 days [7]. Previous studies demonstrated that deletion of IL-22 suppressed diabetes progression led by autoreactive T cells [10], and that rmIL-22 treatment did not block streptozotocin-induced type 1 diabetes [11]. We confirmed that IL-22 was active and reached the pancreas by assessing downstream STAT3 activation in whole pancreas of NOD mice, 30 min after the administration of 200 ng/g rmIL-22 (see electronic supplementary material [ESM] Fig. 1a). It is also possible that, by 6 weeks of age, when treatment was commenced, NOD mice had been exposed to environmental cues and beta cell ER stress that had already initiated presentation of self-antigens and precipitated T cell-driven autoimmunity.

Of note, IL-22RA1 expression in the pancreas of NOD mice also diminished with ageing (ESM Fig. 1b), which could be due to increased infiltration within the islets. Using a tolerogenic vaccination approach, in utero delivery of a proinsulin Fc fusion protein in G9C8 developing embryos has recently been demonstrated to be beneficial in preventing disease onset [12]. This early delivery method may overcome the issue of treatment timing in this autoimmune model [12] and the barrier of IL-22 placental transport [13]. However, the viability for clinical translation of such early intervention approaches is doubtful.

Circulating LPS remained unchanged after therapeutic delivery of IL-22, in contrast to our previous findings in obese mice, in which the improvement of mucosal integrity by IL-22 therapy led to changes in gut microbiota, improved intestinal barrier function and lowered serum LPS [14]. Leaky intestinal mucosal barrier [15], altered microbiota and reductions in short-chain fatty acids are a common occurrence in both types of diabetes [16, 17]. This lack of change may illustrate disparities in endotoxin/LPS levels in different preclinical rodent models of diabetes, or differences in the detection method used for mucosal integrity. We believe future work should focus on understanding whether different IL-22 dosing regimens or IL-22 fusion proteins would (1) resolve autoimmunity-induced ER stress in the beta cell, and (2) affect microbiota or pancreatic antimicrobial peptides [18], to improve efficacy in autoimmune diabetes.