Autoreactive T Cells and Cytokine Stress Drive β-Cell Senescence Entry and Accumulation in Type 1 Diabetes

Abstract

Type 1 diabetes (T1D) results from a complex dialogue between the immune system and islets characterized by T cell–mediated autoimmune destruction of pancreatic β-cells. In this dialogue, β-cell stress responses have emerged as drug targets for slowing T1D progression, including a subpopulation of senescent β-cells that accumulate during T1D in humans and nonobese diabetic (NOD) mice. However, the mechanisms that cause β-cells to activate senescence in T1D are not known. Here, we show that β-cell senescence entry and accumulation are driven by damage inflicted by autoreactive CD4⁺ and CD8⁺ T cells in the late presymptomatic stages of T1D. Genetically immune-deficient NOD strains showed reduced frequencies of senescent β-cells, and adoptive transfer of diabetogenic splenocytes was sufficient to activate β-cell senescence in immune-deficient mice. Modulation of antigen-specific CD4⁺ T cells using an intermittent paradigm of CD3 antibody in immune-competent wild-type NOD mice led to reduced senescence, but did not affect other responses, concomitantly with slowing disease progression. Depletion of CD4⁺ or CD8⁺ T cells phenocopied the effect of CD3 antibody on β-cell senescence. CD3 antibody and senolytic ABT-199 had a complementary effect in reducing senescent β-cell burden, consistent with these agents acting in different pathways. Mechanistically, exposure to T1D-related inflammatory cytokines recapitulated stable phenotypes of senescence in human islets and β-cells. Our results demonstrate that β-cell senescence is a stress response that depends on progressive autoreactive CD4⁺ and CD8⁺ T-cell damage in T1D and suggests a novel mechanism of action for CD3 immunotherapy in limiting the accumulation of senescent β-cells.

Article Highlights

• Senescence is a β-cell stress response in type 1 diabetes (T1D), the origins of which are not understood. We wanted to determine the role of the T cell–mediated autoimmune process in β-cell senescence during T1D.

• In the nonobese diabetic mouse model, β-cell senescence largely depended on damage inflicted by autoreactive CD4⁺ and CD8⁺ T cells during the development of T1D. Chronic exposure to sublethal doses of proinflammatory cytokines associated with the diabetogenic process was sufficient to elicit stable senescence phenotypes in human islets in culture.

• Our findings suggest that autoreactive T cells trigger not only β-cell death but also β-cell senescence, potentially via cytokine-dependent mechanisms in T1D. This finding has implications for understanding the mechanisms of action and beneficial impacts of immunotherapy using CD3 antibodies in T1D.

Introduction

Type 1 diabetes (T1D) results from a chronic T cell–mediated autoimmune process that leads to selective and progressive destruction of insulin-producing pancreatic β-cells (1). In addition to the role of T cells, a growing body of work has shown that β-cell stress responses promote progression of the disease (2,3). These include endoplasmic reticulum stress leading to activation of the unfolded protein response (UPR) (4) and the type I interferon (IFN) response leading to inflammatory gene expression and upregulation of MHC class I (5). Modulating the UPR in β-cells slows disease progression in the nonobese diabetic (NOD) autoimmune T1D mouse model (6–10) and mitigates the decline in residual β-cell function in individuals with newly diagnosed T1D (11,12). Similarly, small-molecule targeting of downstream effectors in the type I IFN response alleviates diabetes in NOD mice (13,14) and allays residual β-cell functional decline in humans with symptomatic disease (15). Together, these studies indicate that β-cell stress pathways play a causal role in T1D.

In addition to the UPR and IFN response, a small subpopulation of β-cells (∼5–10%, on average) undergo a senescence stress response during T1D in NOD mice and humans (16). β-Cell senescence is characterized by DNA damage, expression of cyclin-dependent kinase inhibitors p21 and p16^INK4A, a prosurvival phenotype marked by elevated expression of Bcl-2 family proteins, senescence-associated β-galactosidase activity, and a senescence-associated secretory phenotype (SASP) involving a variety of cytokines, chemokines, growth factors, and proteases (17). Studies in NOD mice suggest that β-cell senescence accelerates disease progression and diabetes onset. Treatment of NOD mice with senolytic compound ABT-199 or ABT-737 to trigger apoptosis in the senescent β-cell population (16), bromodomain extraterminal domain inhibitors to suppress SASP (18,19), or palmitic acid hydroxy stearic acids mitigates β-cell senescence and slows disease progression in NOD mice (20). However, it remains to be determined what causes β-cells to activate senescence in T1D.

β-Cells expressing 53BP1, a marker of DNA repair, are more frequent in donors with new-onset T1D compared with normoglycemic control donors and correlate with islet lymphocyte infiltration (21), suggesting a potential connection between β-cell DNA damage and the autoimmune process. In NOD mice, senescent β-cell accumulation progresses during the late prediabetic stage of disease from 8 to 16 weeks of age (16), coinciding with the final phases of CD4⁺ and CD8⁺ T-cell effector destruction. However, the relationships between the T cell–mediated autoimmune process and β-cell senescence during T1D have not been investigated. Here, we demonstrate that the autoimmune process led by autoreactive T cells is both necessary and sufficient to drive activation of the senescence program and accumulation of senescent β-cells during the development of T1D in NOD mice. Sublethal doses of T1D-associated inflammatory cytokines IFN-γ, tumor necrosis factor-α (TNF-α), and interleukin-1β (IL-1β) led to stable activation of senescence phenotypes in human islets and EndoC-βH5 cells, suggesting that cytokines may contribute to β-cell senescence. Together, these results provide insights into the mechanisms of β-cell senescence in T1D and reveal a novel mechanism of action of CD3 immunotherapy in the mitigation of this maladaptive β-cell stress response.

Research Design and Methods

Mouse Husbandry

All mouse procedures were approved for ethics by the Animal Care Committee at the University of Manitoba. Female NOD (strain 1976; NOD/ShiLtJ) mice and NOD.Scid-γ (strain 5557; NSG) mice were from The Jackson Laboratory. All mice were housed at the University of Manitoba Children’s Hospital Research Institute of Manitoba animal care facility or central animal care facility and fed a chow diet on a typical 12-h light–dark cycle. Body weight was measured weekly from age 9 to 13 weeks, and random blood glucose was measured once per week from week 11 to 13 from the tail vein, using a Contour Next glucose monitor and test strips. Mice were diabetic (hyperglycemic) if a single fasted blood glucose measurement was ≥7 mmol/L or if two consecutive random measurements were ≥14 mmol/L.

Treatment of NOD Mice With Experimental Agents

All treatments were administered to confirmed normoglycemic female NOD mice. Female NOD mice aged 9 weeks were treated with either 2.5 mg/kg anti-CD3ε F(ab)₂ (Bio X Cell) or 2.5 mg/kg species-matched immunoglobulin G (IgG) F(ab)₂ (Bio X Cell) via i.p. injection once per week for 4 weeks. For combination treatment, NOD mice aged 9 weeks received one of the following for 4 weeks: vehicle only (60% Phosal 50 PG, 30% PEG400, and 10% ethanol via oral gavage), ABT-199 plus IgG (25 mg/kg ABT-199 oral gavage twice per week and 2.5 mg/kg Armenian hamster IgG F[ab]₂ i.p. once per week), or ABT-199 plus anti-CD3 (25 mg/kg ABT-199 twice per week and 2.5 mg/kg anti-CD3ε F[ab]₂ i.p. once per week), following the dosing of ABT-199 from previous studies (16,22).

Adoptive Transfers

Adoptive transfers were performed according to previous methods (23), with some modifications. Splenocytes from diabetic female NOD mice were resuspended in PBS for injection (2 × 10⁷ cells in 200 μL per mouse or PBS alone as a control) and injected i.p. into 8-week-old female NSG mice (n = 6 per group).

Intraperitoneal Glucose Tolerance Testing

Intraperitoneal glucose tolerance testing was performed according to previous methods (24), with some modifications. Filter-sterilized d-glucose was administered to 13.5-week-old female NOD mice after overnight (18 h) fasting via i.p. injection at 2 g/kg after measuring fasted blood glucose levels for time 0. Blood glucose concentration was measured at the tail vein at 15, 30, 60, and 120 min postinjection.

Mouse Islet Isolation

Islets from female NOD mice were isolated according to previously established digestion methods via perfusion of the pancreas with 1.4 mg/mL collagenase P in Hanks’ balanced salt solution containing 0.5% BSA, Ca²⁺, and Mg²⁺ (Millipore-Sigma) through the cannulated bile duct (25,26). Islets were cultured in RPMI-1640 containing 2 mmol/L l-glutamine, 10% FBS, and 1X antibiotic–antimycotic (Gibco) and rested for 1–3 h or 24 h at 37°C with 5% CO₂ before being handpicked for experiments.

Human Donor Islet Studies

Human islets were procured from the Alberta Diabetes Institute Islet Core and by the National Institute of Diabetes and Digestive and Kidney Diseases–funded Integrated Islet Distribution Program at City of Hope. All experiments were completed with donor preparations of ≥75% purity (determined by dithizone staining), and islets were always handpicked to ensure purity before experiments. All human islet donor information is provided in Supplementary Table 1. On arrival, islets were rested for 24 h in RPMI-1640 media containing 10% FBS, 2 mmol/L l-glutamine, 5.5 mmol/L glucose, and 1X antibiotic–antimycotic at 37°C and 5% CO₂. For senescence induction, we used our previous protocol specifying treatment with the DNA-damaging agent bleomycin sulfate at 50 μmol/L for 48 h followed by 4-day washout in regular islet media lacking drug (27). Alternatively, human islets were left untreated as a control or treated with inflammatory cytokines (StemCell Technologies) consisting of 2 ng/mL IL-1β, 10 ng/mL TNF-α, and 10 ng/mL IFN-γ (cytokine mix dissolved in water) and replaced daily with fresh cytokine-containing media for 48 h and then cultured in regular islet media without cytokines for 4 days before collection (drug washout).

EndoC-βH5 Cell Culture

The EndoC-βH5 human female fetal β-cell line (28) was purchased from Human Cell Design and cultured, and senescence was induced with bleomycin sulfate according to previous methods (27). Cells were cultured in media containing 35 μmol/L bleomycin for 48 h, which was then replaced with regular growth media for 4 days. Alternatively, cells were treated with media containing 2 ng/mL IL-1β, 10 ng/mL TNF-α, and 10 ng/mL IFN-γ (cytokine mix) for 96 h before a 24-h cytokine washout (all cytokines were from StemCell Technologies) or left untreated in regular media as a control.

Viability Assays

Viability assays in human islets or EndoC-βH5 cells were performed as previously described (29) using the PrestoBlue HS Viability Assay (Invitrogen) on a Cytation 5 instrument. After viability assays, islets were washed twice with 1 mL PBS before being harvested for RNA isolation.

RNA Isolation and Quantitative RT-PCR

Islets or cells were stored in 150–300 μL Trizol reagent (Zymogen) at −80°C. Total RNA was extracted using the Zymogen RNA MicroPrep Kit as previously described (27). Total RNA from EndoC-βH5 cells was isolated from frozen pellets using the RNeasy Micro Kit (Qiagen) as described previously (29). Template cDNA was synthesized using 100–300 ng RNA with the Luna RT SuperMix Kit (BioLabs). Data were analyzed with the delta-delta C_T method normalized to the housekeeping gene Ppia or PPIA for mouse or human islets and EndoC-βH5 cells, respectively. Primer sequences for this study are listed in Supplementary Table 4.

Immunohistochemistry

Immunohistochemistry (IHC) was performed as previously described (22). Fixed pancreas blocks from 12-week-old female NOD.Rag1-knockout (NOD.Rag1^−/−) mice (strain 3729) were purchased from The Jackson Laboratory, and fixed pancreas blocks from 14-week-old female NOR/LtJ mice (strain 2050; The Jackson Laboratory) were a gift from A. Bhushan (University of California San Francisco Diabetes Center). Slides were stained with primary and fluorophore-conjugated secondary antibodies (Supplementary Table 3) and counterstained with ProLong Diamond antifade media with DAPI (Thermo Fisher Scientific, Waltham, MA). Markers were scored on >10 randomly selected islets per mouse using several slides from two different regions of the pancreas spaced by 100 μm. Insulitis was scored in insulin-positive islets as intact, peri-insulitis, and insulitis based on islet-infiltrating lymphocytes as determined by DAPI staining. Sample sizes are reported in figure legends, and additional details are provided in the Supplementary Material.

Flow Cytometry

Flow cytometric analysis of splenocytes was performed according to standard methods. Splenocyte suspensions from each mouse were stained with antibodies or tetramers to the following targets: mouse CD45, mouse CD3, mouse CD4, mouse CD8b.2, mouse insulin_13–21 MHC II tetramer (30), or mouse IGRP_206–214 MHC I tetramer (31) (Supplementary Table 3). The Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) was used with anti-CD4 and Foxp3 to stain regulatory T cells. Zombie Near-IR Viability Dye (BioLegend) was used in Cell Staining Buffer (BioLegend) for all flow cytometric experiments, and >10,000 live cells were scored in each sample. Sample gating for each staining panel is shown in Supplementary Figs. 1 and 8.

Western Blot Analysis

Human islets were harvested and lysed in ice-cold radioimmunoprecipitation lysis and extraction buffer (Thermo Fisher Scientific) containing 1X Halt complete protease and phosphatase inhibitor cocktail, EDTA free (Thermo Fisher Scientific), and protein concentration was determined by BCA assay (Thermo Fisher Scientific). Antibody information is reported in Supplementary Table 3. Signals were detected with horseradish peroxidase–conjugated secondaries (1:100,000) and enhanced chemiluminescence reagents (SuperSignal West Pico Plus; Thermo Fisher Scientific) after exposure to X-ray film, with multiple exposures to confirm signal was in linear range.

mRNA Sequencing Analysis

RNA was isolated as described above from five to 10 handpicked and frozen islets isolated from 13.5-week-old anti-CD3– or IgG-treated NOD mice at 2 or 3 days after the last dose, respectively (n = 6 mice per group). Total RNA was submitted to Azenta Life Sciences (South Plainfield, NJ) for next-generation sequencing services including poly(A) RNA enrichment, library preparation, and paired-end 150-bp mRNA sequencing on the Illumina Nova X Plus platform to obtain ∼20 million reads per sample. Additional details of bioinformatic analysis can be found in the Supplementary Material. Differential expression was calculated using DESeq2, and normalized read counts and differentially expressed genes (≥1.5 fold change; adjusted P < 0.05) are listed in Supplementary Table 2.

Statistical Analysis

All group sample sizes (n) and definitions of samples measured are shown in figure legends. Data are presented as mean ± SD of at least three to six biologic replicates. In each experiment, we considered biologic replicates as either individual mice or samples/wells of cells or islets treated in parallel. Statistical calculations were performed in GraphPad Prism (version 10; GraphPad Software, La Jolla, CA). Data were considered significant at P < 0.05 during two-tailed t test or one-way ANOVA. Additional details on quantifications and statistics are provided in the Supplementary Material.

Data and Resource Availability

The data sets generated are available in the National Center for Biotechnology Information Gene Expression Omnibus repository under accession GSE275350. No applicable resources were generated or analyzed during the current study.

Results

Genetic Disruption of Autoimmunity Leads to Reduced Accumulation of Senescent β-Cells in NOD Mice

To investigate whether the autoimmune process is associated with β-cell senescence in T1D, we performed IHC for established senescence markers in well-characterized genetically immune-deficient NOD mouse models affecting adaptive immunity. These included NSG and NOD.Rag1^−/− strains, neither of which develop diabetes because of the profound lack of mature T and β-cells (32,33). We compared these with wild-type control NOD females at 12–14 weeks of age, corresponding to the late prediabetic stage where senescent β-cell accumulation is most pronounced (16). In both immune-deficient strains, β-cells expressing p21 as an early senescence marker and Mmp2 as an SASP and late senescence marker (17) (Fig. 1A) were substantially reduced relative to wild-type NOD females, although not abolished (Fig. 1B and C). We also performed IHC on the T1D-resistant strain NOR/LtJ, which harbors a protective C57BL6 polymorphism but is matched at the NOD MHC class II alleles and accordingly does not develop destructive insulitis (34). Interestingly, NOR mice showed a similar phenotype to the NSG and NOD.Rag1^−/− mice of reduced frequency of β-cells expressing p21 or Mmp2 compared with controls (Fig. 1A–C). These data suggest that an active autoimmune destruction process is involved in the accumulation of senescent β-cells in NOD mice.

Figure 1.

Senescent β-cell accumulation is reduced in strains of NOD mice genetically deficient in adaptive immunity or harboring disease-resistant polymorphisms. A: IHC for p21 (red) or Mmp2 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD wild-type (WT), NOD.Prkdc^Scid/scid;Il2rg^−/− (NSG), NOD.Rag1^−/−, and NOR mice (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). B: Quantification shows mean percentage of p21⁺ β-cells per islet scored from NOD WT (n = 32 from 3 mice), NSG (n = 40 from 4 mice), NOD.Rag1^−/− (n = 40 from 4 mice), and NOR (n = 40 from 4 mice). C: Quantification shows mean percentage of Mmp2⁺ β-cells per islet scored from NOD WT (n = 38 from 3 mice), NSG (n = 40 from 4 mice), NOD.Rag1^−/− (n = 40 from 4 mice), and NOR (n = 40 from 4 mice). Error bars indicate SD. **P < 0.005, ***P < 0.0005, ****P < 0.0001 by one-way ANOVA with Dunnett multiple comparison test. KO, knockout.

Diabetogenic Splenocytes Trigger β-Cell Senescence Accumulation

To directly ascertain whether the autoimmune destructive process drives β-cell senescence, we adoptively transferred splenocytes from newly diabetic NOD female mice into 8-week-old female NSG mice (Fig. 2A). We monitored glycemia in the recipient mice for 3–4 weeks and euthanized the cohort when the first mouse became diabetic (Fig. 2B) at 11.5 weeks of age to perform IHC of senescence markers, insulitis scoring in the pancreas, and flow cytometry of T cells. Flow cytometry confirmed repopulation of the spleen with CD3⁺ T cells (Fig. 2C), and insulitis scoring confirmed recruitment of the transferred immune cells into the pancreas (Fig. 2D and E). Remarkably, immunostaining showed that there was a significant increase in the percentage of β-cells expressing p21 or Mmp2 in NSG mice receiving adoptive transfer compared with buffer-only controls (Fig. 2F and G). Together, these findings demonstrate that the autoimmune process is both necessary and sufficient to drive senescence activation and accumulation of senescent β-cells during the development of T1D.

Figure 2.

Adoptive transfer of diabetogenic splenocytes triggers β-cell senescence in immune-deficient NSG mice. A: Schematic representation of adoptive transfer. Eight-week-old female NSG recipient mice were injected i.p. with 2 × 10⁷ total splenocytes from diabetic female wild-type NOD/ShiLtJ mice (n = 6) or injected with PBS buffer as control (n = 6), and glycemia was monitored for 3–4 weeks before euthanizing all mice. B: Glycemia in NSG control and NSG splenocyte-injected mice (n = 6 mice per group) at 3.5 weeks postadoptive transfer. Mice with at least one reading ≥14 mmol/L were considered potentially diabetic. C: Flow cytometric analysis of mean percentage of splenic total CD3⁺ T cells in NSG PBS vs. NSG plus splenocyte mice (n = 6 mice per group). D: Representative IHC images for insulitis scoring stained with insulin (green) and DAPI (blue): left, no infiltration; middle, peri-insulitis; and right, insulitis (scale bar, 40 μm). E: Islet insulitis scoring in NSG control (n = 60 from 6 mice) and NSG plus splenocyte mice (n = 53 from 4 mice). Data are mean percentage of islets showing no infiltration (intact), peri-insulitis, or insulitis. F: IHC for p21 (red), insulin (green), and DAPI (blue) in pancreas sections from NSG control and NSG plus splenocyte mice (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of p21⁺ β-cells per islet scored from NSG (n = 60 from 6 mice) and NSG plus splenocyte (n = 53 from 4 mice). G: IHC for Mmp2 (red), insulin (green), and DAPI (blue) in pancreas sections from NSG control and NSG plus splenocyte mice (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of Mmp2⁺ β-cells per islet scored from NSG (n = 60 from 6 mice) and NSG plus splenocyte (n = 63 from 4 mice). Error bars indicate SD. **P < 0.005, ***P < 0.0005, ****P < 0.0001. ns, not significant by two-tailed unpaired t test (B, C, F, and G) or χ² test (E). Created in BioRender.

Intermittent CD3 Antibody Modulates Autoreactive CD4⁺ T Cells and Slows Disease Progression in NOD Mice

We next addressed the role of autoreactive CD4⁺ and CD8⁺ T cells in senescent β-cell accumulation during disease progression in wild-type NOD mice. Modulation of cytotoxic T cells using CD3 antibody delays symptomatic T1D (35) and slows decline of β-cell function in new-onset T1D in humans (36) and both prevents and reverses symptomatic disease in NOD mice (37,38). Therefore, we devised a treatment paradigm with an established nonmitogenic monoclonal CD3 antibody, anti-CD3ε (F[ab]₂ fragments of clone 145-2C11) (39). Female NOD mice were dosed weekly with 2.5 mg/kg over the period of 9–13 weeks during the late prediabetic phase when senescent β-cells accumulate and compared with species-matched F(ab)₂ fragments of IgG as a control (Fig. 3A). This CD3 antibody reverses overt diabetes in NOD mice when given for 5 consecutive days in a manner dependent on regulatory T cells (37) and prevents diabetes when a single dose is administered in 1- to 3-week-old mice (38). However, no studies have tested an intermittent regimen and characterized its effect at the late prediabetic stage in NOD mice (40).

Figure 3.

Modulation of antigen-specific CD3⁺ T cells improves glucose tolerance and slows disease progression in late prediabetic NOD mice. A: Schematic representation of intermittent anti-CD3 treatment paradigm. NOD mice were administered monoclonal anti-CD3 145-2C11 F(ab)₂ fragments or species-matched IgG F(ab)₂ fragments once per week from 9 to 13 weeks, followed by metabolic and immune phenotyping and histologic assays. B: Flow cytometric analysis of mean percentage of total CD3⁺ T cells from anti-CD3– and IgG-treated mice (n = 12 mice per group). C: Flow cytometric analysis of mean percentage of insulin tetramer–positive/CD3⁺ T cells from anti-CD3– and IgG-treated mice (n = 18 mice per group). D: Flow cytometric analysis of mean percentage of IGRP tetramer–positive/CD3⁺ T cells from anti-CD3– (n = 18 mice) and IgG-treated mice (n = 17 mice). E: Flow cytometric analysis of mean percentage of total CD4⁺ T cells from anti-CD3– and IgG-treated mice (n = 12 mice per group). F: Flow cytometric analysis of mean percentage of Foxp3⁺/CD4⁺ T cells from anti-CD3– and IgG-treated mice (n = 12 mice per group). G: Intraperitoneal glucose tolerance tests (IPGTT) were carried out in NOD mice after 4-week treatment with anti-CD3 or IgG control (n = 18 mice per group). Data are mean glucose concentration for each time point. Time points not significant (ns) unless noted. H: Area under the curve (AUC). I: Quantification of β-cell area in anti-CD3 and IgG NOD mice. Data are mean insulin-positive islet area in squared microns; each point represents single mouse from IgG (n = 6 mice; 88 islets scored) and CD3 (n = 6 mice; 96 islets scored). J: Islet insulitis scoring in anti-CD3– and IgG-treated NOD mice. Data are mean percentage of islets showing no infiltration (intact), peri-insulitis, or insulitis from IgG (n = 88 from 6 mice) and CD3 (n = 96 from 6 mice). Error bars indicate SD. *P < 0.05, ****P < 0.0001. ns, not significant by two-tailed unpaired t test (B–I). ***P < 0.0005 by χ² test (J). Created in BioRender.

A single i.p. treatment of NOD mice with CD3 antibody led to rapid and transient depletion of total CD3⁺ T cells in the spleen and peripheral blood mononuclear cells starting from 2 h and peaking at 24 h compared with a species-matched control IgG treatment (Supplementary Fig. 1A–C and Supplementary Fig. 2A–C). Two days after the last dose, total CD3⁺ T cells remained partially depleted in the spleen (Fig. 3B). Tetramer staining on splenocytes to quantify MHC II–presented insulin-reactive T cells (30) and MHC I–presented IGRP-reactive T cells (31) showed that anti-CD3 led to a reduction in the former (Fig. 3C) but not the latter (Fig. 3D). Consistent with this, total CD4⁺ T cells were also reduced in the spleens of the anti-CD3 treated mice (Fig. 3E), whereas CD8⁺ T cells were not affected (Supplementary Fig. 3A–C). However, there was no difference in the frequency of regulatory T cells (Fig. 3F) in contrast with other treatment regimens (41). Anti-CD3–treated mice showed improved glucose tolerance (Fig. 3G and H), reduced random blood glucose at 13 weeks of age (Supplementary Fig. 3D), an increased islet insulin-positive area as a proxy for β-cell mass (Fig. 3I), and reduced insulitis (Fig. 3J) compared with IgG control mice. These phenotypes occurred without any differences in body weight (Supplementary Fig. 3E). Glucose-stimulated insulin secretion assays on isolated islets of anti-CD3–treated mice showed similar insulin release during high glucose or KCl compared with IgG controls (Supplementary Fig. 3F and G). Thus, our CD3 antibody model slowed disease progression in NOD mice through depletion of antigen-specific CD4⁺ T cells, leading to preservation of β-cells and improved glucose tolerance.

CD3 Antibody Treatment Reduces Accumulation of Senescent β-Cells

We next analyzed β-cell stress and identity markers in the CD3 antibody model. IHC for senescence markers revealed reduced frequency of p21⁺ β-cells in the anti-CD3–treated mice relative to IgG controls (Fig. 4A) and, similarly, fewer β-cells expressing SASP marker Mmp2 (Fig. 4B). Remarkably, there were no differences in the frequency of β-cells expressing nuclear localized sXBP1 (Fig. 4C), the main effector of the IRE1α branch of the adaptive UPR (4). Interestingly, senescence was related to islet insulitis stage, because p21⁺ β-cells were more frequently found in islets with insulitis compared with peri-insulitis and intact islets in IgG control mice, whereas this relationship was lost in anti-CD3–treated mice (Fig. 4D). The frequencies of glucagon-positive cells (α-cells) were not different between IgG- and CD3 antibody–treated groups (Supplementary Fig. 4A), and no differences were found in proliferation as judged by Ki-67⁺ β-cells per islet (Supplementary Fig. 4B). Staining for β-cell identity marker Pdx1 showed that it was maintained in insulin-positive cells in anti-CD3–treated mice and IgG controls (Supplementary Fig. 4C). Changes in senescence markers were also reflected at the mRNA level (Fig. 4E), because isolated islets from anti-CD3–treated mice showed lower expression of Cdkn1a (encoding p21) and SASP genes Mmp2 and Il6 compared with IgG control mice (Fig. 4F). SASP factor secretion from islets of anti-CD3–treated and IgG control mice by antibody array showed reduction of β-cell SASP factor Serpine1 (16) as well as two additional putative SASP factors, cystatin C (42) and osteopontin (43), in anti-CD3–treated mice compared with IgG controls (Supplementary Fig. 4D).

Figure 4.

Anti-CD3 treatment reduces senescent β-cell accumulation in late-stage prediabetic NOD mice. A: IHC for p21 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with anti-CD3 or IgG from 9 to 13 weeks (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of p21⁺ β-cells per islet scored from IgG (n = 88 from 6 mice) and CD3 (n = 96 from 6 mice). B: IHC for Mmp2 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with anti-CD3 or IgG from 9 to 13 weeks (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of Mmp2⁺ β-cells per islet scored from IgG (n = 93 from 6 mice) and CD3 (n = 100 from 6 mice). C: IHC for sXBP1 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with anti-CD3 or IgG from 9 to 13 weeks (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of sXBP1⁺ β-cells per islet scored from IgG (n = 46 from 6 mice) and CD3 (n = 54 from 6 mice) per group. Each data point represents single islet scored, and ≥10 islets were scored for each mouse from ≥2 different regions of pancreas. D: Quantification of p21⁺ β-cells based on islet insulitis score in IgG- and anti-CD3–treated mice from IgG (n = 52 from 6 mice) and CD3 (n = 57 from 6 mice). E: Overview of islet quantitative RT-PCR (RT-qPCR). Islets from NOD mice treated with anti-CD3 or IgG from 9 to 13 weeks were isolated and rested for 24 h in media before harvesting for RNA. F: RT-qPCR of Cdkn1a, Mmp2, and Il6 normalized to Ppia in islets from anti-CD3– (n = 17–25) or IgG-treated mice (n = 18–26). G: Overview of islet mRNA sequencing (mRNA-seq). Islets from NOD mice treated with anti-CD3 or IgG from 9 to 13 weeks were isolated and harvested for mRNA-seq (n = 6 mice per group). H: Volcano plot depicting differentially expressed genes (294 upregulated, 110 downregulated) between anti-CD3 and IgG control mice islets (≥1.5-fold change). I: mRNA-seq normalized read counts for sample of genes differentially expressed, including ribosome and mitochondrial genes (Rpl35, Cox7b), regenerating islet genes (Reg3a, Reg3b), and endoplasmic reticulum (ER) stress and UPR genes (Xbp1, Ddit3). Error bars indicate SD. P values are Hochberg adjusted and significant at P < 0.05 (H and I). *P < 0.05, ****P < 0.0001 by two-tailed unpaired t test (A and B) or two-way ANOVA with Tukey multiple comparisons test (D). ns, not significant. Created in BioRender.

To characterize the global transcriptomic changes in islets after anti-CD3 treatment, we performed mRNA sequencing on freshly isolated islets from IgG control and anti-CD3–treated NOD mice (n = 6 mice per group) (Fig. 4G and Supplementary Fig. 5A). We identified a total of 404 genes differentially expressed, 294 genes upregulated (≥1.5 fold change higher in anti-CD3 treatment group; adjusted P < 0.05), and 110 genes downregulated (≥1.5 fold change higher in IgG controls; adjusted P < 0.05) (Fig. 4H). Gene ontology analysis showed that differentially expressed genes were strongly enriched for ribosome biogenesis, mitochondria, and energy metabolism (Supplementary Fig. 5B and Supplementary Table 2), such as Rpl35 and Cox7b (Fig. 4I). Notably, several regenerating islet (Reg) family genes with anti-inflammatory effects and roles in β-cell protection were among the most highly upregulated (approximately two- to eightfold) in anti-CD3 islets (Fig. 4I, Supplementary Table 2). There was also upregulation of UPR genes Xbp1 and Ddit3 (also known as Chop) in anti-CD3 islets (Fig. 4I), suggesting improved UPR transcriptional activity, which declines during disease progression in NOD mice (6). Quantitative RT-PCR on isolated islets after a 24-h rest period to examine basal UPR transcriptional activity showed reduced expression of Atf4, Ddit3 and Xbp1 compared with IgG controls, whereas sXbp1 levels were not changed (Supplementary Fig. 5C and D), consistent with the IHC of sXBP1 (Fig. 4C). Thus, although CD3 antibody treatment did not alter the extent of sXbp1 mRNA or sXBP1⁺ β-cells, it led to improved UPR gene expression. Importantly, there were no significant differences in islet hormone genes (Ins1, Ins2, Gcg, Sst) or key genes involved in β-cell/islet cell identity or maturation between anti-CD3–treated mice and IgG controls (Supplementary Table 2).

CD4⁺ and CD8⁺ T Cells Are Both Required for Senescent β-Cell Accumulation

To investigate the individual contributions of CD4⁺ and CD8⁺ T cells in β-cell senescence, we depleted each T-cell population in NOD mice from 9 to 13 weeks of age (Supplementary Fig. 6A). CD4 antibody effectively depleted only CD4⁺ T cells (Supplementary Fig. 6B), whereas CD8β antibody similarly depleted only CD8⁺ T cells (Supplementary Fig. 6C), and both treatments led to reduced insulitis compared with IgG controls (Supplementary Fig. 6D). Depletion of either CD4 or CD8 T cells resulted in reduced frequency of p21⁺ β-cells and Mmp2⁺ β-cells compared with IgG control mice (Supplementary Fig. 6E and F). Thus, both CD4⁺ and CD8⁺ T cells are required for β-cell senescence.

CD3 Antibody Is Additive With Senolytic ABT-199 in Slowing Disease Progression and Reducing Accumulation of Senescent β-Cells

Because intermittent anti-CD3 acts to limit β-cell senescence via a different mechanism than senolytics, we reasoned that it would have a complementary effect with the senolytic ABT-199 (16) on disease progression and senescent β-cell accumulation. To test this idea, we treated prediabetic 9-week-old NOD female mice with vehicle only, ABT-199 plus IgG, or ABT-199 plus anti-CD3 for 4 weeks (Fig. 5A). Mice treated with ABT-199 plus anti-CD3 showed the most reduced insulitis (Fig. 5B) and improved glucose tolerance, whereas the ABT-199 plus IgG group showed an intermediate phenotype relative to vehicle controls (Fig. 5C and D), and no differences were found in the insulin-positive stained area between treatment groups (Fig. 5E). In addition, no differences in random blood glucose or body weight were apparent between groups (Supplementary Fig. 7A and B).

Figure 5.

Anti-CD3 has additive effect with senolytic compound ABT-199 in prediabetic NOD mice. A: Schematic representation of combination therapy. NOD mice were administered anti-CD3 (145-2C11 F[ab]₂ fragments) once per week from 9 to 13 weeks and ABT-199 twice per week from 10 to 13 weeks, species-matched IgG F(ab)₂ fragments once per week from 9 to 13 weeks and ABT-199 twice per week from 10 to 13 weeks, or vehicle twice per week from 10 to 13 weeks. B: Quantification of degree of islet immune infiltration in anti-CD3/ABT-199–, IgG/ABT-199–, and vehicle-treated NOD mice after treatment by IHC staining. Data are average percentage of islets showing no infiltration (intact), peri-insulitis, or insulitis from vehicle (n = 87 from 6 mice), IgG/ABT-199 (n = 91 from 6 mice), and CD3/ABT-199 (n = 75 from 6 mice). C: Intraperitoneal glucose tolerance tests (IPGTTs) were carried out in NOD mice after 4-week treatment with anti-CD3 or IgG control (n = 6 mice per group). Data are mean glucose concentration for each time point. No time points showed significant differences by one-way ANOVA with Dunnett multiple comparisons test. D: Area under curve (AUC). E: Quantification of β-cell area in CD3/ABT-199, IgG/ABT-199, and vehicle NOD mice after treatment. Data are squared microns from vehicle (n = 87 from 6 mice), IgG/ABT-199 (n = 91 from 6 mice), and CD3/ABT-199 (n = 75 from 6 mice). F: IHC for p21 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with anti-CD3/ABT-199, IgG/ABT-199, and vehicle (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of p21⁺ β-cells per islet scored from vehicle (n = 87 from 6 mice), IgG/ABT-199 (n = 91 from 6 mice), and CD3/ABT-199 (n = 75 from 6 mice). G: IHC for Bcl-2 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with CD3/ABT-199, IgG/ABT-199, and vehicle (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of Bcl-2⁺ β-cells per islet scored from vehicle (n = 70 from 6 mice), IgG/ABT-199 (n = 58 from 6 mice), and CD3/ABT-199 (n = 58 from 6 mice). H: IHC for Mmp2 (red), insulin (green), and DAPI (blue) in pancreas sections from NOD mice treated with CD3/ABT-199, IgG/ABT-199, and vehicle (scale bar, 40 μm). White box shows zoomed-in view of costained cell (scale bar, 10 μm). Quantification shows mean percentage of Mmp2⁺ β-cells per islet scored from vehicle (n = 58 from 6 mice), IgG/ABT-199 (n = 52 from 6 mice), and CD3/ABT-199 (n = 54 from 6 mice). Error bars indicate SD. **P < 0.005 by χ² test (B). *P < 0.05, ***P < 0.0005, ****P < 0.0001. ns, not significant by one-way ANOVA with Dunnett multiple comparisons test (D–H) . Created in BioRender.

We next assessed β-cell senescence marker p21; prosurvival marker Bcl-2, which is the target of ABT-199–mediated senolysis (16); and SASP marker Mmp2 (Fig. 5F–H). β-Cells expressing the senescence marker p21 and SASP marker Mmp2 were less abundant in both ABT-199 plus IgG and ABT-199 plus anti-CD3 groups relative to vehicle controls, but the magnitude of the reduction was greater in the islets of mice treated with ABT-199 plus anti-CD3 (Fig. 5F and H). However, β-cells expressing Bcl-2 were reduced to a similar extent in ABT-199 plus IgG and ABT-199 plus anti-CD3 groups compared with vehicle controls (Fig. 5G). Notably, there were no differences in sXBP1⁺ β-cells between the groups (Supplementary Fig. 7C). These data indicate that anti-CD3 limits senescent β-cell accumulation in a manner that is complementary to senolytic therapy.

T1D-Related Inflammatory Cytokines Recapitulate Stable Phenotypes of Senescence in Human Islets and EndoC-βH5 Cells

We next investigated the mechanisms by which autoreactive T cells could provoke senescence activation and progression and the relevance to human β-cells. Proinflammatory cytokines IFN-γ and TNF-α are produced by activated helper and cytotoxic T cells (44), and together with macrophage cytokine IL-1β, a mix of these factors is commonly used to mimic the inflammatory islet microenvironment and study β-cell death in T1D. However, it remains unclear whether they can induce a stable senescence response in islets/β-cells.

We first tested whether these cytokines could provoke stable senescence phenotypes in human islets independently of the effects on viability. We tested IFN-γ, TNF-α, and IL-1β (cytokine mix) treatment for 48 h (28,45,46) followed by a 4-day washout period to confirm persistence of the phenotype, alongside our established DNA damage–induced senescence model (27), using 48-h treatment with 50 μmol/L bleomycin and a 4-day washout period as a positive control for stable senescence induction (Supplementary Fig. 8A and Fig. 6A). Notably, human islet viability was consistently no different from that of the control after treatment with the cytokine mix for 48 h with an additional 4-day washout period (Supplementary Fig. 8B and C). Quantitative RT-PCR analysis on islets showed that bleomycin induced expression of CDKN1A but not CDKN2A (encoding p16^INK4A) and also upregulated SASP genes CXCL8 and GDF15 relative to controls, consistent with a DNA damage–mediated senescence transcriptional program (27) (Fig. 6B). DNA damage senescence with bleomycin did not upregulate UPR gene CHOP (also called DDIT3), consistent with the lack of change in UPR genes in this model as previously reported (27). In contrast, inflammatory cytokines stably induced expression of both senescence genes CDKN1A and CDKN2A and SASP gene CXCL8, whereas another SASP gene, GDF15, was not affected, nor was UPR gene CHOP (Fig. 6B). Thus, cytokines led to an alternative senescence transcriptional response compared with bleomycin in that cytokines also induced CDKN2A. Notably, p21 was stably upregulated in islets at the protein level at the same time point (Fig. 6D and E), and levels of p21 started to increase after 48 h of cytokine treatment (Supplementary Fig. 8D).

Figure 6.

Sublethal doses of T1D inflammatory cytokines IL-1β, IFN-γ, and TNF-α partially replicates senescence phenotypes in human islets. A: Schematic representation of human islet experiments. Human islets were treated with media only (untreated); media containing 2 ng/mL IL-1β, 10 ng/mL TNF-α, and 10 ng/mL IFN-γ (cytokine mix; 48-h treatment, 4-day washout); or media containing 50 μmol/L bleomycin (bleomycin; 48-h treatment, 4-day washout) as DNA-damage positive control for senescence induction. Details of human islets used for each experiment are reported in Supplementary Table 1. B: Quantitative RT-PCR (RT-qPCR) analysis for CDKN1A, CDKN2A, CXCL8, and GDF15 normalized to PPIA in human islets after treatment with cytokine mix (n = 28–29 independently treated wells of islets from 6 donors) or bleomycin (n = 18 independent treated wells of islets from 4 donors) or no-treatment control (n = 28–29 independently treated wells of islets from 6 donors) 4 days postdrug washout. Data are means. C: RT-qPCR analysis for CHOP normalized to PPIA in human islets after treatment with cytokines or bleomycin or no-treatment control 4 days postdrug washout from same donors and independently treated wells of islets as in panel B. Data are means. D: Western blot analysis and relative quantification of p21 in male donor islets after treatment with cytokine mix or bleomycin 4 days postwashout. Quantification shows average p21 levels normalized to β-actin as loading control for n = 3 biologic replicates from single donor. E: Western blot analysis and relative quantification of p21 in female donor islets after treatment with cytokine mix or bleomycin 4 days postwashout. Quantification of p21 levels normalized to β-actin as loading control for n = 4 biologic replicates from single donor. F: Schematic representation of EndoC-βH5 experiment. EndoC-βH5 cells were treated with media only (untreated); media containing 2 ng/mL IL-1β, 10 ng/mL TNF-α, and 10 ng/mL IFN-γ (cytokine mix; 96-h treatment, 24-h washout); or media containing 35 μmol/L bleomycin (bleomycin; 48-h treatment, 72-h washout). G: RT-qPCR analysis for CDKN1A, CDKN2A, GDF15, and CXCL8 normalized to PPIA for EndoC-βH5 cells after treatment with cytokine mix or bleomycin. Data are mean of n = 7–8 biologic replicates. H: ELISA for GDF15 on conditioned media from EndoC-βH5 cells treated as in panel F. Data are mean of n = 4 biologic replicates per group. Error bars indicate SD. *P < 0.05, **P < 0.005, ***P < 0.0005, ****P < 0.0001. ns, not significant by one-way ANOVA with Dunnett multiple comparisons test. Created in BioRender.

To further investigate the role of cytokine stress in triggering stable senescence programs specifically in human β-cells, we turned to EndoC-βH5 cells (Fig. 6F and Supplementary Fig. 8E), a fetal-derived human β-cell line that has been negatively selected to remove proliferating cells (28). These cells also mount a stable senescence-like stress response to DNA damage with bleomycin similar to human islets (27) and can model β-cell cytokine-induced death responses (28,47). Treatment of these cells for 96 h with cytokines followed by a 24-h washout led to a modest reduction in viability (Supplementary Fig. 8F), but the remaining viable cells showed higher expression of CDKN2A, CXCL8, and GDF15 compared with controls (Fig. 6G). This slightly differed from bleomycin treatment, which induced CDKN1A but not CDKN2A and similarly upregulated SASP genes CXCL8 and GDF15 (Fig. 6G). Flow cytometry showed that live cytokine-stressed EndoC-βH5 cells increased intracellular expression of p21 but not SASP factor GDF15 (Supplementary Fig. 8G–I). However, GDF15 secretion was increased from cytokine-exposed EndoC-βH5 cells (Fig. 6H). Taken together, these data suggest that chronic exposure to cytokine stress can recapitulate some of the stable senescence phenotypes in human islets and EndoC-βH5 cells elicited by DNA damage.

Discussion

Senescence is a mechanism of β-cell stress that contributes to T1D in NOD mice and occurs during early presymptomatic stages of T1D in humans (17). However, the mechanisms that lead β-cells to enter senescence and why this subpopulation accumulates during T1D development are not known. Here, we demonstrate that senescence in β-cells is largely, although not exclusively, dependent on damage from autoreactive CD4⁺ and CD8⁺ T cells during T1D development in NOD mice. Studies on genetically disease-resistant mouse models demonstrated a marked reduction in accumulated senescent β-cells, and this could be reconstituted through adoptive transfer of diabetogenic splenocytes in NSG mice. Intriguingly, β-cell senescence was not abolished in the immune-deficient models, demonstrating that β-cell–intrinsic and disease-independent mechanisms are involved and lead to some senescence, even in the absence of destructive autoimmunity.

To probe the specific role of autoreactive T cells in NOD mice during the late prediabetic disease stages where CD4⁺ and CD8⁺ T-cell effector function is dominant and senescent β-cells accumulate, we established an intermittent anti-CD3 paradigm. Our approach differed from the many earlier regimens tested with this agent (40), particularly those used for diabetes reversal in NOD mice with repeated dosing for 5 consecutive days (37) or very early stage single dosing in young NOD mice (38). Once-weekly intermittent depletion of CD3⁺ T cells with anti-CD3 reduced the CD4⁺ T-cell population and MHC II–presented insulin-specific T cells but did not affect MHC I–presented IGRP-specific T cells or total CD8⁺ T cells. This was consistent with the previously reported preferential effect of this CD3 antibody on the CD4⁺ T-cell population as a result of their higher CD3 expression compared with CD8⁺ T cells in NOD mice (48). Our anti-CD3 model did not increase regulatory T-cell numbers, as observed with previous anti-CD3 regimens (41); nevertheless, it led to improved glucose tolerance, reduced insulitis, and increased β-cell area. Remarkably, anti-CD3 treatment specifically reduced the senescent β-cell population without also reducing the frequency of β-cells expressing sXBP1, the main marker of the IRE1α branch of the adaptive UPR (4). However, anti-CD3 improved UPR gene expression, shown by increased Xbp1 and Ddit3 levels in mRNA sequencing analysis and reduced basal UPR gene expression in islets from anti-CD3 mice compared with IgG controls after resting. β-Cell proliferation, α-cell numbers, and β-cell differentiation/identity genes were also not affected after anti-CD3 treatment, ruling out that these other pathways were responsible for the phenotypic effect. We also established that both CD4⁺ and CD8⁺ T cells drive β-cell senescence. Anti-CD3 treatment had an additive effect with senolytic ABT-199 in reducing senescent β-cell accumulation, consistent with its different mechanism of action, and again, no effect was found on sXBP1⁺ β-cells. These data together strongly support the conclusion that senescent β-cell accumulation is reliant on the destructive activity of autoreactive CD4⁺ and CD8⁺ T cells, whereas the IRE1α-sXBP1 adaptive UPR does not rely on the autoimmune process to the same extent.

How autoreactive CD4⁺ and CD8⁺ T cells drive β-cell senescence remains unclear and is difficult to investigate in vivo. Therefore, we explored a role for T1D-related inflammatory cytokines and ascertained relevance to human β-cells in vitro. Using our established human islet and EndoC-βH5 cell DNA damage–mediated senescence models as controls (27), we showed that sublethal exposure to proinflammatory cytokines IL-1β, IFN-γ, and TNF-α can trigger stable senescence responses at the transcriptional and protein levels. Although these cytokines are widely used to study β-cell death, comparatively few studies have interrogated stable, nonlethal stress responses activated by these factors. We monitored cell viability and incorporated a washout period to rule out cytokine-induced transcriptional responses that are reversible (49,50). Cytokine-mediated stress may contribute to β-cell senescence activation and progression, but additional studies involving coculture of autoreactive T cells with β-cells will be required to address additional mechanisms of T cell–mediated stress aside from inflammatory cytokines that could provoke the senescence response.

Our study has implications for understanding the mechanisms by which CD3 immunotherapy slows progression during presymptomatic stage 2 T1D in humans (1). Our findings suggest that reduced β-cell senescence might be an additional mechanism by which teplizumab slows disease progression from stage 2 to 3 T1D in humans. In conclusion, our findings provide insights into the mechanisms that lead to β-cell senescence activation and accumulation during T1D development. Our work also suggests a novel mechanism of action for CD3 immunotherapy in T1D and supports the rationale for combining senescence-targeted therapy with anti-CD3 to further slow disease progression and protect β-cells in presymptomatic stages of T1D.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29129000.

Funding Statement

J.P. was supported by a University of Manitoba Graduate Fellowship, the Canada Graduate Scholarship–Master’s Award, and a health sciences PhD studentship from Children’s Hospital Research Institute of Manitoba and Research Manitoba. N.R.M. was supported by a postdoctoral training award from the Canadian Islet Research and Training Network and Breakthrough T1D. The laboratory of P.J.T. was supported by a Canadian Institutes of Health Research project grant (PJT-479641), a new investigator operating grant from Research Manitoba (5351), the Canada Foundation for Innovation John R. Evans Leaders Fund (43168), and the 2023 Manitoba Medical Services Foundation Allen Rouse Basic Science Career Development Award.