Extending Remission and Reversing New-Onset Type 1 Diabetes by Targeted Ablation of Autoreactive T Cells

Type 1 diabetes develops silently and clinical presentation occurs only after the destruction of a biologically sufficient mass of insulin-producing β-cells, yet an estimated 30% of β-cell mass remains at disease onset (1). The preservation of insulin and C-peptide, the portion of the proinsulin molecule cleaved and secreted in equimolar concentration to insulin, is known to have significant clinical benefits in preventing severe hypoglycemia, retinopathy, nephropathy, and neuropathy (2–4). To preserve the β-cells, however, subsequent autoimmune attack must be prevented. Although detectable insulin production can persist for years or even decades in some patients (5), the functional capacity of β-cells decreases in response to continued immune-mediated damage by an average of 40% within a year after diagnosis (6). Thus, interventions that halt further immune damage or induce the durable re-establishment of immune tolerance to β-cells remain both imperative and elusive. Moreover, without functional immune tolerance to β-cells, recent technical strides in restoring, regrowing, or transplanting islets are likely to fail.

Although a durable means of controlling islet antigen–specific T cells remains to be elucidated, the salient qualities of an effective therapy are clear. Given the established efficacy of insulin replacement therapy, any direct control or elimination of autoreactive T cells must be well tolerated with minimal off-target effects, must be specific and spare naive and memory T cells required for robust immunity to pathogens, and must target the bulk of autoreactive T cells, even those for which the antigen specificity is unknown. As such, the ideal approach would target one or more intrinsic traits shared by all activated autoreactive effector cells at the time of disease onset.

Recently, we identified such requisite properties in activated effector T (Teff) cells that permit their targeted elimination while sparing naive T-cell, regulatory T (Treg) cell, and memory T-cell subsets (7). Teff cells are exceptional for their extraordinarily rapid rate of cell division (8) and exhibit significant spontaneous DNA breakage and concomitant DNA damage response (DDR) upon activation (7). Initiation of the DDR results in either cell cycle arrest, where progression through the cell cycle is prevented until DNA damage is repaired, or in apoptosis. This cell fate decision depends on the graded accumulation of activated p53 (phospho-Ser15) in the cell (9). Because activated p53 accumulates, the tipping point is reached at which the cell undergoes apoptosis if the damage is too severe to repair (9). We reasoned that since Teff cells balance on this edge of apoptosis, by using small-molecule inhibitors targeting distinct proteins in the DDR pathway such as CHK1/2 and WEE1, which govern cell cycle arrest (in both S phase and at the G2/M checkpoint), Teff cells would enter mitosis prematurely (7,10–12), and we could force the selective apoptosis of Teff cells with their amplified p53 while sparing other T-cell subsets shielded by virtue of their slower proliferative rate (13) and diminished DDR (7). When these are combined with an inhibitor of mouse double minute 2 homolog (MDM2; the negative regulator of p53), the accumulation of activated p53 is further potentiated (14), the net result of which is a synergistic increase in apoptosis. Importantly, cells not in cycle or without a concomitant DDR, such as naive T cells, Treg cells, and quiescent memory T cells, would be largely spared. We previously showed that the combination of these inhibitors, which we have termed “p53 potentiation with checkpoint abrogation” (PPCA), conferred significant therapeutic efficacy in treating mouse models of multiple sclerosis (experimental autoimmune encephalomyelitis) and hemophagocytic lymphohistiocytosis (7).

We hypothesized that PPCA administration at the first onset of disease would specifically reduce or eliminate diabetogenic Teff cells, thus halting further damage to β-cells and sustaining residual endogenous insulin production. Herein, we present evidence that PPCA is effective and well tolerated in the treatment of type 1 diabetes in multiple clinically relevant circumstances by selectively targeting activated diabetogenic T cells, while exhibiting minimal off-target effects on naive T-cell, Treg cell, and memory T-cell populations. Thus, by manipulating p53 and cell cycle checkpoints, we can exploit the endogenous response of activated T cells in a novel and promising new approach to re-establishing immune tolerance in type 1 diabetes in a durable, selective, and nonharmful way.

NOD/ShiLtJ (NOD), NOD.Cg-Tg(TcraBDC2.5, TcrbBDC2.5)1Doi/DoiJ (BDC2.5.NOD), NOD.129S7(B6)-Rag1^tm1Mom/J (NOD.Rag^−/−), NOD.B6-Ptprc^b/6908MrkTacJ (CD45.2 NOD), and C57BL/B6 (B6) mice were purchased from The Jackson Laboratory; bred; and maintained under specific pathogen-free conditions in accordance with institutional animal care guidelines at Cincinnati Children’s Hospital Medical Center Vivarium.

Prediabetic NOD mice were monitored beginning at 10 weeks of age for clinical signs of prediabetes (hyperglycemia, assessed by nonfasting blood glucose [BG] testing). Diabetes onset was determined via two consecutive BG readings of ≥200 mg/dL. Mice were monitored triweekly from this point until they developed end-stage disease (BG ≥600 mg/dL); lost >20% body weight; established long-term normoglycemia for at least 60 days; or died.

All chemotherapeutics were administered intraperitoneally, and were dosed as follows: AZD1775 (WEE1 inhibitor [WEE1i]; ChemieTek), 40 mg/kg; AZD7762 (CHK1/2 inhibitor [CHKi]; Selleck Chemicals), 25 mg/kg; and nutlin-3 (MDM2 inhibitor nutlin-3 [MDM2i]; Cayman Chemical), 50 mg/kg. Vehicle was made as previously described (7).

Single-cell suspensions of spleen and pancreatic lymph nodes were separated by density gradient centrifugation using Lympholyte-M and incubated with tetramer (National Institutes of Health Tetramer Core). Cells were then stained for surface expression of the indicated antibodies. For intracellular staining, cells were fixed with the FoxP3 fixation/permeabilization buffer and permeabilized with permeabilization buffer. Flow data were collected using an LSR Fortessa System using FACS Diva software and analyzed using FlowJo (Tree Star).

Single-cell suspension of pancreatic cells was prepared using the Miltenyi Biotec gentleMACS Dissociator, filtered through a 70 μm cell strainer, and prepared according to the Miltenyi Biotec Debris Removal Solution protocol. Isolated lymphocytes were then stained in accordance with the methods listed above.

Pancreas was formalin fixed and paraffin embedded. Sections were stained with VectaFluor R.T.U. Antibody Kit (Vector Laboratories). Images were taken and analyzed using NIS Elements Imaging Software. Insulitis was scored as follows: 0, no visible infiltration; 1, peri-insulitis; 2, insulitis with <50% islet infiltration; 3, insulitis with >50% islet infiltration; and 4, islet scar.

T cells were prepared for adoptive transfer in accordance with the MojoSort Mouse CD3 T Cell Isolation Kit (BioLegend) protocol. A total of 10⁷ cells were injected via retro-orbital i.v. injection in 250 µL of DMEM.

PolyA mRNA purification, quality control, and next-generation sequencing (NGS) were performed by the University of Cincinnati Genomics, Epigenomics, and Sequencing Core on an Illumina HiSEq 2500 System at 50 million raw reads per sample, 150–base pair paired ends. AltAnalyze was used to map reads (via built-in Kallisto software), create expression files, and perform initial GO-Elite functional annotation with reads per kilobase of transcript per million mapped reads (less than one removed; GO-Elite filtered based on the minimum fold change of 2.0; P < 0.05, t test). Principal components analysis was performed within both AltAnalyze and MATLAB. Gene set enrichment analysis (GSEA) was run through the corresponding modules within GenePattern. Gene sets run on GSEA were collected from the Molecular Signatures Database version 6.1 and ImmGen. Additional comparisons of upregulated and downregulated genes (2.0-fold, P < 0.05) were compared with annotated gene lists using ToppGene. Heat maps were generated in R.

Islet transplants were performed as described previously (15). Briefly, streptozotocin (STZ) (200 mg/kg) was administered to NOD.Rag^−/− mice to induce diabetes, after which ∼300 islets isolated from B6 mice were transplanted under the left kidney capsule.

All statistical analyses were performed using one-way or two-way ANOVA, log-rank Mantel-Cox test, or multiple t test using GraphPad Prism version 6.07 for Windows.

At the onset of type 1 diabetes, some residual β-cells remain, yet, absent intervention, they will continue to be destroyed by acutely activated diabetogenic T cells over time. Therefore, we reasoned that since PPCA specifically targets activated, proliferating Teff cells, the treatment of new-onset disease in NOD mice with PPCA should affect long-term glycemic control. We used two different PPCA combinations, a CHKi, and a WEE1i, each coupled with the MDM2i. Since inhibiting either the CHK1/2 or WEE1 checkpoint kinase prevents the inhibitory phosphorylation of CDKs, cycling Teff cells skirt the normal S and G2 DNA damage regulatory checkpoints (10,11,16) and enter mitosis prematurely; the concurrent inhibition of p53 negative regulator MDM2 (14) augments the accumulation of phospho-p53, and, in the setting of mitotic dysregulation, drives a strong and synergistic increase in apoptosis of these Teff cells. To test this, we treated NOD mice immediately after the onset of spontaneous diabetes (200–250 mg/dL BG) for 3 successive days with PPCA and monitored subsequent changes in BG levels. This treatment strategy was used because 3 days of treatment was shown to be effective in reducing BG levels while still being well tolerated. Given that PPCA requires Teff cells to be in rapid cell cycle for maximum efficacy and that not all islet-reactive T cells are in the same synchronous activation state, it was not surprising that we had to retreat most new-onset mice multiple times to maintain a BG level <200 mg/dL; the number of retreatments varied, but mice receiving WEE1i+MDM2i received on average 7.5 courses of treatment. NOD mice with untreated new-onset diabetes were used as a control group for comparison. The mice with new-onset diabetes treated with either PPCA combination therapy maintained glycemic control (BG <200 mg/dL) for 51% (WEE1i+MDM2i) and 53% (CHKi+MDM2i) of the enrolled days (a period ranging from 21 to 234 days) (Fig. 1A). This is in stark contrast to the untreated mice, which were only able to maintain glycemic control for 5% of the days they were enrolled (a period ranging from 34 to 72 days) (Fig. 1A). Although PPCA ameliorated disease in all spontaneously diabetic NOD mice, it was most efficacious when given to NOD mice with new-onset disease with BG levels between 200 and 250 mg/dL at the time of disease onset, suggesting that PPCA is most efficient when used early, at a time when significant β-cell function remains. In addition, PPCA-treated mice had significantly decreased incidence of end-stage diabetes (BG ≥600 mg/dL), a metric used because it is indicative of a point at which endogenous insulin production, as measured by C-peptide levels (<150 ng/mL) (Supplementary Fig. 1), was minimal. Although all untreated mice progressed to end-stage diabetes in an average of 25 days, 71% of PPCA-treated mice never reached end-stage disease (seven of nine mice treated with WEE1i+MDM2i and five of eight mice treated with CHKi+MDM2i) (Fig. 1B). Both PPCA combinations yielded comparable results overall, but WEE1i+MDM2i performed slightly better at preventing progression to end-stage disease; therefore, we opted to use this combination for subsequent experiments, unless otherwise noted. These results corresponded with significant preservation of functional β-cells at day 60 or end-stage disease after onset and treatment. Mice treated with PPCA showed a marked preservation of islet mass, with comparable levels of insulitis and islet damage to prediabetic mice and significantly fewer severely infiltrated or destroyed islets than vehicle-treated controls (Fig. 1C and D).

To determine the effects of PPCA therapy on immune responses, we quantified the T-cell response in mice treated with or without PPCA. Lymphocytes were isolated from the spleen on day 4 after disease development after 3 days of treatment and were stained with tetramers for insulin-specific T cells (Fig. 2A). After treatment with PPCA, the absolute numbers of both total activated (CD44⁺) CD4⁺ and CD8⁺ Teff cells as well as islet antigen–specific T cells were significantly decreased compared with those isolated from the spleens of vehicle-treated and prediabetic controls (CD4⁺, Fig. 2B; CD8⁺, Fig. 2C), which is consistent with the decreased number of activated diabetogenic cells displaying increased levels of DNA damage markers in PPCA-treated mice compared with vehicle-treated controls (Supplementary Fig. 2).

Because insulin and islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) tetramer reagents only identify a fraction of the total diabetogenic T cells, we assessed the overall pathogenicity of the T-cell compartment after PPCA in toto by adoptive transfer. As in previous experiments, we treated NOD mice with new-onset diabetes for 3 days with either PPCA or vehicle and harvested lymphocytes 1 day after the last treatment. We then enriched lymphocytes from the spleen and pancreatic lymph nodes for CD3⁺ T cells, adoptively transferred 10⁷ CD3⁺ T cells (CD45.1⁺) into CD45.2 NOD.Rag^−/− mice, and monitored for the development of disease. Mice that were injected with cells from PPCA-treated mice developed disease with a significantly lower incidence than those receiving cells from vehicle-treated controls, demonstrating that PPCA ablates diabetogenic T cells such that they are significantly less able to transfer disease (Fig. 3) and suggesting that PPCA reduced the overall islet antigen–reactive pools of CD4⁺ and CD8⁺ T cells.

Importantly, the reduction in islet antigen–specific Teff cells in both the spleen and the pancreas occurs without significantly decreasing the absolute numbers of other T-cell subsets, such as naive (CD44^loCD62L⁺), regulatory (CD4⁺Foxp3⁺), and memory populations (Fig. 4A–C and Supplementary Fig. 3). This is consistent with data showing that only PPCA treatment significantly decreased cells with DNA damage markers in activated diabetogenic cells, and not in naive or regulatory cells (Supplementary Fig. 4). Moreover, the diminution in Teff cells without significant loss in total and antigen-specific Treg cells resulted in a significant increase in the Treg/Teff ratio, both in the spleen and, more prominently, in the pancreas (Fig. 4D and E). Interestingly, PPCA is significantly less effective in prolonging the remission in new-onset disease when disease was induced using cyclophosphamide (Supplementary Fig. 5), which is known to deplete Treg cells (17). Together, this suggests that, even without a total elimination of all diabetogenic Teff cells, the clinical benefit resulting from their significant reduction is likely due to the ability of the largely unaltered islet-specific Treg-cell population to exert regulatory control and functional tolerance.

We would predict from our mechanistic suppositions that after PPCA therapy the remaining activated T cells would manifest an enriched Treg cell molecular signature, while at the same time they would display reduced molecular evidence for cell cycle and DNA replication. To this end, we undertook NGS analysis of RNA from T cells sort purified from PPCA- or vehicle-treated new-onset diabetic NOD mice, which were, as described above, treated with PPCA or vehicle for 3 days, and on day 4 CD4⁺ T cells were sort purified into naive (CD44^−/lo), activated (CD44^hi), and activated insulin-specific (CD44^hitet⁺) subsets; lysed; and subjected to NGS. As expected, the naive CD4⁺ T cells from both groups had RNA expression signatures that differed from all the activated T-cell subsets but not from each other, suggesting that PPCA does not alter the gene expression profile of naive T cells (Supplementary Fig. 6). However, as predicted, both the activated and activated insulin antigen-specific subsets from PPCA- and vehicle-treated mice showed disparate expression profiles, which stratified by drug treatment (Fig. 5A) but not tetramer reactivity, as both subsets of CD44^hiCD4⁺ T cells (tetramer reactive and nonreactive) showed similar and highly related expression profiles, and as such could be effectively treated as replicates rather than as distinct subpopulations (Fig. 5A and Supplementary Fig. 6). Using GSEA (18), we observed the highly significant and predicted loss of G2/M phase gene expression in the activated subsets of T cells from PPCA-treated mice when compared with those from vehicle-treated mice (Fig. 5B). Similar levels of both Ki67 and γH2AX in the spleen and pancreas indicate that this is also an accurate reflection of the apoptosis of activated T cells in the islets (Supplementary Fig. 7). Importantly, PPCA preferentially augmented the Treg cell molecular signature in activated CD4⁺ T cells (Fig. 5C); moreover, the leading-edge subset of Treg cell–expressed genes (Fig. 5D) included Ikzf4 (Eos), which is essential for inhibitory Treg cells (19,20). We also obtained results showing that 11 of 200 annotated Treg cell genes were upregulated in the PPCA subsets (Supplementary Fig. 8). Together, these data, along with the preservation of Foxp3⁺ Treg cells (Fig. 4), provide the best evidence to date that PPCA therapy largely spares natural and antigen-specific Treg cells, while dramatically reducing the numbers of activated T cells in the cell cycle (specifically at the critical G2/M border).

One of the aims of reestablishing immune tolerance to islets is the ability to restore endogenous insulin production by islet transplantation. To determine whether PPCA could protect islet grafts from autoreactive T cells, ∼300 islets from B6 mice were transplanted under the kidney capsules of NOD.Rag^−/− mice rendered diabetic by prior treatment with STZ. After transplantation, the NOD.Rag^−/− mice regained euglycemia within 24–48 h, were rested for >12 days to allow for graft vascularization, and were then challenged with 10⁶ diabetogenic T cells (i.v.) isolated from BDC2.5 TcR/NOD.Rag^−/− mice. Previously published studies show that 10⁶ T cells transfer diabetes with near 100% efficacy in islet-engrafted NOD.Rag^−/− mice (15). After T-cell transfer, transplanted mice received either PPCA or vehicle (day 2–3 post-transfer); all but one mouse receiving PPCA was protected from diabetes, whereas mice that received vehicle developed end-stage diabetes by day 30, with one exception likely the result of suboptimal cell transfer (Fig. 6). Upon nephrectomy of the engrafted kidney, transplanted PPCA-treated mice uniformly and immediately became diabetic again (BG >600 mg/dL), indicating that the islet graft was functional and responsible for the normoglycemia observed (Fig. 6). The robustness of PPCA in the transplant setting has the added advantage of engendering a more synchronous reactivation of T cells, in this case islet-reactive T cells, making them more effectively targeted by PPCA therapy.

Many patients with type 1 diabetes retain β-cell function and substantial endogenous insulin and C-peptide secretion at diagnosis (5), and, given the clinical significance of preserving this production and the rapidity with which it falls within a year after diagnosis (2,6), a tractable intervention to halt disease progression early in the course of disease is imperative. We have demonstrated that PPCA is able to significantly preserve β-cell function by preventing further lymphocytic infiltration through the ablation of activated diabetogenic Teff cells, which notably is not accompanied by a decrease in the naive T-cell, Treg cell, or memory T-cell populations, or by off-target effects on the gut, bone marrow, and thymocytes, or the ability to clear concurrent viral infections (7). Our data suggest that the mechanistic durability of PPCA lies in its reversal of the balance of self-reactive Teff cells and their Treg cell counterparts; the decrease in antigen-specific Teff cells with the attendant preservation of Treg cells resulted in a significantly enhanced Treg/Teff ratio in the spleen, and far more extensively in the pancreas itself (Fig. 4). We also demonstrate a significant enrichment of Treg cell gene expression in activated PPCA-treated cells (Fig. 5C). Although the initial break in tolerance in type 1 diabetes seems to be due in part to defects in the Treg cell compartment, there is also evidence that the Teff cells are resistant to regulation (21) and that even a minor reduction in Treg cells due to reduced interleukin-2Rβ signaling is enough to accelerate diabetes onset (22), suggesting that reducing Teff cells may be the most effective strategy in attempting to reverse new-onset disease. This is supported by a recent study (23) showing protection from disease development in NOD mice due to gut metabolite-induced changes, either reducing the autoreactive Teff cell compartment or augmenting the Treg cell compartment, although each independently provided protection from disease, the diet that reduced the autoreactive Teff cells showed greater efficacy.

Why some effector cells escape PPCA is unclear; however, this may reflect the short temporal window in which PPCA operates and the activation state of diabetogenic T cells at the time of treatment or it may also be explained by the modulation of Bcl-2 family members in response to genomic stress. It is also possible that administering additional treatments after remission has been achieved and the Treg/Teff ratio has been altered, we can continue to ablate the surviving autoreactive cells, based on a recent study showing that Treg cells induce DNA damage in Teff cells via metabolic competition (24).

Although many similarities exist between the NOD model and human type 1 diabetes, there are both physiological and environmental differences that may impact the translation of PPCA treatment to human disease. The NOD mouse model is a far more fulminant form of diabetes, wherein mice lose insulin production with greater rapidity than is generally seen in humans (25). In another marked difference from the situation that would occur in human diabetes, when mice developed clinical disease, we did not administer exogenous insulin. The exposure to insulin, whether from the higher levels of endogenous production or exogenous administration, would serve to further exacerbate the response of autoreactive cells specific to insulin antigens, almost certainly increasing the percentage of autoreactive cells that are activated and the amount of DDR signaling present. In addition, we have previously found that human T cells have an even greater sensitivity to DDR manipulation than mouse T cells and exhibit a greater selectivity to the inhibitors used (7). Together, this would suggest that PPCA is likely to be even more effective in humans with new-onset diabetes than in mice.

In addition to its implications for insulin and C-peptide preservation in new-onset disease, PPCA has significant translational potential in islet transplantation. Through a variety of methodologies (regrowth of islets, autologous graft through induced pluripotent stem cell differentiation, or allograft), islet restoration or transplantation has become an increasingly technically feasible option; however, to date the treatment showing the most efficacy is long-term therapy with broad immunosuppressive drugs (26). In contrast, PPCA has the potential to be equally or more effective without the need for protracted use, and without any impact on protective immunity, including on defenses against concurrent infections, and thus has the potential to succeed as an immune intervention against both autoreactive and alloreactive T cells, providing durable tolerance to transplanted islets. Moreover, our data model the coordinated activation that occurs when islets are restored in a patient; by giving them recognizable antigen, the resulting temporally uniform recall response provides an optimal therapeutic window for maximal PPCA efficacy, potentially eliminating islet-reactive cells from the T-cell repertoire entirely. Our data (Fig. 6) suggest that in the case of this concurrent cellular expansion, PPCA is so efficacious in most cases that there is no need to reset a regulatory balance; the adoptively transferred diabetogenic T cells are from BDC2.5.TcR/NOD.Rag1^−/− mice that lack Treg cells, underlining the extent of effector cell ablation that can be achieved, although the possibility that some Teff cells may have converted to Treg cells has not been formally excluded. PPCA thus would be an excellent candidate therapy adjunctive to, or even in replacement of, the current long-term immunomodulatory therapies used to promote islet graft tolerance.

In conclusion, we have found that the targeted manipulation of p53 and cell cycle checkpoints selectively kills activated autoreactive T cells in vivo; that this is an effective treatment strategy for type 1 diabetes in multiple clinically relevant circumstances; and that it shows no effect on naive T-cell, Treg cell, or memory T-cell populations, and significantly increases the ratio of islet-specific Treg/Teff cells. These results provide the foundation for a promising new avenue of immune intervention in type 1 diabetes.

Acknowledgments. The authors thank the National Institutes of Health Tetramer Core for the Insulin A_94–108 I-A^g7, Insulin B_12–20 I-A^g7, Proinsulin_47–64 I-A^g7, IGRP_206–214 H-2K^d, and LCMV GP_33–41 D^b tetramers. The authors also thank the Cincinnati Children's Hospital Medical Center (CCHMC) Pathology Core for assistance in processing histological samples; the University of Cincinnati Genomics, Epigenomics, and Sequencing Core for assistance with RNA sequencing; as well as Dr. Emily Miraldi (CCHMC) and Dr. Harinder Singh (CCHMC) for invaluable feedback and help with RNA sequencing analysis.

Funding. This work was supported by National Institute of Health/National Institute of Diabetes and Digestive and Kidney Diseases grants R01-DK-081175 (to D.A.H. and J.D.K.) and R01-AI-109810 (to D.A.H. and M.B.J.) and National Institute of Arthritis and Musculoskeletal and Skin Diseases grant P30-AR-047363 (to CCHMC Cincinnati Rheumatic Diseases Center Animal Models of Inflammatory Disease Core).

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

Author Contributions. K.R.C. performed experiments and acquired data; interpreted and analyzed data; contributed to research design, analysis, and final approval; and wrote the manuscript. E.E.E. performed experiments and acquired data. J.J.S. performed experiments and acquired data and interpreted and analyzed data. J.P.M. interpreted and analyzed data and contributed to research design, analysis, and final approval. D.A.H. interpreted and analyzed data; contributed to research design, analysis, and final approval; and reviewed and edited the manuscript. M.B.J. contributed to research design, analysis, and final approval and reviewed and edited the manuscript. J.D.K. interpreted and analyzed data; contributed to research design, analysis, and final approval; wrote the manuscript; and reviewed and edited the manuscript. J.D.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.