Reversal of Experimental Autoimmune Diabetes With an sCD39/Anti-CD3 Treatment

The purine ATP, released from cells as extracellular ATP (eATP), participates in cellular homeostasis (1,2) by signaling via purinergic (P2; ionotropic–P2X, and metabotropic–P2Y) receptors expressed on the cell surface of several tissues (1,2). eATP is released by stressed and dying cells and triggers proinflammatory signals by binding P2X7R (3,4). Abrogating eATP signaling, through its receptor P2X7R, has been proven to successfully inhibit the alloimmune response in different settings of allotransplantation and to delay the onset of experimental autoimmune diabetes in NOD mice (1,2,5–9). eATP is cleared by the ectonucleotidase CD39, which converts ATP into AMP, and by the ectonucleotidase CD73, which dephosphorylates AMP into adenosine (10–13). CD39 and CD73 activities are both essential for the conversion of eATP and could have important implications in mediating immune responses in the pancreatic microenvironment and/or systemically, particularly in light of the potent anti-inflammatory properties of adenosine (14,15). So far, very few treatments have been capable of halting the destruction of pancreatic β-cells in patients newly diagnosed with type 1 diabetes (T1D) (16,17). This is partially related to the complex etiopathogenesis of T1D, which remains elusive, and a better understanding of the underlying physiopathological mechanisms may be needed in order to assist in developing effective therapies that could better preserve the functional β-cell mass (18). The current study is aimed at assessing the role of eATP and its degradation during the onset of experimental autoimmune diabetes.

Under protocols approved by the University of Miami Institutional Animal Care and Use Committee (A-3224-01), mice were obtained from The Jackson Laboratory. The strains used in the current study included NOD/ShiLtJ (001976; NOD), NOD.NON-Thy1^a/1LtJ (004483; NOD.Thy1.1), NOD.Cg-Tg (TcraBDC2.5,TcrbBDC2.5) 1Doi/DoiJ (004460; NOD.BDC2.5.Thy1.2), and NOD.Cg-Tg (TcraTcrbNY8.3) 1Pesa/DvsJ (005868; NY8.3-NOD) T-cell receptor (TCR)–transgenic NOD mice, and C57BL/6J mice (000664). Diabetes was defined as nonfasting glycemic values ≥250 mg/dL in 2 consecutive days. eATP levels were measured in serum samples obtained from different groups of NOD mice and were analyzed with the Enliten ATP assay system (FF2000, Promega) (19). Soluble CD39 (4 units/day; sCD39/apyrase [APY]; Sigma-Aldrich) was administered for 2 weeks via subcutaneous osmotic pumps (Alzet.com), with or without a 5-day course (intravenous injection of 50 μg/day dose) of hamster anti-mouse CD3 IgG antibody (Ab) (αCD3; clone 145-2C11 from ATCC.org) (20). Histology and insulitis score was performed as described by our group (21), and analysis was performed using ImageJ software (imagej.nih.gov/ij) (22).

For in vivo functional assays, CD4⁺ splenocytes from NOD.BDC2.5.Thy1.2 mice were isolated by positive selection with microbeads (LT34; Miltenyi Biotech), followed by FACS-enrichment based on CD4⁺Vβ4⁺ expression (achieving ≥99% purity). Similarly, CD8⁺ splenocytes from NY8.3-NOD mice were positively selected using CD8a microbeads (Ly-2; Miltenyi Biotech), followed by FACS-enrichment based on CD8⁺Vβ8⁺ expression (achieving ≥99.5% purity). After CellTrace Violet (CTV; Life Technologies) labeling, aliquots of 1.7 × 10⁶ CD4⁺ cells or 2 × 10⁶ CD8⁺ cells were adoptively transferred to NOD.Thy1.1 female mice that had osmotic pumps (Durect Corp.) releasing PBS or APY implanted subcutaneously 1 week before adoptive transfer. Animals were humanely sacrificed 3 days after adoptive transfer, and their pancreatic lymph nodes (pLNs), inguinal LNs, spleen, and peripheral blood monocyte cells (PBMCs) were collected for phenotypic analyses of the expression of Foxp3 and CD39 in host CD4⁺ and CD8⁺ T cells, as well as in donor CD4⁺Vβ4⁺Thy1.2⁺ and CD8⁺Vβ8⁺Thy1.2⁺ T cells in which proliferation was also assessed by CTV dilution.

Spleen, pLNs, and pancreas were digested (37°C, 20 min) with collagenase-D (2 mg/mL; Roche.com). PBMCs were purified from whole blood on Ficoll-PaquePlus gradient (GElifesciences.com), and erythrocytes were lysed with ACK (Ammonium-Chloride-Potassium) Lysing Buffer (Cytiva). Nonspecific binding was prevented by incubation with rat–anti-CD16/32 Ab (clone 24G2). Cells were then stained with fluorescence-conjugated Abs against mouse CD45, CD3, CD4, CD8, CD11b, CD11c, B220, GR-1, CD39, and CD73 (BioLegend). Intracellular staining relied on a fluorescence-conjugated anti-mouse/rat Foxp3 Ab kit (clone FJK-16s; eBioscience). Samples were acquired and analyzed using an LSRII cytometer (BD Biosciences) and Kaluza Flow Analysis Software (Beckman Coulter), respectively, with assistance from the DRI Flow Cytometry Core.

Splenic CD4⁺ T cells isolated by positive selection with CD4 microbeads (LT34; Miltenyi Biotech) were labeled with 5 mmol/L CTV and cultured at a 1:1 bead-to-cell ratio for 72 h in 96-well plates with mouse αCD3/CD28 monoclonal (m)Ab-coated Dynabeads (Cytiva). Cells were cultured for 4 days in the presence of no cytokines (Th0 polarization), interleukin 2 (IL-2) (10 ng/mL; R&D Systems), and transforming growth factor-β (10 ng/mL; R&D Systems), (regulatory T cells [Treg]); or IL-6 (5 µg/mL; R&D Systems), transforming growth factor-β (5 ng/mL), anti–IL-4 (10 µg/mL; Peprotech), and anti–interferon-γ (10 µg/mL; Peprotech) (T helper 17 [Th17]). Then, cells were washed, pulsed 4 h with phorbol myristate acetate (PMA; 50 ng/mL; Sigma-Aldrich), ionomycin (1.0 µg/mL; Sigma-Aldrich), and GolgiStop (1.0 µL/mL; BD Biosciences) and then analyzed by FACS for expression of activation markers (CD44 and CD25), cytokines (IL-2, IL-10, IL-4, IL-17, and interferon-γ), and transcription factors specific for Tregs (Foxp3) and Th17 (RORγt). For analysis, cells were gated in the lymphocyte gate based on forward and side scatter (FSC/SSC), doublets excluded by FSC/SSC-height and FSC/SSC-width. Dead cells were excluded by gating on cells negative for the viability marker Live/Dead (Cytiva). Viable and activated CD4⁺ T cells were defined based on their CD25⁺ expression. Further analysis was done within the activated CD4 gate (CD4⁺CD25⁺).

Pancreata were collected in 10% buffered formalin solution or frozen at −80°C in optimal cutting temperature compound (Tissue-Tek; VWR). Pancreatic sections were stained with the following Abs: primary Abs used were rabbit αCD3 (1:100; Cell Marque), rabbit anti-CD31 (1:50; Abcam.com), sheep anti–mouse-CD39 (1:25; R&D Systems), sheep anti–mouse-CD73 (1:10; R&D Systems), rat anti-mouse/human B220 (1:50; eBioscience), rat anti–mouse-CD8 (1:25; BD Biosciences), rat anti-mouse/rat Foxp3 (1:25; eBioscience), rabbit anti-glucagon (1:100; BioGenex), and polyclonal guinea pig anti-insulin (1:100; Dako). Secondary Abs (all from Cytiva) included goat anti-guinea pig Alexa Fluor 488 and 647 (1:200), goat anti-rat Alexa Fluor 568 and 488 (1:200), goat anti-rabbit Alexa Fluor 488 and 555 (1:200), and donkey anti-sheep Alexa Fluor 568 and 647 (1:500). Images were obtained on a SP5 inverted confocal microscope (Leica). Analysis was performed using ImageJ software (https://imagej.nih.gov/ij/) (22).

Survival curves were compared by log-rank test using GraphPad Prism 7.0 software, and data were expressed as median with range. Two-tailed, unpaired Student t test for two-group comparisons and one-way ANOVA for multiple comparisons were used. Data are expressed as mean ± SEM. P values <0.05 were considered statistically significant.

In the view to determine whether eATP levels and CD39/CD73 expression play a role during the onset of T1D, we assessed the peripheral levels of eATP as well as the expression of CD39 and CD73 in NOD mice during the progression of experimental autoimmune diabetes. eATP levels were increased at the time of diabetes onset in NOD mice, while being reduced in those NOD mice that were naturally protected from diabetes (e.g., long-term normoglycemic) (Fig.1A). The expression of CD39 and CD73 was evident in insulin⁺ and glucagon⁺ cells, while being also expressed by CD31⁺ endothelial cells and CD3⁺ T cells (Fig.1B–E and Supplementary Fig. 1A and B). CD39 expression colocalized with a proportion of CD3⁺ cells in 4-week-old NOD mice, in long-term normoglycemic NOD mice, and in hyperglycemic NOD mice (Fig.1B and C and Supplementary Fig. 1A and B). A proportion of CD3⁺ cells coexpressed CD39 and CD73 in all different groups of NOD mice, particularly in the prediabetic NOD mice (Fig.1C). Our data indicate increased eATP levels were observed in hyperglycemic mice, so overtly diabetic mice and the presence of an expression of ectonucleotidases within α-/β-cells, T cells, and endothelial cells.

Flow cytometric analysis revealed an increased percentage of CD3⁺CD73⁺CD39⁺ cells within the pancreas and the pLNs of long-term normoglycemic NOD mice (Fig.2A), and the absolute number of CD3⁺ T cells in the pancreas retrieved, respectively, from 4- and 10-week-old, hyperglycemic and long-term normoglycemic NOD mice is shown in Supplementary Table 1. Higher proportions of CD3⁺CD73⁺CD39⁺ cells were detected within islet-infiltrating T cells from hyperglycemic NOD compared with prediabetic 4-week-old NOD mice (Fig.2A and Supplementary Fig 2A–C).

Next, to determine whether dysregulated expression of CD39 and CD73 was present and could be correlated to the functional status of CD4⁺ T cells during the progression of experimental autoimmune diabetes, CD4⁺ T cells obtained from NOD mice were exposed to different in vitro Th-polarizing conditions (Fig.2B–E). Under Th17 polarization, CD4⁺ splenocytes obtained from hyperglycemic NOD mice displayed the highest expression of IL-17 compared with the other groups (Fig.2B and Supplementary Fig. 2D). During Th17 polarization, the frequency of CD4⁺CD25⁺CD73⁺CD39⁺ cells was significantly higher in long-term normoglycemic NOD mice (Fig.2C). Long-term normoglycemic NOD mice displayed significantly higher Foxp3 expression compared with other groups under Treg polarization conditions, having the highest percentage of CD73⁺CD39⁺ Tregs compared with hyperglycemic and 10-week-old NOD mice (Fig.2D and E and Supplementary Fig. 2E). These data indicate that the CD25⁺CD73⁺CD39⁺ population decreased in hyperglycemic NOD mice relative to long-term normoglycemic NOD mice.

Encouraging results obtained in inflammation models have prompted the development of recombinant human APY or sCD39 capable of degrading eATP (23,24). By administering sCD39 in NOD mice, we evaluated the effect of the modulation of eATP signaling. We first observed that serum levels of eATP decreased significantly in sCD39-treated NOD mice compared with controls (Fig.3A). Flow cytometric analysis on splenocytes obtained from sCD39-treated mice showed an increased frequency and absolute numbers of host CD4⁺Foxp3⁺ and CD8⁺Foxp3⁺ regulatory T cells compared with controls (Fig.3B and C, Supplementary Fig. 2F, and Supplementary Table 2). Treatment with sCD39 also resulted in significantly increased CD39 expression in both splenic CD4⁺ and CD8⁺ T cells (Fig.3D and E). To determine whether sCD39 treatment has an effect on the distribution of pathogenic autoreactive T cells in vivo, we adoptively transferred NOD.Thy1.1 mice with autoreactive CD4⁺Vβ4⁺NOD.BDC2.5.Thy1.2⁺ or CD8⁺Vβ8⁺NOD.8.3Thy1.2⁺ cells, and at the time of adoptive transfer, we injected these recipient mice with sCD39. At harvesting, mice injected with sCD39 displayed decreased frequency and absolute numbers of host Thy1.2⁺CD4⁺ cells within their pLNs compared with controls (Fig.3F and G and Supplementary Table 2). Paralleling the decrease in autoreactive T cells, significantly lower frequencies and absolute numbers of proliferating CD4⁺ autoreactive T cells (Fig.3H and I and Supplementary Table 2), as well as higher proportions and absolute numbers of host Tregs (Thy1.2⁺CD4⁺Foxp3⁺ cells) (Fig.3J and K and Supplementary Table 2) and higher Treg proliferation rates and absolute numbers, were also found in the pLNs of sCD39-treated NOD mice than in the controls (Fig.3L and M and Supplementary Table 2). On the contrary, no significant changes were noted in the percentage of Thy1.2⁺Vβ8⁺CD8⁺ cells within the pLNs of sCD39-treated NOD mice and controls (Fig.3N and O and Supplementary Table 2), albeit lower proliferating CD8⁺ autoreactive T cells were noted in the former (Fig.3P and Q). Importantly, mice treated with sCD39 showed significantly higher frequencies and absolute numbers of peripheral host CD4⁺Foxp3⁺ (Fig.3R and Supplementary Table 2) and of splenic host CD4⁺Foxp3⁺ and CD8⁺Foxp3⁺ Tregs compared with controls (Fig.3S and T and Supplementary Table 2). These data show that sCD39 administration may affect the function of autoreactive T cells and Tregs.

We then tested the effect of sCD39 administration on the onset of experimental autoimmune diabetes in a prevention study in NOD mice. After 2 weeks of treatment, serum levels of eATP were significantly lower in sCD39-treated mice than in controls (Fig.4A). Pancreata of sCD39-treated mice displayed improved insulitis scores and higher proportions of insulitis-free islets compared with controls (P < 0.0001) (Fig.4B and C). Immunophenotyping performed after 2 weeks of treatment revealed an increased percentage of Foxp3⁺ on CD8⁺ T cells from sCD39-treated NOD mice compared with controls (Fig.4D). The percentage of CD4⁺/CD8⁺ CD25⁺Foxp3⁺ T cells was significantly higher in pLN/splenocytes (Fig.4E) and of CD8⁺ CD25⁺Foxp3⁺ T cells in PBMCs (Fig.4F) of sCD39-treated NOD mice compared with controls. Finally, higher frequencies of CD4⁺CD39⁺, of CD4⁺CD73⁺, and of CD4⁺CD39⁺CD73⁺ T cells were observed in sCD39-treated NOD mice compared with controls (Fig.4G–I). Furthermore, while sCD39 only slightly delayed the onset of hyperglycemia (Fig.4J), a reduced peri-insular cell infiltrate was observed compared with controls (Fig.4K). Overall, these data show that prediabetic mice treated with sCD39 displayed a more benign course of T1D, while showing a more immunoregulatory phenotype.

The effect of sCD39 was further explored in vivo in a diabetes-reversal study. We sought to determine whether a combination of short-course treatment with αCD3 and sCD39 could promote diabetes remission in newly hyperglycemic NOD mice (Fig.4L and M). While, all control mice received the insulin pellet, only 12 mice returned to hyperglycemia within 1 month (median 17.5 days) (Fig.4M), and hyperglycemic NOD mice that were treated with sCD39 alone returned to hyperglycemia within a median of 27 days (Fig.4M). Transient diabetes remission was observed in hyperglycemic NOD mice that received αCD3 alone, with hyperglycemia recurring in 80% of mice after a median time of 46 days (P = 0.0399 in αCD3 alone vs. sCD39+αCD3) (Fig.4M). Importantly, a combination therapy of sCD39 and αCD3 showed a strong synergism with an extended duration of the remission of the disease, and 45% of the treated mice remained diabetes-free until the end of their follow-up (median 88 days) (Fig.4M). αCD3 and sCD39 treatment synergistically reverted experimental autoimmune diabetes in NOD mice.

In this study, we delineated the role of the eATP/CD39 pathway during the onset of experimental autoimmune diabetes. We observed an increase in the peripheral levels of eATP in NOD mice and an expression of CD39/CD73 in the islets and in infiltrating T cells. In a prevention study in NOD mice, the administration of sCD39 abrogated the proliferation of autoreactive T cells, while its coadministration with αCD3 synergistically reversed experimental autoimmune diabetes in NOD mice. eATP is known to activate proinflammatory pathways in lymphocytes (1,2,5), and it is also involved in the priming of T lymphocytes during the onset of T1D (7). The putative sources of eATP can be endocrine islet cells, vascular structures, monocytes, and passenger leukocytes; indeed, eATP can be released by the islets and can be sensed by many immune cells in the islet microenvironment (7,19,25). An altered β-cell health might be associated with supraphysiological eATP release in the islet microenvironment and may disrupt the immune system homeostasis, thus facilitating the onset of T1D (7). Interestingly, CD39 and CD73, the main regulators of eATP levels, are expressed by β-cells and islet-infiltrating T cells, thus confirming pancreatic expression shown previously (26–28). It is conceivable that the net balance between eATP release in response to stress conditions, along with its degradation by the ectonucleotidases CD39/CD73 expressed on endocrine and immune cells, determines the degree of microenvironment activation and islet cell fate in subjects at risk to develop T1D. Interestingly, enhancing eATP degradation promotes Treg expansion (29) and increases the frequency of immune cells with a putative “regulatory” phenotype (e.g., expression of both CD39 and CD73).

Indeed, based on our published data and on published reports, where we delineated a new role for the eATP/P2X7R axis during Th17 generation in T1D autoimmunity (7), we thus studied in depth Th17/Treg generation. Furthermore, we and others have reported that eATP can be sensed by P2X7R, which is considered a fundamental step of T-cell activation during antigen recognition, Th1/Th17 generation, and alloimmunity, so we assessed the percentage of CD39/CD73 during Th17/Treg generation. A CD39⁺CD73⁺ T-cell population has been previously associated with a regulatory phenotype in mice (12,30), because Tregs use both CD39 and CD73 to degrade eATP into adenosine and suppress T-cell responses (12,31). CD73⁺CD39⁺ immune cells were particularly enriched in the pancreas and pLNs of long-term normoglycemic NOD mice. The administration of sCD39 was associated with a reduction in the peripheral levels of eATP in NOD mice as well as with a reduction of the proliferation and the homing of autoreactive T cells and with an expansion of islet specific Tregs. This could be related to the early role of eATP in the priming/generation of Th17/Tregs, which is in line with several published data confirming that the modulation of CD39 boosted/reinforced Tregs function in Crohn disease (32). Notably, administration of sCD39 induced remission of diabetes in 25% of hyperglycemic NOD mice and synergized with αCD3 in an even stronger remission rate. This has important therapeutic implications, particularly in light of the promising results of targeting CD3 T cells in patients at risk for or newly diagnosed with T1D in clinical trials (17,33–39).

The current study has a few limitations, which could be assumed as follows: 1) the relevance of the NOD murine model for human T1D; 2) the lack of any data related to the long-term efficacy of the therapy >100 days; 3) the need of starting the treatment close to the onset of the disease; and finally, 4) the use of an anti-CD3 mAb, which may be mitogenic and thus may induce eATP increase along with the known “flu-like” syndrome. The testing of an anti-CD3 Ab, which contains nonmitogenic F(ab')2 fragments, is deemed safer and should be considered as future perspective. However, we should also mention that a certain degree of T-cell activation is observed with both mitogenic and nonmitogenic anti-CD3 mAbs (38,39). 5) We acknowledge that we treated a larger number of mice (n = 24 NOD mice) with αCD3 compared with the other groups; indeed, the rationale behind this choice resides in the large variability of diabetes remission observed in NOD mice, which may vary from 100%, 80%, to 15%, as reported in the literature (40–42).

Collectively, our study indicates that eATP signaling may have a relevant role in the pathophysiology of experimental autoimmune diabetes in NOD mice and demonstrates that systemic modulation of eATP signaling is an appealing adjuvant strategy to enhance the efficacy of immunotherapies and preserve β-cell mass in T1D.

Acknowledgments. The authors are grateful to Drs. Alberto Pugliese and Jay S. Skyler (University of Miami), and Fabio Grassi (Institute for Research in Biomedicine, Bellinzona, Switzerland) for constructive intellectual input. Special thanks for technical assistance to Alejandro Tamayo-Garcia, Yelena Gadea, and Elsie Zahr-Akrawi (DRI Preclinical Cell Processing and Translation Models Core), Kevin Johnson (DRI Histology Laboratory), and Marcia Boulina (DRI Imaging Core).

Funding. This study received support from the Diabetes Research Institute Foundation (www.diabetesresearch.org). P.F. is supported by the Italian Ministry of Health grant RF-2016-02362512. V.U. is supported by the Fondazione Diabete Ricerca (FO.DI.RI) Società Italiana di Diabetologia (SID) Fellowship.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

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

Author Contributions. C.F., R.D.M., A.L.B., and A.P. designed and performed experiments, analyzed data, and wrote the manuscript. C.F. and A.P. conceived the studies. M.B.N., U.U., H.B.B., M.E.L., V.U., A.J.S., S.A.K., C.L.S., and J.G. performed experiments. O.U., C.A.F., A.J.M., C.R., and P.F. provided reagents, intellectual feedback, and reviewed the manuscript. A.P. 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.