PD-1, but Not PD-L1, Expressed by Islet-Reactive CD4⁺ T Cells Suppresses Infiltration of the Pancreas During Type 1 Diabetes

NOD mice were purchased from Taconic (Germantown, NY). NOD.BDC2.5 TCR mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and crossed to NOD.Thy1.1⁺ mice. C57BL/6.PD-1–deficient mice (25) were backcrossed 13 generations, and PD-L1–deficient mice (7) were backcrossed 15 generations to the NOD background. PD-1 and PD-L1 knockout (KO) NOD.BDC2.5.Thy1.1 mice were generated by crossing NOD.BDC2.5.Thy1.1 with NOD.PD-1^+/− (backcross 13) and NOD.PD-L1^+/− (backcross 15) mice, and F1 mice were intercrossed to produce NOD.BDC2.5.Thy1.1.PD-1^−/− and NOD.BDC2.5.Thy1.1.PD-L1^−/− mice, respectively. Prediabetic NOD mice were used as recipients for BDC2.5 T cells between 7 and 12 weeks of age. Animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Minnesota.

Naïve NOD.BDC2.5.Thy1.1⁺ T cells were enriched from spleen and inguinal, axillary, brachial, cervical, and periaortic lymph nodes (LNs) between 4.5 and 6.5 weeks of age using negative selection with magnetic columns (Miltenyi Biotec, Auburn, CA). Anti-biotin microbeads (Miltenyi) were used to deplete cells based on biotinylated anti–granulocyte differentiation antigen 1, major histocompatibility class (MHC)-II, CD8α, CD117, CD24, CD25, TER119, CD19, B220, and CD44. Cells were ≥98% CD44^low. We transferred 7,500 or 1 × 10⁶ naïve CD4⁺.BDC2.5 T cells intravenously into NOD recipients. In vitro activated BDC2.5 T cells were generated as previously described (5). Pancreata were harvested for histology at diabetes onset.

Anti–PD-L1 antibody (clone M1H6) was generated as previously described (26). Rat IgG was purchased from Jackson ImmunoResearch (West Grove, PA). Antibody (200–250 μg/mouse/injection) was injected intraperitoneally every other day prior to harvest for two to three injections or diabetes onset as indicated in the figure legends. Blood glucose levels were measured, and >250 mg/dL was considered diabetic (LifeScan, Milpitas, CA).

BDC2.5 T cells were enriched from single cell suspensions from spleen and/or pooled non–antigen draining LN samples (27). Anti-Thy1.1-APC (eBioscience, San Diego, CA) and anti-APC microbeads were used to enrich BDC2.5.Thy1.1 cells by positive selection magnetic columns (Miltenyi). Single cell suspensions from pancreas were generated using collagenase P digestion (Roche, Indianapolis, IN) and discontinuous percoll separation as previously described (28). Antibodies included anti-Thy1.1, -CXCR3, –PD-1, –interleukin (IL)-2, –γ-interferon (IFN-γ), –tumor necrosis factor (TNF)-α (eBioscience), and -Ki67 (BD Biosciences, San Jose, CA). Cell numbers were determined using AccuCheck Counting Beads (Life Technologies, Grand Island, NY).

Flow cytometry was performed using BD LSRII and Fortessa instruments (BD). Data were analyzed using FlowJo (Tree Star, Ashland, OR). The gating strategy to define BDC2.5 T cells was singlet⁺, CD3⁺ lineage⁻ (B220⁻, CD11b⁻, and CD11c⁻), CD4⁺, and Thy1.1⁺ cells.

Pancreata were frozen in optimal cutting temperature (Tissue-Tek; VWR, Arlington Heights, IL), cut at 7 μm, and fixed using acetone. Antibodies included guinea pig anti-swine insulin (Dako, Denmark), donkey anti–guinea pig AF488 (Jackson ImmunoResearch), Thy1.1-PE (clone OX7; BioLegend, San Diego, CA), and CD3e-APC (eBioscience). Slides were mounted using Prolong Gold with DAPI (Life Technologies) and imaged on a Leica epifluorescent DM5500 microscope (Wetzlar, Germany). Data were compiled using MetaMorph software (Molecular Devices, Sunnyvale, CA).

Single cell suspensions were stimulated in complete Dulbecco’s modified Eagle’s medium (5) with 10 μg/mL brefeldin A, 100 ng/mL phorbol 12-myristate 13-acetate (PMA), and 1,000 ng/mL ionomycin (Sigma-Aldrich, St. Louis, MO). Cells were incubated for 4 h at 37°C and 5% CO₂. BD cytofix/cytoperm fixation/permeabilization kit was used for intracellular cytokine staining.

Purified BDC2.5 T cells were labeled with carboxyfluorescein-succinimidyl ester (CFSE) using the manufacturer’s protocol (Life Technologies). Intracellular staining for Ki67 (BD) was performed using a Foxp3 fixation/permeabilization kit according to the manufacturer’s protocol (eBioscience).

Statistical analysis was performed using GraphPad Prism 5 software (La Jolla, CA). P values were calculated using unpaired two-tailed Student t tests or Mann-Whitney U tests using a 95% CI.

To investigate the role of PD-1 in regulating islet-reactive CD4⁺ T cells during T1D progression, we established a model to track these cells in vivo. We used the BDC2.5 transgenic mouse, which contains CD4⁺ T cells specific for an islet peptide–MHCII (pMHCII) complex (15,16). We did not provide exogenous antigen or adjuvant but, rather, relied on presentation of endogenous levels of autoantigen to activate the cells. We speculated that since levels of endogenous islet-pMHCII complexes are low, precursor frequency would dictate the ability of T cells to become activated (11). To test this hypothesis, we transferred naïve BDC2.5 cells into NOD mice and compared the frequency of cells that increased CD44 expression after a low (7,500 cells) or high (1 × 10⁶ cells) transfer (Fig. 1A and B). At day 6 posttransfer, a significantly higher frequency of BDC2.5 cells had converted to CD44^high in the low transfer compared with the high transfer in the pancreas-draining LN (pLN) and spleen and non–antigen draining LNs (nondLNs) (Fig. 1A and B). Importantly, at a later time point (day 27), many of the BDC2.5 cells remained naïve in the high-transfer situation, while nearly the entire population in the low-transfer setting had encountered autoantigen (Fig. 1B). For this reason, the remainder of this study used the low-transfer model.

To further characterize the low-transfer BDC2.5 model, we transferred 7,500 naïve BDC2.5 cells into NOD mice. Approximately 10–15% of the BDC2.5 cells survived the transfer (data not shown), providing a naïve preimmune repertoire of ~750 cells. Although this is higher than what has recently been reported for endogenous preimmune CD4⁺ T cells using pMHCII tetramers (20–400 cells) (29–31), it is orders of magnitude closer to reality than other transfer models (5,6). After low transfer, the majority of BDC2.5 cells (73 ± 6%) in the pLN increased CD44 expression indicating that they encountered autoantigen by day 3 posttransfer, while a minority of cells in the spleen and nondLNs expressed CD44 (34.8 ± 6.2) (Fig. 1C). At time points 28 days and later, there was no significant difference in CD44 levels between these organs (Fig. 1C). We conclude from these data that the entire population converted to the CD44^high phenotype by day 28, though the majority was CD44^high by day 7 (Fig. 1C).

We next enumerated BDC2.5 cells to determine the shape of the clonal expansion curve. Endogenous priming led to a 2.5-fold increase in BDC2.5 cells at day 7 (Fig. 1D), which was low compared with a classic acute infection response (30). Interestingly, we did not measure a contraction of antigen-specific T cells at later stages (post–day 21). Importantly, BDC2.5 cells accumulated in the pancreas and localized to the infiltrate surrounding the islet (peri-insulitis) (Fig. 1E and F). BDC2.5 were rarely found deep within the islet core (insulitis) (Fig. 1F). The frequency and localization of BDC2.5 cells in the pancreas differ substantially from in vitro activated and transferred BDC2.5 cells, which destroy the pancreas within 1 week (5,8) by mediating more destructive insulitis (Fig. 1G) than the low-transfer model (Fig. 1F).

Lastly, we determined the effector potential of transferred BDC2.5 cells. We measured the transcription factor T-bet, which controls the differentiation of Th1 cells (32). After activation, the majority of BDC2.5 cells differentiated into T-bet⁺ effector cells with significant production of IFN-γ in the pLN and pancreas (Supplementary Fig. 1A–C). These cells did not express Foxp3 (data not shown). This low-transfer system provided a novel tool to examine the role of PD-1 in the regulation of an autoreactive CD4⁺ T-cell population in vivo.

We next used the low-transfer model to evaluate the requirement for PD-1/PD-L1 interactions during T1D. We determined that BDC2.5 cells in the pLN were PD-1⁺ at day 3, consistent with recent T-cell activation (Fig. 2A). PD-1 expression on the BDC2.5 T cells in the spleen and nondLNs was lower at day 3 than in the pLN, as was expected, since a higher frequency of BDC2.5 cells was naïve in this location (33). At day 7, there was a decrease in PD-1 expression in the pLN compared with day 3, though the significance of this decline is unclear. By day 21 posttransfer, PD-1 levels were elevated in the pLN, spleen, and nondLNs, and these levels remained elevated. By day 7, PD-1 levels were highest in the pancreas and remained highest in this organ (Fig. 2A and B), which could reflect chronic antigen engagement.

We next determined whether blocking PD-1 influenced the BDC2.5 cells. Three weeks posttransfer, anti–PD-L1 resulted in a significant increase in BDC2.5 cells in the pLN and pancreas (Fig. 2C). There was no statistically significant increase in BDC2.5 cells in the non–antigen draining inguinal LN (iLN) after PD-1 blockade (Fig. 2C), consistent with PD-1 functioning in an antigen-dependent manner (34). An increase in the number of BDC2.5 cells was also observed in the spleen, which was likely due to increased migration of cells from the pLN, since antigen was unlikely to be present. Lastly, we sought to determine whether anti–PD-L1 administration impacted the spatial organization of BDC2.5 cells within the pancreatic infiltrate. At 3 weeks posttransfer, the majority of pancreas infiltrate was peri-insulitis (Fig. 2E). However, anti–PD-L1 administration caused a substantial conversion from peri-insulitis to destructive insulitis (Fig. 2E), consistent with previous work (18). Importantly, BDC2.5 cells were present within the islet core after PD-1 blockade (Fig. 2D). Anti–PD-L1 accelerated T1D in this model (Supplementary Fig. 1D), consistent with previous reports in WT NOD mice (18). These data highlight the significant pathogenic effect of anti–PD-L1 administration on BDC2.5 cells in the pancreas during T1D.

Previous work identified a role for PD-1 in regulating T-cell proliferation, survival, and cytokine production using in vitro assays and high–cell transfer models in vivo (6,7,24,35). However, the role for PD-1 inhibition in an endogenous antigen system has not been tested. Additionally, since the majority of work focused on early time points (<10 days posttransfer), little is known about how PD-1 regulates CD4⁺ T cells at chronic time points. We therefore tested the role of PD-1 in regulation of BDC2.5 cell proliferation early during the activation phase (day 7) and late during the chronic phase (day 42). At day 7, anti–PD-L1 resulted in a significant increase in the frequency of BDC2.5 cells that fully diluted CFSE in the pLN (Fig. 3A and B), consistent with previous work (6). We did not measure a significant difference in the frequency of CFSE-diluted BDC2.5 cells in the iLN or the pancreas. We also found an increase in CFSE-diluted BDC2.5 cells in the spleen after PD-L1 blockade and speculate that the splenic increase is due to increased trafficking from the pLNs.

We next tested the role of PD-1 regulation on proliferation during the chronic phase. We treated with anti–PD-L1 or control antibody 6 weeks posttransfer and measured Ki67, a molecule associated with cell cycle entry. As predicted, the BDC2.5 cells from control-treated animals were quiescent (Ki67⁻) (Fig. 3C and D). PD-1 blockade resulted in a significant increase in the frequency of BDC2.5 cells entering cell cycle (Ki67⁺) in the spleen, pLN, and pancreas (Fig. 3C and D). PD-1 blockade did not significantly increase Ki67⁺ BDC2.5 cells in the iLN, as expected, since this organ lacks pancreatic antigens (Fig. 3C and D). Collectively, these data suggest that PD-1 regulates CD4⁺ T-cell proliferation early during priming (Fig. 3A and B) and late during the persistent phase of the response (Fig. 3C and D). These findings extend our knowledge of how PD-1 regulates proliferation, demonstrating that loss of PD-1 can restore proliferative capabilities in CD4⁺ T cells at late time points in vivo.

Our results indicated that PD-1 blockade increased BDC2.5 cell infiltration of the pancreas (Fig. 2C). Therefore, we determined whether PD-1 regulated expression of a chemokine receptor associated with trafficking to the pancreas. CXCR3 sensing CXCL10 promotes effector Th1 cell trafficking to pancreatic islets during T1D (36). Additionally, CXCR3 is required for optimal Th1 differentiation in antigen-draining LNs (37). PD-1 blockade resulted in the significant increase of CXCR3 expression on BDC2.5 cells in the spleen and pLN (Fig. 4A and B). PD-1 blockade also resulted in a highly significant increase in the absolute number of CXCR3-expressing BDC2.5 cells in the spleen and pLN (Fig. 4C and D). Interestingly, the highest CXCR3 expression was on Ki67^high T cells after PD-1 blockade (Fig. 4E). This increased CXCR3 expression may aid in optimal Th1 differentiation and trafficking of islet-reactive CD4⁺ T cells to the pancreas during T1D pathogenesis.

We hypothesized that PD-1 blockade would enhance the capacity for BDC2.5 cells to produce effector cytokines similarly to findings with islet-reactive CD8⁺ T cells (35,38). Contrary to our prediction, PD-1 blockade did not alter the frequency of BDC2.5 T cells producing the cytokines IFN-γ, TNF-α, and IL-2 (Fig. 5A–C and Supplementary Table 1). However, PD-1 blockade resulted in a significant increase in the number of cytokine-producing BDC2.5 cells in the pLN and pancreas (Fig. 5D and E). These data suggest that PD-1 blockade does not alter the capacity for cytokine production but significantly affects the quantity of cytokine-producing cells.

Our results demonstrate that PD-1/PD-L1 blockade enhanced proliferation and infiltration of the pancreas in this low-transfer model. Previous work has demonstrated roles for both PD-1 and PD-L1 expressed by effector CD8⁺ T cells in regulating function (38,39); however, the specific requirement for PD-1 or PD-L1 on antigen-specific CD4⁺ T cells during T1D has not been studied. Therefore, we generated BDC2.5 mice deficient for PD-1 or PD-L1 (Supplementary Fig. 1F). NOD.BDC2.5.PD-1^−/− and NOD.BDC2.5.PD-L1^−/− mice developed normally, and both strains displayed accelerated T1D compared with WT BDC2.5 mice (Supplementary Fig. 1E). However, disease mean onset was delayed compared with reports of non-TCR transgenic NOD.PD-1^−/− and NOD.PD-L1^−/− mice (7,17). Importantly, all donor mice used for these experiments were euglycemic at the time of BDC2.5 cell transfer.

We next tested the role of PD-1 and PD-L1 on BDC2.5 cells in the low–cell transfer model in NOD recipients. This approach allowed us to determine the requirement for PD-1 and PD-L1 on the T cell in a WT environment. The park rates between WT and KO BDC2.5 cells were comparable (data not shown). At day 7 posttransfer, we observed a statistically significant increase in PD-1–deficient BDC2.5 cells in the spleen, pLN, and pancreas compared with PD-1–sufficient controls (Fig. 6A). There was not a significant increase in PD-1–deficient BDC2.5 cells in the iLN. Conversely, there was no difference in the number of PD-L1–deficient BDC2.5 cells compared with PD-L1–sufficient controls in any organ (Fig. 6B). These data suggest that PD-1, but not PD-L1, expressed by the BDC2.5 cell was required to suppress proliferation and infiltration of the pancreas.

We next measured CXCR3 expression on PD-1-KO.BDC2.5 cells after low transfer and found significantly higher levels of CXCR3 in the spleen and pLN compared with WT (Fig. 6C and D). Importantly, CXCR3 was not elevated in the intact PD-1-KO.BDC2.5 mouse compared with WT BDC2.5 mice (data not shown). Additionally, since naïve cells were transferred (Supplementary Fig. 1G), all BDC2.5 cells were CXCR3⁻ at the time of transfer.

Lastly, we determined how the loss of PD-1 selectively on the BDC2.5 cells impacted localization within the pancreas. Interestingly, PD-1-KO.BDC2.5 cells could be detected both around and deep within the islet core (Fig. 6E). Mice that received PD-1-KO.BDC2.5 cells displayed more severe insulitis compared with mice that received WT BDC2.5 (Fig. 6F). This suggests that loss of PD-1 on the BDC2.5 cells promotes the ability to penetrate into the core of the islet. These data support a model where PD-1 expressed by islet-reactive CD4⁺ T cells regulates proliferation, CXCR3 expression, and pancreas infiltration.

The critical link between the PD-1/PD-L1 pathway and constraint of T1D has been reported (17,18); however, the cellular mechanisms are unclear. In this study, we developed a new model system to investigate how autoreactive CD4⁺ T cells respond to endogenous autoantigen and the role for PD-1. Our study showed that PD-1 on the BDC2.5 cell regulated proliferation, CXCR3 expression, pancreas infiltration, and autoimmune diabetes.

Current knowledge of how CD4⁺ T cells are regulated during spontaneous T1D progression is limited owing to a lack of tools to track these cells in vivo. Our study used the transfer of a low number of naïve autoreactive T cells to mimic rare self-reactive T cells in the preimmune repertoire (27,29). Precursor frequency can influence the magnitude of an immune response, and previous reports showed that transferring nonphysiologically high numbers of TCR transgenic T cells can lead to abnormal T-cell activation and function (12–14). We reasoned that precursor frequency would be critically important in T1D because of limited endogenous self-pMHC complexes. Indeed, analysis of CD44^high frequencies in BDC2.5 cells after low or high transfer revealed that a significantly higher frequency of cells encountered antigen and became activated in the low-transfer system. Importantly, in the high transfer a large fraction of cells failed to encounter antigen at later time points even though antigen persisted, presumably because of competition for limiting pMHCII (11). In addition to antigen being a limiting factor in activation, we postulate that tolerance mechanisms in NOD mice are also limiting. Overwhelming the system with large numbers of either naïve or previously activated cells reduces the likelihood that endogenous regulatory mechanisms (e.g., PD-L1 in the islet) could induce or maintain tolerance. For studying how peripheral tolerance is maintained and why it breaks down, we believe models that more closely mimic physiological conditions should be used. Lastly, this low-transfer approach allowed us to investigate diabetogenic T-cell dynamics in the natural inflammatory environment in NOD mice during T1D pathogenesis, which cannot be recapitulated with in vitro systems, priming regimens, or mice that have transgenic expression of model antigens in the pancreas (5,8,24,35,38,40). By limiting the precursor frequency, we were able to define the dynamics of autoreactive T-cell activation in response to endogenous autoreactive pMHCII complexes.

The role for PD-1/PD-L1 interactions in regulating T cells has been investigated previously in vitro and in vivo (19). Current knowledge of how PD-1 regulates CD4⁺ T cells during T1D in vivo is largely limited to the secondary lymphoid organs, while less is known about the pancreas. Work from Guleria and colleagues investigated mechanisms of PD-1–mediated regulation of antigen-specific CD4⁺ T cells in NOD mice (6), including BDC2.5 cells. However, a high number of BDC2.5 cells were transferred into recipient mice, and a large fraction of these cells failed to proliferate in the pLN even after anti–PD-L1 administration. In our low-transfer model, the majority of the BDC2.5 cells in the pLN diluted CFSE after anti–PD-L1 administration, highlighting an important difference between low– and high–T-cell transfer numbers during diabetes pathogenesis. We extended the results from the previous study by demonstrating that PD-1 blockade caused quiescent cells to reenter cycle during a later and chronic stage of disease. During this time, Ki67⁺ BDC2.5 cells increased in the pancreas; however, it is unclear whether this increase correlated with trafficking or proliferation within the pancreas. It is interesting to speculate that islet-reactive CD4⁺ T cells could enter cell cycle in the pancreas, since T cells persist in this organ in NOD mice for weeks to months prior to T1D onset. Additionally, we showed that loss of PD-1 signaling correlated with increased expression of CXCR3. CXCR3 is not only involved in diabetogenic T cell homing to the pancreas but also optimizes Th1 differentiation in antigen-containing LNs (37). These data support a model where PD-1 expressed by the islet-reactive CD4⁺ T cell suppresses proliferation and limits CXCR3 expression, thereby inhibiting optimal Th1 differentiation and infiltration of the pancreas.

Lastly, we used a genetic approach to determine whether the effects that we measured with anti–PD-L1 were due to a direct effect of PD-1 or PD-L1 on the CD4⁺ T cells or indirect effects such as alternative ligand binding (e.g., PD-1/PD-L2 or B7-1/PD-L1). Our data highlight the importance of PD-1, but not PD-L1, acting directly on the CD4⁺ T cell and suggest that the PD-L1/B7-1 (41,42) interaction plays little or no role in regulating these cells. The lack of expansion with PD-L1-KO.BDC2.5 cells further excludes the PD-L1/B7-1 pathway in this model. While the PD-1-KO result underscores the importance of PD-1 on the CD4⁺ T cell in regulating T1D, there are important differences between anti–PD-L1 administration and selective loss of PD-1 on BDC2.5 cells. Anti–PD-L1 induced rapid onset of diabetes, consistent with reports in intact NOD mice (18). Paradoxically, PD-1-KO.BDC2.5 cells did not accelerate T1D in NOD mice despite the overwhelming number of cells in the pancreas and direct invasion of the islet early posttransfer. The lack of T1D observed with PD-1-KO.BDC2.5 cell transfers differs from previous work using PD-1-KO OT-I T-cell transfers into RIP-mOVA mice (38). We speculate that this discordance reflects differences in CD4⁺ and CD8⁺ T cells during T1D. Since CD8⁺ T cells are responsible for direct killing of β-cells, it stands to reason that PD-1-KO diabetogenic CD8⁺ T cells could have a more direct pathogenic effect on disease than PD-1-KO diabetogenic CD4⁺ T cells. Our histological data suggest that anti–PD-L1 is impacting other cells within islets in addition to BDC2.5 cells. Additional work is needed to define the role of anti–PD-L1 on the dynamic interplay between diabetogenic CD4⁺ and CD8⁺ T cells, dendritic cells, and regulatory T cells in the pancreas. By understanding how the PD-1 pathway functions during normal T1D development, we will strengthen our ability to target this pathway in humans.

The studies were supported by the JDRF (2-2011-662 to B.T.F.), the American Diabetes Association (ADA 07-09-JF-21 to B.T.F.), and the University of Minnesota Graduate School Fellowship (to K.E.P.).

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

K.E.P. designed and conceptualized the research project, did the experiments, analyzed data, prepared the figures, interpreted data, and wrote the manuscript. M.K.J. designed and conceptualized the research project, contributed to discussion, and reviewed and edited the manuscript. M.A. provided anti–PD-L1 hybridoma. B.T.F. designed and conceptualized the research project, did the experiments, analyzed data, prepared the figures, interpreted data, and wrote the manuscript. B.T.F. 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.

The authors thank Dr. Bruce Blazar, Department of Pediatrics, University of Minnesota; Dr. David Masopust and Dr. Vaiva Vezys, Department of Microbiology, University of Minnesota; Dr. Kristin Hogquist, Department of Laboratory Medicine and Pathology, University of Minnesota; and Dr. Todd Eagar, Department of Pathology, Immunology, University of Texas-Southwestern for critical review of the manuscript and scientific discussions. The authors thank T. Obu, J. Mader, and N. Sahli from the Department of Medicine, University of Minnesota, for animal care and husbandry.