Deficiency in B7-H1 (PD-L1)/PD-1 Coinhibition Triggers Pancreatic β-Cell Destruction by Insulin-Specific, Murine CD8 T-Cells

H-2^b C57BL/6 (B6) mice were obtained from Janvier (Le Genets-St-Isle; France). B6.SJL-Ptprc^aPep3^b/BoyJ (CD45.1⁺ B6) mice (Jackson #002014), B7-H1^−/− mice (25), PD-1^−/− mice (26), and RIP-B7.1 mice (11) were bred and kept under standard pathogen-free conditions in the animal colony of Ulm University (Ulm, Germany). All studies were conducted after Institutional Board approval in accordance with the Federal German Animal Protection Law.

Mice were immunized intramuscularly into the tibialis anterior muscle or injected intravenously with 3 × 10⁶ magnetic-activated cell sorting (MACS)-purified splenic CD8 T-cells (cat. no. 130-090-859, Miltenyi Biotech). When indicated, mice were treated with blocking B7-H1 antibody (clone MIH5; cat. no. 16-5,982-85, eBioscience). Diabetes was diagnosed if two consecutive blood glucose values exceeded 250 mg/dl, i.e., 13.8 mmol/l (Disetronic Freestyle, Sulzbach, Germany).

Histology was performed as described previously (12,14).

Pancreata were perfused in situ with collagenase P (cat. no. 11213865001, Roche) dissolved in 1 mg/ml Hank′s balanced salt solution (HBSS), removed, digested again with collagenase P for 8 min at 37°C, and washed twice with cold HBSS supplemented with 10% FCS. Pancreatic cells were purified with Histopaque-1077 (cat. no. 10771, Sigma-Aldrich, Germany) by centrifugation for 15 min at 2,400 rpm.

Pancreatic cells (10⁵/100 μl) were incubated for 14–16 h in UltraCULTURE medium with 5–20 μg/ml of the indicated peptides in the presence of brefeldin A (0.5 μg/ml) (cat. no. 15870; Sigma, Taufkirchen, Germany). Cells were harvested, surface stained with APC-conjugated anti-CD8 antibody (cat no. 17–0081–83, BD Biosciences, Heidelberg, Germany), fixed with 2% paraformaldehyde, resuspended in permeabilization buffer (HBSS, 0.5% BSA, 0.5% saponin, 0.05% sodium azide), and stained with fluorescein isothiocyanate-conjugated anti-IFNγ antibody (cat. no.554411; BD Biosciences, Heidelberg, Germany). Frequencies of IFNγ⁺ CD8 T-cells were determined by flow cytometry (FCM).

A single injection of the pCI/preproinsulin DNA encoding murine preproinsulin efficiently induced hyperglycemia in RIP-B7.1 mice, but not in wild-type B6 mice (Fig. 1,A) (13,14). PCI/preproinsulin-immunized RIP-B7.1 mice with early EAD (blood glucose levels of 300–440 mg/dl) were treated 3 weeks after immunization with anti-CD8 antibody (Fig. 1,B). This antibody treatment efficiently depleted CD8 T-cells within 2–3 days and transiently cured EAD (Fig. 1,B). CD8 T-cell levels were restored 3–4 weeks after the anti-CD8 antibody treatment was discontinued, and EAD reappeared concomitant with the re-emerging CD8 T-cells (Fig. 1 B). Anti-CD4 antibody treatment did not inhibit diabetes progression (data not shown). The effector phase of EAD thus depends on diabetogenic CD8 T-cells.

EAD development in preproinsulin-immunized RIP-B7.1 mice was accompanied by increasing infiltrations of CD8 T-cells into the pancreatic target tissue (13,14). CD8 T-cells isolated from immunized, diabetic RIP-B7.1 mice specifically recognized the immunodominant K^b-restricted A_12–21 (i.e., preproinsulin_101–110) epitope of preproinsulin (Fig. 1,C) (13,27). Ex vivo stimulation of preproinsulin-primed, pancreas-infiltrating CD8 T-cells with the antigenic A_12–21 peptide, but not with all other peptides of a preproinsulin-specific library, revealed a CD8 T-cell population with specifically inducible IFNγ expression (Fig. 1 C). In the course of EAD, we also found a significant influx of other lymphoid cells (e.g., CD4 T-cells and B cells) of unknown specificities into the pancreata (data not shown) (14). It is under investigation whether these bystander cells contribute to the diabetogenic immune response (5,28).

We usually detected 4–8 × 10³ CD8 T-cells per pancreas in preproinsulin-immunized, diabetic RIP-B7.1 mice (with blood glucose levels of >400 mg/dl). The frequency of A_12–21-specific IFNγ⁺ CD8 T-cells in these pancreatic CD8 T-cell populations was low (0.5–1.5%) (Fig. 1,C). We identified an epitope variant (A_12-N21A) with an alanine (A) exchange for the COOH-terminal asparagine (N) at position A_21. This variant facilitated in vitro detection of primed CD8 T-cells (14). RIP-B7.1 mice immunized with either pCI/preproinsulin or the pCI/preproinsulin_N110A variant (encoding preproinsulin with the mutant A_12-N21A epitope) developed a similar EAD (data not shown). Pancreas-infiltrating IFNγ⁺ CD8 T-cells expanded in vitro more efficiently after stimulation with the mutant A_12-N21A than the A_12–21 peptide (Fig. 1 D). Similarly, a monospecific A_12–21-encoding vector induced EAD in RIP-B7.1 mice (14), but pancreas-infiltrating CD8 T-cells preferentially expanded in vitro with the mutant A_12-N21A peptide (data not shown). Hence, the alanine at position A₂₁ of the mutant epitope apparently modulates the specific steric structure of the peptide (29) and improves its major histocompatibility complex class I-presentation properties in vitro. We used the variant epitope to detect preproinsulin-specific CD8 T-cell responses.

Preproinsulin-specific immunization efficiently induced EAD in RIP-B7.1 mice, but not in B6 mice (Fig. 1,A). However, low numbers of CD8 T-cells were reproducibly found in the periphery of some (<2%) islets in immunized B6 mice (Fig. 2,A). These CD8 T-cell populations were not found either in nonimmunized B6 (Fig. 2 A) or nonimmunized RIP-B7.1 mice (13,14), suggesting that immunization had induced preproinsulin-specific CD8 T-cells in B6 mice.

We used adoptive transfer experiments to test whether functional preproinsulin-specific CD8 T-cells are primed in B6 mice. B6 mice were immunized with pCI/preproinsulin_N110A. Their splenic CD8 T-cells isolated 14 days after immunization were adoptively transferred into RIP-B7.1 mice (30). RIP-B7.1 hosts developed EAD after transfer with primed (Fig. 2,B, group 2), but not with nonprimed (Fig. 2,B, group 1), CD8 T-cells. Preproinsulin-specific CD8 T-cells accumulated in the pancreata of transplanted and diabetic hosts (Fig. 2,C, group 2). IFNγ is critical for inducing EAD in RIP-B7.1 hosts because adoptively transferred CD8 T-cells from immunized IFNγ^−/− mice did not induce EAD (Fig. 2 B, group 3). Similarly, immunization of RIP-B7.1⁺/IFNγ^−/− mice with preproinsulin also did not induce EAD (13).

We tested whether EAD can be triggered in preproinsulin-immune B6 mice by blocking B7-H1-mediated coinhibition. B6 mice injected with either pCI (Fig. 3,A) or pCI/preproinsulin_N110A (Fig. 3,B and C) were treated at days 12 and 15 after immunization, either with blocking B7-H1 antibody (31) (Fig. 3,A and B) or an isotype control antibody (Fig. 3,C). PCI/preproinsulin_N110A-immunized B6 mice rapidly developed hyperglycemia after injection of the anti-B7-H1, but not control antibody (Fig. 3,B and C). Control mice (injected with pCI and treated with anti-B7-H1 antibody) did not develop EAD (Fig. 3 A).

Within 6–9 days after anti-B7-H1 antibody injection, mice developed either severe hyperglycemia (Fig. 3,B, group 1) or moderate and transient hyperglycemia (Fig. 3,B, groups 2a/b). Severe disease correlated with an influx of high levels of A_12-N21A-specific IFNγ⁺ CD8 T-cells into the islets and an almost complete loss of insulin-producing islet β-cells (Fig. 3,D and E; group 1). In mice with transient diabetes, only low numbers of CD8 T-cells were detectable in the islets during the hyperglycemic stage, and insulin expression was reduced, but still intact (Fig. 3,D and E; group 2a). After recovery to normoglycemia, CD8 T-cells were no longer detectable in the islets (Fig. 3 D and E; group 2b). Pancreatic β-cells that lose B7-H1 coinhibition are hence susceptible (at least transiently) to attack by preproinsulin-specific CD8 T-cells.

We used B7-H1^−/− mice (25) to confirm that EAD induction by preproinsulin depends on B7-H1. A single injection of pCI/preproinsulin_N110A-induced hyperglycemia in B7-H1^−/− mice. EAD developed in both male and female B7-H1^−/− mice with a median onset of 3–5 weeks after immunization and a cumulative diabetes incidence of 80% by week 5 (Fig. 4,A, data not shown). B7-H1^−/− mice did not develop spontaneous EAD after injection with the noncoding pCI vector (Fig. 4,A). Using the preproinsulin-specific peptide library described above (see Fig. 1 C), we detected only A_12–21/A_12-N21A-specific IFNγ⁺ CD8 T-cells in the pancreata of immunized and diabetic B7-H1^−/− mice (data not shown). Thus, CD8 T-cells with this specificity play a prominent role in the destructive autoimmune response in these mice.

We used adoptive cell transfers to exclude that B7-H1-deficiency in CD8 T-cells is critical for EAD development in B7-H1^−/− mice. RIP-B7.1 and B7-H1^−/− mice were immunized with pCI/preproinsulin_N110A. We adoptively transferred 3 × 10⁶ CD8 T-cells (derived from spleens of early diabetic mice) into RIP-B7.1 hosts (Fig. 4,B). RIP-B7.1 hosts reconstituted with primed, RIP-B7.1- or B7-H1^−/−-derived CD8 T-cells induced EAD with a similar efficacy (Fig. 4,B). Transfer of nonprimed CD8 T-cells from healthy RIP-B7.1 or B7-H1^−/− donors did not induce EAD (data not shown). Thus, B7-H1-deficiency in preproinsulin-specific CD8 T-cells did not alter their diabetogenic potential. In contrast, transfer of identical RIP-B7.1-derived CD8 T-cell preparations into B7-H1^−/− hosts inefficiently induced late EAD (Fig. 4,C). Only 1 of 8 B7-H1^−/− hosts developed EAD at 18 weeks after transfer (Fig. 4,C). The efficacy of EAD could be increased in adoptively transferred B7-H1-deficient hosts if the stimulatory B7.1 molecule is coexpressed in pancreatic β-cells of RIP-B7.1⁺/B7-H1^−/− mice (Fig. 4 C).

We used PD-1^−/− knockout mice (26) to test whether induction of preproinsulin-specific EAD depends on the coinhibitory PD-1/B7-H1 interaction (19). A single injection of the pCI/preproinsulin_N110A (but not of pCI vector) into PD-1^−/− mice induced A_12-N21A-specific IFNγ⁺ CD8 T-cell responses and EAD (Fig. 4,D; data not shown), suggesting that B7-H1/PD-1 coinhibition is critical to induce EAD in preproinsulin-immune mice. To confirm this, we generated bone marrow chimeras. To distinguish between donor- and host-derived T-cells in bone marrow (BM) chimeras, we used wild-type CD45.1⁺ B6 mice in this set of experiments. BM cells from CD45.1⁺ donor mice (B7-H1⁺ PD-1⁺) were transferred into lethally irradiated, congenic B7-H1^−/− or PD-1^−/− (CD45.2⁺), and BM cells from B7-H1^−/− or PD-1^−/− donor mice were transferred into wild-type CD45.1⁺ hosts (Table 1). At 6 to 7 weeks after transplantation, chimeric mice contained >90% of donor T-cells (data not shown). Chimeric mice were immunized 7 weeks after transplantation with pCI/preproinsulin_N110A (or control pCI). As is evident from Table 1, wild-type/B7-H1^−/− (group 1) and PD-1^−/−/wild-type chimeras (group 7), but not wild-type /PD-1^−/− (group 3) and B7-H1^−/−/wild-type chimeras (group 5), developed EAD after immunization with pCI/preproinsulin_N110A. EAD manifestation in groups 1 and 7 correlated with an influx of A_12-N21A-specific IFNγ⁺ CD8 T-cells into the target tissue (data not shown). Hence, either the selective deficiency of B7-H1 on target cells (group 1) or the deficiency of PD-1 on T-cells (group 7) triggered CD8 T-cell–mediated EAD. Binding of PD-1 expressed by activated CD8 T-cells to B7-H1 expressed by pancreatic β-cells thus seems to control the diabetogenic immune response.

There is increasing evidence from patients with type 1 diabetes that autoreactive CD8 T-cells specific for preproinsulin are involved in β-cell destruction (32,,,,,–38). We established mouse models to study the pathogenic cross-talk between preproinsulin-specific CD8 T-cells and preproinsulin-expressing β-cells. RIP-B7.1 tg and two well defined (B7-H.1^−/−, PD-1^−/−) mouse lines allowed us to identify critical checkpoints for the control of diabetogenic, preproinsulin-specific CD8 T-cells: 1) the costimulator molecule B7.1 (CD80) expressed on β-cells (RIP-B7.1 mice); 2) the coinhibitor B7-H1 (PD-L1) expressed by β-cells (B7-H1^−/− mice); or 3) the coinhibitor PD-1 molecule expressed by CD8 T effector cells (PD-1^−/− mice).

Transgene-driven B7.1 (CD80) expression in the pancreatic β-cells of RIP-B7.1 mice makes them susceptible to autoreactive CD8 T-cell attack. The nonphysiologic, costimulatory B7.1/CD28 interaction in RIP-B7.1 mice (9,10) may allow efficient effector function delivery by autoreactive CD8 T-cells. CD28-deficient RIP-B7.1⁺/CD28^−/− mice did not develop EAD after immunization with preproinsulin (data not shown). There is thus strong evidence that the interaction of B7.1 on the surface of β-cells with the costimulator molecule CD28 on T-cells is an essential component of T-cell–mediated EAD in RIP-B7.1 mice. Preproinsulin-specific immunization induced EAD in almost all RIP-B7.1 mice with a strikingly similar time course and histopathology. Expression of B7.1 in islet β-cells facilitated diabetes development by adoptively transferred preproinsulin-specific CD8 T-cells (Fig. 4 C) (30). The RIP-B7.1 diabetes model is thus well suited to study distinct events in the priming and effector phase of preproinsulin-specific CD8 T-cells. For example, we previously showed that expression and processing of preproinsulin antigens in the endoplasmic reticulum favor priming of autoreactive CD8 T-cells (14). Thus, direct loading of the A_12–21 epitope on newly synthesized K^b-molecules in the endoplasmic reticulum may be an essential step for enabling presentation of this epitope (14). We currently establish a HLA-A*0201⁺/RIP-B7.1 mouse model to define the diabetogenic potential of HLA-A*0201-restricted preproinsulin epitopes associated with human type 1 diabetes (35,–37). Furthermore, RIP-B7.1 mice are a useful tool to identify novel β-cell antigens that are targets for CD8 T-cell–triggered diabetes (4).

We consider the key observation of this report to be the de novo induction of preproinsulin-specific CD8 T-cells (and EAD) in mice and their control by B7-H1/PD-1 interaction. It has been shown that PD-1 and its ligands, B7-H1 (PD-L1) and PD-L2, deliver inhibitory signals that regulate the balance between T-cell activation, tolerance, and immunopathology (19,39). B7-H1 (PD-L1) is expressed on antigen-presenting cells as well as effector CD8 T-cells, whereas PD-1 is expressed on T-cells. Using bone marrow chimeric mice, we confirmed that the deficiency of B7-H1 on target cells or PD-1 on T-cells was essential to trigger preproinsulin-specific, CD8 T-cell–mediated EAD by DNA-based immunization (Table 1). Binding of PD-1 on activated T-cells to B7-H1 expressed by pancreatic β-cells may hence downmodulate the diabetogenic potential of preproinsulin-specific CD8 T-cells. We further showed that B7-H1-deficiency in preproinsulin-specific CD8 T-cells did not alter their diabetogenic potential (Fig. 4,B). B7-H1^−/− mice inefficiently developed EAD after adoptive transfer of preproinsulin-specific CD8 T-cells, but coexpression of the stimulatory B7.1 molecule in B7-H1-deficient pancreatic β-cells (RIP-B7.1⁺/B7-H1^−/− mice) significantly accelerated disease induction (Fig. 4,C). Transgene-driven B7.1 costimulation in pancreatic β-cells was thus more potent than the loss of B7-H1 coinhibition in promoting the pathogenic immune response of adoptively transferred preproinsulin-specific CD8 T-cells. In contrast, immunization of B7-H1^−/− and RIP-B7.1 mice with preproinsulin-encoding DNA induced EAD with similar efficacies and kinetics (see Figs. 1,A, 4 A). We assume that additional factors (e.g., professional antigen-presenting dendritic cells or cells from the innate immune system) (40) are triggered by DNA-based immunization that facilitate expansion of preproinsulin-specific CD8 T-cells and/or maintain their diabetogenic potential in B7-H1^−/− mice.

The B6 diabetes model described in this study is attractive to characterize distinct events in the regulation of β-cell susceptibility to manifest or control preproinsulin-specific, CD8 T-cell–mediated EAD. B6 is not a privileged strain for type 1 diabetes studies, but it was unexpectedly easy to prime preproinsulin-specific CD8 T-cells in male and female B6 mice. CD8 T-cells revealed their diabetogenic potential after adoptive transfer into congenic RIP-B7.1 hosts or after conditioning the pancreatic target tissue by antibody-mediated blockage of B7-H1 coinhibiton (Figs. 2 and 3). Preproinsulin-specific CD8 T-cells in immunized B6 mice thus have a diabetogenic potential, but pancreatic β-cells are protected from immune attack by these cells. Similarly, in the lymphocytic choriomeningitis virus diabetes model, the pancreatic target tissue must be exposed to stimulatory signals from the innate immune system to become susceptible to the destructive CD8 T-cell attack (8). The B6 model is a good example for translational medicine since our data illustrated the central role of the β-cell as a gatekeeper in the preproinsulin-specific insulitic process. Beta cells per se prevent the deleterious cross-talk with preproinsulin-specific CD8 T-cells. Changes in the β-cell milieu—e.g., by antibody treatment (Fig. 3), by interferon (8), or by viral infections (41)—can favor the susceptibiliy of β-cells for the CD8 T-cell–mediated immune attack. Further manipulations of the pancreatic β-cells or distinct arms of the immune system by specific drugs (42) or by using different mouse strains with defects in specific cell types or immune-mediators may define conditions that inactivate (tolerize/anergize) autoreactive CD8 T-cells. “Translation” of these approaches to human type 1 diabetes (3,41,–43) could be helpful to design prophylactic vaccines.

This work was supported by grants from the Deutsche Forschungsgemeinschaft: SFB 518/A01 and GRK-1041 (to R.S. and B.O.B.) and DFG SCHI-505/4-1 to R.S.

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

T.R., H.B., C.S., and A.S. researched data and contributed to discussion. B.O.B. contributed to discussion and reviewed/edited the manuscript. L.C. reviewed/edited the manuscript. J.R. contributed to discussion and wrote the manuscript. R.S. researched data and wrote the manuscript.

The authors thank Claudia Heilig, Katrin Ölberger, and Andrea Wissmann (Department of Internal Medicine I, University Hospital of Ulm, Ulm, Germany) for outstanding technical assistance, and Dr. Tasuku Honjo (Department of Immunology and Genomic Medicine, Kyoto University, Graduate School of Medicine, Kyoto, Japan) for PD-1^−/− mice.