A Missing PD-L1/PD-1 Coinhibition Regulates Diabetes Induction by Preproinsulin-Specific CD8 T-Cells in an Epitope-Specific Manner

Coinhibitory PD-1/PD-L1 (B7-H1) interactions provide critical signals for the regulation of autoreactive T-cell responses. We established mouse models, expressing the costimulator molecule B7.1 (CD80) on pancreatic beta cells (RIP-B7.1 tg mice) or are deficient in coinhibitory PD-L1 or PD-1 molecules (PD-L1−/− and PD-1−/− mice), to study induction of preproinsulin (ppins)-specific CD8 T-cell responses and experimental autoimmune diabetes (EAD) by DNA-based immunization. RIP-B7.1 tg mice allowed us to identify two CD8 T-cell specificities: pCI/ppins DNA exclusively induced Kb/A12–21-specific CD8 T-cells and EAD, whereas pCI/ppinsΔA12–21 DNA (encoding ppins without the COOH-terminal A12–21 epitope) elicited Kb/B22–29-specific CD8 T-cells and EAD. Specific expression/processing of mutant ppinsΔA12–21 (but not ppins) in non-beta cells, targeted by intramuscular DNA-injection, thus facilitated induction of Kb/B22–29-specific CD8 T-cells. The A12–21 epitope binds Kb molecules with a very low avidity as compared with B22–29. Interestingly, immunization of coinhibition-deficient PD-L1−/− or PD-1−/− mice with pCI/ppins induced Kb/A12–21-monospecific CD8 T-cells and EAD but injections with pCI/ppinsΔA12–21 did neither recruit Kb/B22–29-specific CD8 T-cells into the pancreatic target tissue nor induce EAD. PpinsΔA12–21/(Kb/B22–29)-mediated EAD was efficiently restored in RIP-B7.1+/PD-L1−/− mice, differing from PD-L1−/− mice only in the tg B7.1 expression in beta cells. Alternatively, an ongoing beta cell destruction and tissue inflammation, initiated by ppins/(Kb/A12–21)-specific CD8 T-cells in pCI/ppins+pCI/ppinsΔA12–21 co-immunized PD-L1−/− mice, facilitated the expansion of ppinsΔA12–21/(Kb/B22–29)-specific CD8 T-cells. CD8 T-cells specific for the high-affinity Kb/B22–29- (but not the low-affinity Kb/A12–21)-epitope thus require stimulatory ´help from beta cells or inflamed islets to expand in PD-L1-deficient mice. The new PD-1/PD-L1 diabetes models may be valuable tools to study under well controlled experimental conditions distinct hierarchies of autoreactive CD8 T-cell responses, which trigger the initial steps of beta cell destruction or emerge during the pathogenic progression of EAD.


Introduction
Type 1 diabetes (T1D) is an autoimmune disorder, in which insulin-producing beta cells are destroyed by the cellular immune system [1,2,3]. Diabetes development is characterized by progressive infiltration of T-cells into the pancreatic islets and beta cell destruction, resulting in severe hyperglycemia. Disease in man is triggered by poorly defined antigens and factors that finally result in the breakdown of central and/or peripheral tolerance and activation of autoreactive CD4 + and/or CD8 + T-cells [1,4]. There is increasing evidence from patients with T1D that autoreactive CD8 + T-cells are involved in the development of disease but it is difficult to detect these rare lymphocytes and to assign their individual effects during the progression of diabetes [5,6,7]. It is assumed that the nature of an autoantigen-derived peptide and its presentation by MHC class I molecules plays a central role in the development of T-cell-mediated autoimmunity [8]. In the NOD mouse model [9], the binding of insulin-derived self peptides to MHC class I or class II molecules is weak and caused by unfavoured binding registers [10,11,12]. This suggests that nonconventional antigenic epitope processing and presentation may contribute to the induction of autoreactive immune responses [7,13].
Spontaneous diabetes development in the NOD mouse model elucidated many aspects of diabetogenic immune responses [9]. Furthermore, different mouse models have been used to characterize de novo induction of well-defined T-cell responses and their pathogenic cross-talk with beta cells, which selectively express transgene-encoded 'neo-self' antigens under rat insulin promoter (RIP) control [14]. We used transgenic (tg) RIP-B7.1 mice, expressing the costimulatory molecule B7.1 (CD80) on pancreatic beta cells [15], to characterize induction of preproinsulin (ppins)specific CD8 T-cells and experimental autoimmune diabetes (EAD) by DNA-based immunization [16,17,18,19]. A single injection of ppins-encoding DNA (pCI/ppins) efficiently induced CD8 T-cell-mediated EAD in both, male and female RIP-B7.1 tg mice with a median onset of 2-3 weeks post immunization and a cumulative diabetes incidence of .95% by week 4 [17]. In these mice, progressive invasion of insulin A-chain-derived K b /A 12-21specific CD8 T-cells into pancreatic islets precedes hyperglycemia and insulin deficiency. K b /A 12-21 -specific CD8 T-cells and EAD were efficiently induced by pCI/ppins in MHC class II-deficient (Aa 2/2 ) RIP-B7.1 mice (RIP-B7.1 + /MHC-II 2/2 ) with no conventional CD4 T-cells and in RIP-B7.1 tg mice acutely depleted of CD4 T-cells with anti CD4 antibody [17,18]. The RIP-B7.1 tg model hence provides an attractive experimental approach to study CD4 T-cell-independent induction of EAD by ppins-specific CD8 T-cells.
During the course of EAD in RIP-B7.1 tg mice ex vivo stimulation of ppins-primed CD8 T-cells with the K b /A 12-21 peptide, but not with all other peptides of a ppins-specific library, revealed a CD8 T-cell population with specifically inducible IFNc expression [19]. This suggested that the K b /A 12-21 is the only diabetogenic epitope in ppins-immune RIP-B7.1 tg mice. However, a mutant ppinsDA 12-21 antigen (with a deletion of the COOH-terminal A 12-21 sequence) also induced CD8 T-cellmediated EAD in RIP-B7.1 tg mice, indicating that EAD can be induced by CD8 T-cell responses that have specificities other than K b /A 12-21 [18]. In this study, we mapped the alternative CD8 Tcell epitope in pCI/ppinsDA 12-21 -immune RIP-B7.1 tg mice and investigated the antigen expression and processing requirements to prime this CD8 T-cell specificity and EAD. We further used coinhibiton-deficient PD-1 2/2 , PD-L1 2/2 and RIP-B7.1 + /PD-L1 2/2 mice (differing from PD-L1 2/2 mice only in the tg B7.1 expression in beta cells) to determine whether induction of ppinsand ppinsDA 12-21 -specific CD8 T-cell responses and diabetes development depends on specific costimulatory and coinhibitory signals from pancreatic beta cells.

Construction of Expression Plasmids
The sequences of the different ppins antigens were codonoptimized and synthesized by GeneArt (Regensburg, Germany). All constructs were cloned into the pCI vector (cat.no. E1731, Promega, Mannheim, Germany) using the NheI and NotI restriction sites. Batches of plasmid DNA were produced in E. coli by PlasmidFactory GmbH (Bielefeld, Germany).

Characterization of Antigen Expression
Human embryonal kidney cells (HEK-293 cells) were transiently transfected with the indicated plasmid DNAs using the calcium phosphate method. Cells were labeled with 100 mCi 35 S-methionine/cysteine (cat. no. IS103, Hartmann Analytic GmbH) between 36 and 48 h post transfection and subsequently lysed with pH 8.0 lysis buffer (100 mM NaCl, 0.5% NP40 and 100 mM Tris-hydrochloride) supplemented with the protease inhibitors, leupeptin and aprotinin. Extracts were cleared by centrifugation and precipitated with polyclonal rabbit H-86 anti-insulin (cat. no. sc-9168, Santa Cruz Biotechnology) and protein G sepharose. Precipitates were processed for SDS-PAGE (15%) and subsequent fluorography of the gels. Alternatively, non-labeled cells were lysed in a SDS-containing buffer (3% SDS, 50 mM Tris-hydrochloride, 5% b-mercaptoethanol) and, for high resolution of protein bands, samples were directly loaded onto urea-containing SDS-polyacrylamide gels (16%) [32]. Gels were blotted onto a nitrocellulose membrane using the iBlotH Dry Blotting System (cat. no. IB3010-01; Invitrogen, Carlsbad, CA, USA). Membranes were blocked for 20 min at RT in a buffer supplemented with 0.1% Tween 20, 0.1% gelatine and 3% milk powder, followed by successive incubations with rabbit H-86 anti-insulin antibody and HRPconjugated anti rabbit IgG (cat. No. NA9340; GE Healthcare, Chalfont St Giles, UK). Specific protein bands were detected using the Immobilon TM Western Chemoluminescent HRP substrate (cat. No. WBKLS0100; Millipore, Bedford, MA, USA) followed by subsequent exposure of the membranes to an Amersham Hyperfilm ECL (cat. No. 92004; GE Healthcare).

Isolation of CD8 T-cells from Pancreatic Tissues
Pancreata were perfused in situ with collagenase P (cat. no. 11213865001, Roche) dissolved at 1 mg/ml in Hanks Balanced Salt Solution (HBSS), removed, digested again with collagenase P for 8 min at 37uC and washed twice with cold HBSS supplemented with 10% FCS. Pancreatic cells were purified with Histopaque-1077 (cat. no. 10771, Sigma-Aldrich) by centrifugation for 15 min at 2400 rpm.
Furthermore, specific CD8 T-cells were analyzed with K b /B 22-29 tetramers (Glycotope, Heidelberg, Germany). Freshly isolated cells were washed twice in PBS/0.3% w/v BSA/0.1% w/v sodium azide. Non-specific binding of antibodies to Fc-receptor was blocked by preincubating cells with mAb 2.4G2 as described above. Cells were incubated for 30 min at 4uC with FITC-labeled anti-CD8 mAb (BD Biosciences, Heidelberg, Germany) and PEor APC-conjugated tetramers. Cells were washed and analyzed by FCM.

Statistics
The statistical significance of differences in the mean CD8 Tcell frequencies between groups was determined by the unpaired student's t-test. The statistical significance of diabetes induction in immunized mice was determined by the log-rank test. Data were analyzed using PRISM software (GraphPad, San Diego, CA). Values of P,0.05 were considered significant.

Characterization of Antigen Expression Requirements that Favour Priming of K b /B 22-29 -Specific CD8 T-cells and EAD by DNA-based Immunization
The efficient induction of K b /B 22-29 -specific CD8 T-cells and EAD by mutant ppinsDA [12][13][14][15][16][17][18][19][20][21] was unexpected because immunization with different insulin B-chain-encoding vectors did not (or very inefficiently) induce EAD in RIP-B7.1 tg mice. A pCI/SP-B construct (encoding the ER-targeting signal peptide and the insulin B-chain; Figure 3A), inefficiently induced late EAD in one out of eight RIP-B7.1 tg mice ( Figure 3B, group 2). Similarly, a pCI/SP-B-C construct (encoding the ER-targeting signal peptide up to the C-peptide; Figure 3A) did not induce EAD in RIP-B7.1 tg mice within three months post immunization ( Figure 3B, group 3). We thus conclude that efficient priming of K b /B 22-29 -reactive CD8 Tcells by pCI/ppinsDA 12-21 critically depends on specific properties of the mutant antigen itself.
We here identified an alternative mechanism to promote the expansion and influx of diabetogenic K b /B 22-29 -specific CD8 Tcells into the pancreata of pCI/ppinsDA 12-21 -primed and diabetic PD-L1 2/2 mice. Co-immunization of PD-L1 2/2 mice with both, pCI/ppins+pCI/ppinsDA 12-21 vectors (but not with the individual pCI/ppins or pCI/ppinsDA 12-21 vectors) efficiently elicited both, K b /A 12-21 -and K b /B 22-29 -specific CD8 T-cells (Table S1). This suggested that the initial beta cell destruction, triggered by pCI/ ppins/(K b /A 12-21 )-specific CD8 T-cells in PD-L1 2/2 mice, facilitates expansion and invasion of K b /B 22-29 -specific CD8 Tcells but also other bystander cells into the pancreatic target tissue. The specific molecular mechanisms and signals expanding and attracting K b /B 22-29 -specific CD8 T-cells to the pancreas and the role of bystander cells are not well understood [38,39,40]. An initial beta cell death and antigen release could facilitate activation of circulating, pCI/ppinsDA 12-21 -preprimed K b /B 22-29 -specific CD8 T-cells in the regional lymph nodes by professional APCs [42]. Furthermore, an altered local cytokine milieu and expression of cell surface receptors [40] or an enhanced antigen presentation by beta cells [43] in inflamed islets may favour the attraction and/ or activation of K b /B 22-29 -specific CD8 T-cells. Taken together, our findings suggested that the K b /A 12-21 -specific CD8 T-cell response directly initiates beta cell destruction in PD-L1 2/2 mice, whereas a downstream K b /B 22-29 -specific CD8 T-cell response requires additional activation signals in vivo and emerge during the pathogenic destruction of beta cells. Interestingly, distinct hierarchies of diabetogenic T-cell responses were also detectable in the NOD mouse model. The insulin B9-23 domain, containing both, a dominant CD4 and a K d -restricted B [15][16][17][18][19][20][21][22][23] CD8 T-cell epitope, plays a prominent role in the diabetes development in NOD mice Figure 5. Recruitment of different 'bystander' cell populations into the pancreatic target tissue. PD-L1 2/2 mice were immunized with both, pCI/ppins+pCI/ppinsDA 12-21 vectors into the right and the left tibialis anterior muscles, respectively. Pancreata of representative healthy (at 3 days post immunization) (A) or early diabetic mice (at 15-20 days post immunization) (B) were analyzed histologically for insulin expression (insulin) and influx of CD4 + T-cells (CD4 + ), macrophages (F4/80 + ) or DCs (CD11c + ). doi:10.1371/journal.pone.0071746.g005 [10,44,45]. Prasad et al. previously showed that an initial insulin B9-23-specific T-cell response is immunodominant and autoimmune responses to epitope(s) distinct from B9-23 emerge during the pathogenic progression of diabetes in NOD mice [46]. Similarly, CD8 T-cells specific for the islet-specific glucose-6phosphatase catalytic subunit-related protein (IGRP) were detected in NOD mice but not in tg NOD mice tolerant to proinsulin, indicating that an initial T-cell response against proinsulin is necessary for the development of IGRP-specific CD8 T-cells [47]. This suggests that distinct CD8 T-cell responses, triggering the initial steps of beta cell destruction, play a prime role in the induction of diabetes [44].