PD-1 deficiency reveals various tissue-specific autoimmunity by H-2^b and dose-dependent requirement of H-2^g7 for diabetes in NOD mice

Abstract

Although many autoimmune diseases are associated with particular HLA/H-2 haplotypes, the mechanisms through which specific HLA/H-2 haplotypes afford autoimmune susceptibility remain enigmatic. Here, we analyzed the effects of the diabetes-associated (H-2^g7) and an antidiabetogenic (H-2^b) H-2 haplotypes in NOD mice deficient for programmed cell death-1 (PD-1, Pdcd1), a unique model of type 1 diabetes that confers complete penetrance and rapid onset of the disease. NOD-H2^b/bPdcd1^−/− mice were completely protected from diabetes, confirming that H-2^g7 is indispensable for diabetes even in the absence of PD-1. However, NOD-H2^b/bPdcd1^−/− mice developed autoimmune inflammation in multiple tissues including peripheral nerves, stomachs, and exocrine tissues, demonstrating that autoreactive T cells are generated in the context of H-2^b. These autoreactive T cells damaged target tissues only in the absence of PD-1, confirming that PD-1 deficiency unravels the hidden autoimmune susceptibility of the strain by reducing the threshold of T cell activation. Transfer experiments revealed that CD4 T cells are the effector cells of neuritis, and nerve-infiltrating CD4 T cells are strongly deviated toward Th1. Interestingly, neuritogenic T cells were also generated in the context of H-2^g7, in sharp contrast to the strict requirement of H-2^g7 for diabetes. In addition, 60% of the NOD-H2^b/g7Pdcd1^−/− mice developed diabetes, suggesting that H-2^b does not dominantly suppress diabetes and that H-2^g7 induces diabetes in a dose-dependent fashion.

Association with particular HLA/H-2 haplotypes has been reported for most of the autoimmune diseases, albeit to varying degrees. Nonobese diabetic (NOD) mice, a murine model of type 1 diabetes (T1DM) and Sjogren syndrome, have a unique H-2 haplotype termed H-2^g7 that promotes T1DM in a recessive manner (1, 2). Wicker and colleagues replaced H-2^g7 with H-2^b from C57BL10/SnJ mice and found that NOD-H2^b/b mice are protected from insulitis and T1DM but not from sialoadenitis (3, 4). Studies on other congenic NOD mice expressing non-NOD H-2 haplotypes and transgenic NOD mice overexpressing non-NOD H-2 haplotypes collectively suggest that MHC class II molecules play a direct role in determining the susceptibility of NOD mice to spontaneous T1DM (1, 2). However, the precise mechanism through which MHC class II molecules confer autoimmune susceptibility is still a matter of intense debate. Several groups proposed the idea that H-2^g7 endows autoimmune proclivity by supporting the positive selection of autoreactive T cells in general, which may be explained by the low peptide-binding capacity of H-2^g7 (5–7). On the other hand, other reports suggest that H-2^g7 promotes T1DM specifically by either allowing the selection of diabetogenic T cells in the thymus (8, 9), polarizing diabetogenic T cells toward Th1 (10, 11), or facilitating the clonal deletion of regulatory T cells (Tregs) specific to diabetogenic T cells (12–14).

Programmed cell death-1 (PD-1, Pdcd1) is a negative immunoreceptor belonging to the CD28/Cytotoxic T lymphocyte antigen-4 (CTLA-4) family, and its deficiency causes different types of autoimmune diseases on C57BL/6, BALB/c, and NOD backgrounds (15–17). In particular, PD-1 deficiency accelerates T1DM and sialoadenitis but does not induce other autoimmune diseases in NOD mice (18), indicating that PD-1 deficiency accelerates the inherent autoimmune susceptibility without changing its specificity. Accumulating evidence supports the idea that PD-1 plays an essential role in the establishment and/or the maintenance of T cell clonal anergy and that its deficiency prevents and rescues T cells from anergic state (19, 20). Invigoration of anergic T cells may account for the reason why PD-1 deficiency accelerates the tissue-specific autoimmunity inherent in the strain. It should be noted that no polyclonal activation of lymphocytes is observed in Pdcd1^−/− mice on either background, suggesting that PD-1 deficiency does not distort the general immune response. Taken together, Pdcd1^−/− mice may serve as a useful animal model to verify the effects of other genetic and environmental factors in the development of autoimmune diseases (17, 18, 21, 22).

In the present study, we analyzed the effect of the antidiabetogenic H-2 haplotype (H-2^b) in the accelerated T1DM of NOD-Pdcd1^−/− mice. The complete protection from T1DM in NOD-H2^b/bPdcd1^−/− mice indicated that H-2^g7 is definitely required for the diabetic incidence even in the absence of PD-1. However, NOD-H2^b/bPdcd1^−/− mice developed many different autoimmune diseases including polyneuropathy. Interestingly, neuritogenic T cells were also generated in the context of H-2^g7. In addition, the effect of H-2^g7 on T1DM development on NOD background is dose-dependent, and H-2^b does not dominantly suppress T1DM because 60% of the NOD-H2^b/g7Pdcd1^−/−mice developed T1DM.

Results

Development of Peripheral Polyneuropathy in NOD-H2^b/g7Pdcd1^−/− and NOD-H2^b/bPdcd1^−/− Mice.

We generated NOD-H2^b/bPdcd1^−/− mice by mating NOD-Pdcd1^−/− mice with a NOD congenic line whose H-2^g7-containing chromosomal region [D17Mit100 (11.75 cM) to D17Mit105 (21.75 cM)] is replaced with the corresponding region from C57BL/10SnJ mice (H-2^b). As shown in Fig. 1, none of the NOD-H2^b/bPdcd1^−/− mice developed T1DM by 20 weeks of age, whereas the NOD-H2^g7/g7Pdcd1^−/− mice became diabetic within 8 weeks, indicating that H-2^g7 is indispensable for diabetic incidence, even in the absence of PD-1. Histologically, isulitis was slightly exacerbated by PD-1 deficiency in NOD-H2^b/b mice [supporting information (SI) Fig. 6]. Surprisingly, 60% of the NOD-H2^b/g7Pdcd1^−/− mice developed T1DM by 20 weeks of age with severe insulitis, although none of the NOD-H2^b/g7Pdcd1^+/+ mice developed T1DM in accordance with a previous report (3). These results indicate that H-2^g7 induces T1DM in a dose-dependent fashion, suggesting that H-2^b does not dominantly suppress T1DM.

Fig. 1.

Complete protection of NOD-Pdcd1^−/− mice from T1DM by H-2^b/b but not H-2^b/g7. (a and b) Incidence of T1DM in female (a) and male (b) mice is shown for NOD-H2^g7/g7Pdcd1^+/+ (open circles), NOD-H2^b/g7Pdcd1^+/+ (open triangles), NOD-H2^b/bPdcd1^+/+ (open squares), NOD-H2^g7/g7Pdcd1^−/− (closed circles), NOD-H2^b/g7Pdcd1^−/− (closed triangles), and NOD-H2^b/bPdcd1^−/− (closed squares) mice (n = 11 ≈ 28). (c) Representative H&E staining of islets is shown for female NOD-Pdcd1^+/+ (Top) and NOD-Pdcd1^−/− (Middle) mice with H-2^g7/g7 (Left), H-2^b/g7 (Center), and H-2^b/b (Right) haplotypes. Sections were also stained with Abs against CD4 (red), CD8 (blue), and PD-L1 (green) (Bottom). (Original magnification, ×400.)

Despite the protection from T1DM, female NOD-H2^b/bPdcd1^−/− mice spontaneously developed peripheral polyneuropathy (Fig. 2 a and b and SI Movie 1). Similarly, all of the female NOD-H2^b/g7Pdcd1^−/− mice that escaped from T1DM developed polyneuropathy (Fig. 2d). In contrast, male mice with either genotype did not develop neuropathy by the endpoint of the experiment, when they were 25 weeks old (Fig. 2d). Pathological examination of the neuropathic mice revealed a massive infiltration of lymphocytes in the peripheral nerves and dorsal root ganglia with severe demyelination (Fig. 2c); in contrast, the brain and spinal cord were unaffected (T.Y., T.H., and T.O., unpublished data). These clinical and pathological features resemble the hallmarks of Guillain–Barre syndrome (GBS) and chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) in humans (23). Neither polyneuropathy nor neuritis was observed in PD-1-sufficient mice at the same age (Fig. 2 a, c, and d).

Fig. 2.

Spontaneous development of peripheral polyneuropathy by NOD-H2^b/g7Pdcd1^−/− and NOD-H2^b/bPdcd1^−/− mice. (a) Incidence of polyneuropathy is shown for female NOD-H2^g7/g7Pdcd1^+/+ (open circles), NOD-H2^b/g7Pdcd1^+/+ (open triangles), NOD-H2^b/bPdcd1^+/+ (open squares), NOD-H2^g7/g7Pdcd1^−/− (closed circles), NOD-H2^b/g7Pdcd1^−/− (closed triangles), and NOD-H2^b/bPdcd1^−/− (closed squares) mice (n = 11 ≈ 28). NOD-H2^g7/g7Pdcd1^−/− mice were killed at 10 weeks of age because of the severe T1DM. (b) A neuropathic mouse with severe paralysis of the limbs is shown. (c) Luxol fast blue staining of the sciatic nerve (Left) and H&E stainings of the sciatic nerve (Center) and dorsal root ganglion (Right) are shown. (d) Severity of neuropathy in male (closed circles) and female (open circles) mice is shown for the indicated genotypes. (e) Sciatic nerves from NOD-H2^b/bPdcd1^+/+ (Upper) and NOD-H2^b/bPdcd1^−/− (Lower) mice were stained with Abs against CD4 (green), CD8 (blue), and IgM (red) (Left), PD-L1 (green), CD4 (red), and CD8 (blue) (Center), and PD-L1 (green) and CD11c (red) (Right). (Original magnification: c and e, ×200.)

Immunohistofluorescent staining revealed a massive infiltration of CD4 T cells together with a mild infiltration of CD8 T cells and B cells in the sciatic nerves, suggesting that the extensive demyelination was due to an autoimmune attack by the lymphocytes. Interestingly, PD-L1 was strongly expressed on Schwann cells and on a subset of the CD11c⁺ dendritic cells in sciatic nerves that showed inflammation but not in those without inflammation (Fig. 2e). In contrast, PD-L2 expression was not observed (T.Y., T.H., and T.O., unpublished data). These results suggest that Schwann cells express PD-L1 in an attempt to suppress autoreactive T cells, which could express PD-1, as has been suggested for the pancreatic β cell response to T1DM (18, 24).

In addition to polyneuropathy, the NOD-H2^b/bPdcd1^−/− and NOD-H2^b/g7Pdcd1^−/− mice developed sialoadenitis, pancreatitis, and gastritis (SI Fig. 7 and SI Table 1). These results indicate that NOD-H2^b/b mice can support the generation of the autoreactive T cells against multiple tissues and that these autoreactive T cells can induce overt autoimmune inflammation in the absence of PD-1.

CD4 T Cells Play a Pathogenic Role in the Polyneuropathy of NOD-H2^b/bPdcd1^−/− Mice.

To determine the effector cells of the neuritis in the NOD-H2^b/bPdcd1^−/− mice, we performed adoptive transfer experiments. We isolated CD4 T cells from spleen of neuropathic NOD-H2^b/bPdcd1^−/− mice and transferred into C57BL/6-Rag2^−/− mice, whose H-2 haplotype completely matched that of the donor mice. The adoptive transfer of unfractionated splenocytes and CD4 T cells but not CD4-depleted splenocytes induced severe neuritis coupled with mild contracture in 10 weeks, indicating that CD4 T cells are the effector cells of the neuritis (Fig. 3 a–d).

Fig. 3.

CD4 T cells are the effector cells of the polyneuropathy. (a–c) Representative H&E staining of sciatic nerve from C57BL/6-Rag2^−/− mice that had received transfers of unfractionated splenocytes (a), CD4 T cells (b), or CD4-depleted splenocytes (c) prepared from the spleen of neuropathic NOD-H2^b/bPdcd1^−/− mice. (d) The severity of polyneuropathy is shown for each recipient mouse. (e) B cells are dispensable for the development of polyneuropathy. Representative H&E staining of NOD-H2^b/bPdcd1^−/−μMT mouse is shown. (f–i) NOD-H2^b/bPdcd1^−/− mice produce autoAbs against myelin sheath. (f) Sciatic nerve from NOD-SCID mouse was stained with serum from neuropathic NOD-H2^b/bPdcd1^−/− mouse. (g–i) Higher magnification of sciatic nerve stained with Luxol fast blue (g), serum (green), and anti-neuronal class III β-tubulin Ab (red) (h), or serum (green) and anti-MBP Ab (red) (i). Representative data of more than five mice are shown. (Original magnification: a–f, ×200; g–i, ×630.)

Next, we examined the involvement of B cells by mating NOD-H2^b/bPdcd1^−/− mice with B cell-deficient (NOD-μMT) mice. Although all of the neuropathic mice produced IgG autoAbs against myelin sheath (Fig. 3 f–i), B cells were dispensable for the development of polyneuropathy, because all of the female NOD-H2^b/bPdcd1^−/−μMT and nondiabetic female NOD-H2^b/g7Pdcd1^−/−μMT mice examined (n = 6 and 7, respectively) developed polyneuropathy (Fig. 3e). These results indicate that B cells are not required for either the activation or the effector steps of neuritis in NOD-H2^b/bPdcd1^−/− mice.

Strong Th1 Polarization of the Nerve-Infiltrating T Cells and Increased Numbers of Memory T Cells in the Spleens of Neuropathic Mice.

We next analyzed the production and activation status of T cells in the thymus and spleen. The number or cellular composition of either thymocytes or splenocytes was unchanged in neuropathic NOD-H2^b/bPdcd1^−/− mice (SI Fig. 8). However, the splenic CD4 and CD8 T cells were more activated in the NOD-H2^b/bPdcd1^−/− mice (Fig. 4 a and b). In addition, the frequencies of CD44^hiCD122^hi memory CD8 T cells and CD44⁺CD62L⁻ effector memory CD4 T cells were increased in the spleen of NOD-H2^b/bPdcd1^−/− mice compared with that of NOD-H2^b/bPdcd1^+/+ mice (Fig. 4b). These results suggest that PD-1 deficiency may accelerate the generation and/or survival of memory T cells, which may promote the establishment of chronic autoimmune inflammation.

Fig. 4.

Th1-biased activation of the nerve-infiltrating T cells. (a) The fraction of CD44⁺, CD62L⁻, and CD69⁺ cells is shown for CD4 (Left) and CD8 (Right) T cells in the spleen of 25-wk-old NOD-H2^b/bPdcd1^+/+ (white bars) and NOD-H2^b/bPdcd1^−/− (black bars) mice. (b) Memory CD4 T cells (Left) and CD8 T cells (Right) in spleen are shown for NOD-H2^b/bPdcd1^+/+ (Top) and NOD-H2^b/bPdcd1^−/− (Lower) mice. (c) Statistics of the cellular composition of the nerve-infiltrating cells are shown. (d) Expression of indicated molecules on the sciatic nerve-infiltrating cells (thick lines) and splenocytes (filled histograms) is shown for CD4 (Upper) and CD8 (Lower) T cells. (e) Increase of Th1 and inflammatory cytokines in the inflamed nerve. Relative quantities of the mRNAs for the indicated genes in the inflamed nerve in relation to control nerve are shown. Data are the mean + SEM of three mice. (f) Representative FACS profiles of INFγ- and IL-4-stained spleen cells from NOD-H2^b/bPdcd1^+/+ (Left) and NOD-H2^b/bPdcd1^−/− (Center) mice and of nerve-infiltrating cells from NOD-H2^b/bPdcd1^−/− mice (Right) are shown for CD4 (Upper) and CD8 (Lower) T cells. (g) Mean percentage of INFγ- (Left), IL-4- (Center), and IL-17- (Right) producing CD4 and CD8 T cells is shown. Data are the mean + SEM. Representative data of more than three experiments are shown. *, P < 0.05; **, P < 0.01.

Next, we analyzed the nerve infiltrates by FACS. Consistent with the immunohistological analysis (Fig. 2e), 48.1 ± 0.7% of the nerve infiltrates were CD4 T cells, whereas CD8 T cells and B cells accounted for 20.2 ± 0.6% and 19.4 ± 2.4% of the nerve infiltrates, respectively (Fig. 4c). Both CD4 and CD8 T cells were more activated in the sciatic nerves than in the spleen (Fig. 4d).

We then examined the cytokine production in inflamed nerve. As shown in Fig. 4e, we detected higher amounts of Th1 cytokines including IFNγ, IL-12, and IL-15 and inflammatory cytokines including IL-1α and IL-1β in inflamed nerve compared with control nerve. The increase of IL-4 mRNA is much less substantial, and the amounts of IL-5 and IL-13 mRNAs were below the detection level. In addition, the amounts of IL-17 and TGFβ mRNAs were lower in inflamed nerves compared with control nerve, probably because of the suppression of the Th17 response by the Th1 response. Cytoplasmic staining of nerve infiltrates also showed similar tendency (Fig. 4 f and g). Collectively, these results suggest that the Th1 but not the Th2 or Th17 immune response plays a central role in the development of neuritis in this animal model.

Neuritogenic T Cells Can Be Generated in the Context of Either H-2^b Or H-2^g7.

To analyze the H-2-restriction of neuritogenic T cells, we transferred spleen cells from neuropathic NOD-H2^b/g7Pdcd1^−/−mice into NOD-SCID and C57BL/6-Rag2^−/− mice that were homozygous for H2^g7 and H2^b, respectively. Unexpectedly, not only C57BL/6-Rag2^−/− but also NOD-SCID recipients developed neuritis, indicating that neuropathic NOD-H2^b/g7Pdcd1^−/−mice contain both H-2^b- and H-2^g7-reactive neuritogenic T cells (Fig. 5a).

Fig. 5.

Neuritogenic T cells can be generated in the context of either H-2^b or H-2^g7. (a) Total spleen cells from neuropathic NOD-H2^b/g7Pdcd1^−/− mice were injected into C57BL/6-Rag2^−/− and NOD-SCID mice (Left). The severity of polyneuropathy is shown for each recipient mouse (Right). (b) Development of neuritis in the absence of H-2^b. The incidence of severe (black) and moderate (gray) neuritis is shown for female nondiabetic [(NOD-H2^g7/g7Pdcd1^−/− × C57BL/6-H2^b/bPdcd1^−/−)_F1 × NOD-H2^g7/g7Pdcd1^−/−]_BC1 mice with H2^g7/g7 (Left, n = 12) and H2^b/g7 (Right, n = 41) at 24–30 weeks of age.

To address whether these neuritogenic T cells can be generated in the context of H-2^g7 or whether they can be generated only in the context of H-2^b and cross-react with H-2^g7, we took the advantage of mixed genetic background because NOD-H2^g7/g7Pdcd1^−/− mice die from T1DM before 10 weeks of age. After omitting male and diabetic mice from 187 ((NOD-H2^g7/g7Pdcd1^−/− × C57BL/6-H2^b/bPdcd1^−/−)_F1 × NOD-H2^g7/g7Pdcd1^−/−)_BC1 progenies, which we had used in our previous genetic linkage analysis for T1DM-susceptibility loci (18), we obtained 12 and 41 nondiabetic female mice with the H2^g7/g7 and H2^b/g7 haplotypes, respectively. These mice were protected from T1DM probably because their Idd17, Idd20, Iddp1, and/or Iddp2-containing chromosomal region(s) are derived from C57BL/6 (18). Surprisingly, the neuritis incidence (33% and 39% for H2^g7/g7 and H2^b/g7, respectively) and severity (0.50 and 0.61 for H2^g7/g7 and H2^b/g7, respectively) were comparable between these H2^g7/g7 and H2^b/g7 progenies (Fig. 5b), indicating that neuritogenic T cells can be generated even in the context of H-2^g7. Therefore, it is reasonable to speculate that the reason why NOD-H2^g7/g7Pdcd1^−/− mice do not develop neuritis is because they die from T1DM too early. Taken together, neuritogenic T cells can be generated in the context of either H-2^b or H-2^g7, in contrast to the strict requirement of H-2^g7 for diabetogenic T cells.

Discussion

In the present study, we found the complete protection from T1DM in NOD-H2^b/bPdcd1^−/− mice, confirming the requirement of H-2^g7 for the development of T1DM even in the absence of PD-1 (Fig. 1). Despite being protected from T1DM, NOD-H2^b/bPdcd1^−/− mice spontaneously developed autoimmune inflammation in multiple tissues including nerves, stomachs, and exocrine tissues probably because PD-1 deficiency invigorated latent autoreactive T cells intrinsic to NOD-H2^b/b mice as reported (18). Interestingly, however, neuritogenic T cells were generated in the context of either H-2^b or H-2^g7, which is in sharp contrast to the strict requirement of H-2^g7 in T1DM. These results support the idea that H-2^g7 is more prone to generate autoreactive T cell clones (5–7).

Although NOD-H2^b/b mice have been reported to develop only mild sialoadenitis (3), the introduction of PD-1 deficiency unraveled their hidden autoimmune susceptibility. Compared with NOD-H2^b/bPdcd1^−/− mice, C57BL/6-Pdcd1^−/− mice that are homozygous for H2^b develop only mild autoimmunity in joint and kidney (17), indicating that H-2^b alone is not sufficient for the generation of autoreactive T cells against neural, gastric, and exocrine tissues even in the absence of PD-1. Using TCR/insHEL double-transgenic mice on either H-2^k congenic NOD or B10 background, Lesage et al. (25, 26) found that non-H-2 NOD genes allow self-reactive CD4 T cells with high avidity to escape from negative selection. Therefore, it is likely that autoreactive T cells against neural, gastric, and exocrine tissues are generated in the context of H-2^b under the support of non-H-2 NOD genes and induced overt autoimmune diseases in the absence of PD-1. Our observation is consistent with the idea that both H-2 and non-H-2 NOD genes play critical roles in the general autoimmune proclivity of NOD mice.

Another intriguing observation is the incomplete protection from T1DM by the single dose of H-2^b under PD-1 deficiency. Although previous studies have demonstrated that non-NOD H-2 haplotypes dominantly suppress T1DM in NOD mice (1, 2), the actual mechanism through which non-NOD H-2 haplotypes dominantly suppress T1DM is still enigmatic, and diverse possibilities have been raised. These can be divided into two categories; (i) qualitative and (ii) quantitative (amount of H-2^g7) effects. The possible qualitative effects of antidiabetogenic H-2 haplotypes include the deletion of diabetogenic T cells (9, 27), the selection of antidiabetogenic Tregs (13, 14), and the Th2-polarization of potentially diabetogenic T cells (10, 11, 28, 29). The high frequency of T1DM (60%) in NOD-H2^b/g7Pdcd1^−/− mice demonstrates that a sufficient number of diabetogenic T cells for overt T1DM are generated even in the presence of H-2^b. Fife et al. (30) demonstrated that the PD-1 blockade does not abrogate the suppression of T1DM by Tregs using BDC2.5-TCR transgenic system. Keir et al. (24) have shown that PD-L1 on pancreatic β cells but not on hematopoietic cells is responsible for the suppression of T1DM by PD-1 in NOD mice, suggesting that the exacerbation of T1DM by PD-1 deficiency is irrelevant to Tregs. In accordance with these reports, we found no difference in the number and the function of Tregs in NOD-H2^b/bPdcd1^−/− mice (SI Fig. 9). Therefore, it is less likely that very potent antidiabetogenic Tregs are generated by H-2^b. The enhancement of Th1 response by PD-1 deficiency (Fig. 4) (18, 24) suggests that Th2-polirization of potentially diabetogenic T cells may contribute in part to the reduction of T1DM by H-2^b on NOD background. The observation that 100% and 60% of NOD-H2^g7/g7Pdcd1^−/− and NOD-H2^b/g7Pdcd1^−/− mice, respectively, developed T1DM can be most easily explained by the dose effect of H-2^g7. It is clear that H-2^b is not dominantly suppressing TIDM-induction by H-2^g7. Wicker et al. (3, 4) reported that H-2^g7 is “dominant” for T1DM based on the observation that 50% and 3% of the NOD-H2^b/g7 mice developed insulitis and T1DM, respectively. Genetically, however, their observation does not indicate the dominant effect of H-2^g7.

The spontaneous development of polyneuropathy has been reported for B7–2-deficient NOD mice (31) and H3 TKO mice that express I-Aβ^b covalently bound to the Eα chain-derived peptide (Eα52-68) and lack endogenous MHCs (32). IL-2 neutralization induces a wide spectrum of autoimmune diseases, including peripheral polyneuropathy, in NOD mice by reducing the number of CD4⁺CD25⁺ Tregs (33). Because the number and the function of Tregs are unaffected in the current model (SI Fig. 9), the pathomechanism of polyneuropathy in this model may be different from that of IL-2 neutralization in NOD mice. The target peptide of the neuritogenic T cells in NOD-H2^b/bPdcd1^−/− mice is totally different from that in H3 TKO mice because Eα is not expressed in either the H-2^b or H-2^g7 haplotype. Nevertheless, the polyneuropathy in NOD-H2^b/bPdcd1^−/− mice and B7–2-deficient NOD mice shares various features including Th1-predominance and the contribution of CD4 T cells as effector. The major differences between these models are the production of autoAbs and the infiltration of fewer CD11c⁺ dendritic cells in the current model. We found an augmented expression of B7–2 and CTLA-4 in the inflamed nerves of Pdcd1^−/− mice compared with normal nerves (SI Fig. 10), suggesting that the B7–2/CTLA-4 pathway is intact in the Pdcd1^−/− mice. Therefore, the pathomechanisms underlying the polyneuropathy in these two models seem to be different, at least in part.

PD-1 deficiency reduces the threshold of T cell activation and strengthens autoimmunity through various immunological mechanisms, including disruption of anergy, augmentation of cytotoxicity, and promotion of autoAb production (15, 16). We observed an increase in IFNγ- but not IL-4- or IL-17-producing cells in the inflamed nerves of NOD-H2^b/bPdcd1^−/− mice, which is consistent with the augmented Th1 response in the islets of NOD-Pdcd1^−/− mice. NOD mice have been reported to mount a predominantly Th1-biased response, which may be explained partly by the recent findings that macrophages from NOD mice produce increased IL-12, a Th1-inducing cytokine, and that Idd4 is responsible for its overexpression (34, 35). PD-1 deficiency may thus cooperate with the genetic Th1 predominance of NOD mice to induce overt autoimmunity with a variable target that depends on the H-2 haplotype. Further analyses are needed to unravel the precise mechanism by which PD-1 deficiency induces various autoimmune diseases in collaboration with other strain-specific autoimmune susceptibility genes.

Materials and Methods

Animals.

NOD/Shi.Jic and NOD-SCID mice were purchased from Japan Clea. NOD.B10Sn-H2^b/J and NOD.129S2(B6)-Igh-6^tm1Cgn (NOD-μMT) mice were purchased from The Jackson Laboratory and mated with NOD-Pdcd1^+/− mice (18). All mouse protocols were approved by the Institute of Laboratory Animals, Graduate School of Medicine, Kyoto University. Mice were maintained under specific pathogen-free conditions.

Histological Analysis.

Polyneuropathy was scored as follows: grade 0, no change; grade 1, lymphocytic infiltration without clinical signs; grade 2, mild limb paralysis with lymphocytic infiltration; grade 3, severe limb paralysis with lymphocytic infiltration.

Isolation of Sciatic Nerve-Infiltrating Cells.

Sciatic nerves were minced with scissors and incubated in a digestion buffer [PBS, 2% FCS, and 1 mg/ml collagenase (Wako Chemicals)] for 15 min at 37°C, with continuous agitation. Digested nerves were passed through a 22-gauge needle, further incubated for 10 min at 37°C, and washed with staining buffer (PBS, 2% FCS, and 0.01% azide).

Transfer Experiment.

CD4 T cells and CD4-depleted splenocytes were prepared from the spleens of neuropathic NOD-H2^b/bPdcd1^−/− mice by using anti-CD4 microbeads according to the manufacturer's protocol. (Miltenyi Biotec). Next, 2 × 10⁷ unfractionated splenocytes, 1 × 10⁷ CD4 T cells (the average purity of the CD4 T cells was >95%, and the contamination of CD8 T cells was <0.3%), or 2 × 10⁷ CD4-depleted splenocytes (the average contamination with CD4 T cells was <0.5%) were transferred intravenously into 4- to 6-week-old C57BL/6-Rag2^−/− mice.

Real-Time PCR Using TaqMan Low-Density Arrays.

The amount of indicated mRNAs was evaluated by using Taqman Low-Density Array Immuno Profiling Plate (Applied Biosystems) as described in SI Materials and Methods.

Statistics.

The two-tailed unpaired Student's t test was used for statistical analyses.