Abrogation of ICOS/ICOS ligand costimulation in NOD mice results in autoimmune deviation toward the neuromuscular system

Authors


Abstract

NOD mice spontaneously develop insulin-dependent diabetes around 10–40 wk of age. Numerous immune gene variants contribute to the autoimmune process. However, genes that direct the autoimmune response toward β cells remain ill defined. In this study, we provide evidence that the Icos and Icosl genes contribute to the diabetes process. Protection from diabetes in ICOS−/− and ICOSL−/− NOD mice was unexpectedly associated with the development of an autoimmune disorder of the neuro-muscular system, characterized by myositis, sensory ganglionitis and, to a reduced extent, inflammatory infiltrates in the CNS. This syndrome was reproduced upon adoptive transfer of CD4+ and CD8+ T cells from diseased donors to naïve NOD.scid recipients. Our data further show that protection from diabetes results from defective activation of autoimmune diabetogenic effector T cells in ICOS−/− NOD mice, whereas acceleration of diabetes in BDC2.5 ICOS−/− NOD mice is induced by a dominant defect in Treg. Taken together, our findings indicate that costimulation signals play a key role in regulating immune tolerance in peripheral tissues and that the ICOS/ICOSL costimulatory pathway influences the balance between Treg and diabetogenic effector T cells.

Introduction

Type 1 diabetes (T1D) is an autoimmune disease characterized by the activation of autoreactive lymphocytes against pancreatic β-cell antigens. Mechanisms that initiate the failure of immune tolerance to β cells remain elusive in common forms of T1D. The disease develops as a multifactorial process in which environmental factors concur with a highly multigenic susceptibility background 1. However, genes that direct the autoimmune response toward β cells remain ill defined.

In animal models such as the NOD mouse, the predominant role of T lymphocytes is supported by experiments in which T1D is transferred into naïve recipients by T lymphocytes from diabetic or prediabetic animals. Moreover, T1D is prevented by injection of Ab that interfere with T-lymphocyte activation and fails to develop in diabetes-prone mice in which key genes in T-lymphocyte differentiation or activation are nonfunctional 2. Both CD4+ and CD8+ T cells are involved 3–5.

The activation of autoreactive T lymphocytes in T1D requires the recognition of autoantigens expressed by β cells as well as costimulatory signals. Membrane proteins of the CD28 gene family mediate major costimulatory signals through their interaction with members of the B7 family expressed on APC and stromal cells. CD28 6, 7, CTLA4 8, and ICOS 9, 10 are related proteins of the CD28 family that are encoded by genes clustered within a 300 kb region located on chromosome 2. Gene variants within this region have been associated with susceptibility to T1D 11, 12. In vivo deletion of the Cd28 or B7 genes, pharmacological blockade of the CD28 or B7 molecules using monoclonal Ab 13–15, and anti-CTLA4 treatment or the expression of a CTLA4-Ig transgene have all shown to profoundly affect the development of T1D in the NOD mouse 16, 17. The role of the ICOS/ICOS ligand (ICOSL) costimulatory pathway in the pathogenesis of T1D is still controversial. On the one hand, blockade 18, 19 or in vivo deletion 20 of ICOS inhibits the development of T1D in NOD mice; on the other hand, treatment with an ICOS-blocking Ab induces acute T1D in transgenic BDC2.5 NOD mice 21.

To better understand the role of the ICOS/ICOSL costimulatory pathway in the development of T1D, we used a genetic approach, introgressing the Icos and Icosl null mutations 22 onto the NOD background. We confirm the key role of the ICOS/ICOSL costimulatory pathway in the activation of Treg and diabetogenic effector T cells (Teff cells) in the NOD mouse. However, ICOS−/− or ICOSL−/− NOD mice, while protected from T1D, develop an autoimmune disease that affects neural and muscular cells, indicating that local T-cell interaction within the islet environment is a key factor in immune tolerance.

Results

Abrogation of the ICOS/ICOSL costimulatory pathway protects NOD mice from diabetes

To define the role of ICOS/ICOSL costimulation in T1D development, we studied ICOS−/− and ICOSL−/− NOD mice. The spontaneous incidence of T1D was significantly reduced in female ICOS−/− NOD mice (Fig. 1A) since only 16.6% of them became diabetic by 40 wk of age, as compared with 63.6% for female ICOS+/+ NOD mice by 40 wk of age. Since the Icos locus (1C2) is located in proximity of other immuno-modulatory genes, including CD28 and CTLA-4, we analyzed mice deficient for Icosl, which lies in a completely different genetic region. Female ICOSL−/− NOD mice were protected from T1D to a similar extent as female ICOS−/− NOD mice (Fig. 1A). At 40 wk of age, none of them became diabetic, whereas 57.5% female ICOSL+/+ NOD mice developed T1D. Interestingly, both female ICOS−/− and ICOSL−/− NOD mice were also protected from insulitis. At 14 wk of age, 48% of islets from ICOS+/+ mice were infiltrated, with 30% of the islets showing invasive insulitis, whereas only 2% islets were infiltrated in ICOS−/− mice (Fig. 1B and C). At 40 wk of age, ICOS−/− NOD mice were free from insulitis, whereas 9.5% of the islets from ICOSL−/− NOD mice showed peri-insulitis (Fig. 1B). In the light of these findings, we next investigated whether the inability to induce T1D was intrinsic to spleen cells from ICOS−/− NOD mice. To this end, splenocytes from 40-wk-old ICOS−/− or ICOS+/+ mice were activated in vitro by anti-CD3 and anti-CD28 Ab and then adoptively transferred to NOD.scid recipients (Fig. 1D). None of the recipient mice transferred with splenocytes from ICOS−/− NOD mice developed T1D upon transfer within a 90-day time frame. By contrast, all recipient mice transferred with splenocytes from control ICOS+/+ NOD mice became diabetic within 30 days following transfer (Fig. 1D). Interestingly, both ICOS−/− and ICOSL−/− NOD mice were resistant to the induction of T1D by cyclophosphamide at 8 wk of age (data not shown). Taken together, these observations demonstrate that the ICOS/ICOSL pathway is critical in the early steps leading to failure of immune tolerance to β cells.

Figure 1.

ICOS−/− and ICOSL−/− NOD mice are protected from T1D. (A) Spontaneous incidence of T1D in female ICOS+/+ (▪, n=11), ICOS−/− (□, n=12), ICOSL+/+ (•, n=18), ICOSL−/− (○, n=15) NOD mice. *p<0.05 (Logrank Test). (B) ICOS−/− and ICOSL−/− NOD mice are almost free from insulitis. Percentage of normal islets, peri-insulitis, and insulitis in 14-wk-old female ICOS+/+ (n=10 mice) or ICOS−/− (n=9) NOD mice, and in 40-wk-old female ICOS−/− (n=6) or ICOSL−/− (n=6) NOD mice. (C) Representative sections of pancreases from 14-wk-old female ICOS+/+ or ICOS−/− NOD mice. (D) Splenocytes from 40-wk-old ICOS−/− NOD mice do not transfer T1D. Briefly, 8-wk-old NOD.scid mice were transferred with 10×106 stimulated splenocytes from 40-wk-old ICOS−/− (□ ICOS−/− donor total splenocytes, n=4) or ICOS+/+ (▪ ICOS+/+ donor total splenocytes, n=4) NOD mice. **p<0.01 (Logrank Test). Data are representative of three independent experiments.

ICOS expression during the T1D development

To better understand how the ICOS/ICOSL costimulatory pathway may interfere with the diabetogenic process, we studied ICOS expression during T1D development. We found that ICOS expression was the highest in the islet cell infiltrate (Fig. 2A). In total, 29.2% of CD4+ islet-infiltrating T cells expressed ICOS at 8 wk of age, as compared with 7% in pancreatic LN (PLN), 7.5% in axillary LN (ALN), and 9.2% in the spleen. No significant peak of ICOS expression was seen on CD4+ T cells at earlier stages of the NOD autoimmune process, especially in PLN, although ICOS expression was slightly increasing with age (i.e. at 3, 5, and 8 wk of age; Fig. 2B). Since ICOSL can be expressed by various tissues under inflammatory conditions, we studied ICOSL expression in the pancreas. Although ICOSL was expressed, as expected, on infiltrating B cells and APC (data not shown), we observed no ICOSL expression on β cells using flow cytometric analysis (Fig. 2C).

Figure 2.

ICOS is strongly expressed on islet-infiltrating CD4+ T cells. (A) ICOS expression on islet-infiltrating T cells. Flow cytometric analysis of ICOS expression on unmanipulated T cells from spleen, PLN, ALN, pancreas of 8-wk-old ICOS+/+ NOD mice (n=6). Plots show representative staining of ICOS intensity gated on CD4+ living cells. Numbers represent the percentage of ICOS+ cells within CD4+. Data are representative of three independent experiments. (B) ICOS expression during T1D development. Expression of ICOS on unmanipulated T cells from spleen, PLN, ALN, pancreas of 3, 5, and 8-wk-old ICOS+/+ NOD mice (n=6 per group). Each bar corresponds to the mean±SD of values obtained. *p<0.05 (Student's t-test). Data are representative of three independent experiments. (C) β cells do not express ICOSL. The first dot plot on the left shows how β cells were distinguished from non-β cells based on their autofluorescence (FAD content) and cell size, resulting in a population with 90% (insulin-positive) β cells. Second and third dot plots on the right show representative staining of ICOSL intensity gated on unmanipulated β cells pooled from three pancreas of 8-wk-old ICOS+/+ or ICOS−/− NOD mice. Numbers are percentage of ICOSL+ cells within insulin-positive β cells. As positive control, the lower dot plot shows representative staining of ICOSL intensity gated on unmanipulated CD19+ cells pooled from ALN from 8-wk-old ICOS+/+ NOD mice, treated as β cells with collagenase and dissociation buffer. Data representative of five independent experiments.

Abrogation of ICOS/ICOSL costimulatory pathway induces acute diabetes in BDC2.5 NOD mice

Several lines of evidence point toward an important role of the ICOS/ICOSL pathway in the differentiation of Treg, including the observation that anti-ICOS monoclonal Ab treatment accelerates the development of T1D in BDC2.5 NOD mice 21. Since it is difficult to evaluate the level of saturation and accessibility of target cells in studies using blocking Ab, we sought to evaluate the role of ICOS in the development of T1D of BDC2.5 ICOS−/− NOD mice. We observed an increased prevalence of spontaneous T1D in BDC2.5 ICOS−/− NOD mice as compared with BDC2.5 ICOS+/+ NOD mice (Fig. 3A). At 20 wk of age, 66% of female BDC2.5 ICOS−/− became diabetic, as compared with 8.33% of control mice. This observation raises the possibility that the ICOS pathway may control both the activation of autoimmune Teff cells and of Treg, but that T-cell dependence on ICOS may vary depending on the T-cell clone considered.

Figure 3.

Acute T1D and decreased Treg in spleen and PLN of BDC2.5 ICOS−/− NOD mice. (A) BDC2.5 ICOS−/− NOD mice develop acute T1D. Spontaneous incidence of T1D in BDC2.5 ICOS+/+ (♦, n=10 mice) and BDC2.5 ICOS−/− (◊, n=10 mice) NOD females. *p<0.05 (Logrank Test). (B, C) Decreased Treg in BDC2.5 ICOS−/− NOD mice. Percentage (B) and absolute number (C) of Foxp3+ cells within CD4+ T cells in spleen, PLN, ALN of 8-wk-old, unmanipulated, BDC2.5 ICOS+/+ (n=7) or BDC2.5 ICOS−/− (n=8) NOD mice. Each bar represents the mean±SD of values obtained. *p<0.05 and **p<0.01 (Student's t-test). Data representative of three independent experiments.

A significant decrease in the absolute number and the percentage of CD4+Foxp3+ T cells was observed in spleen and PLN obtained from BDC2.5 ICOS−/− NOD mice as compared with WT littermates (Fig. 3B and C). A 26.3–40.0% decrease of percentages of CD4+Foxp3+ T cells was observed in BDC2.5 ICOS−/− NOD mice, respectively, in PLN and spleen. Considering absolute numbers of CD4+Foxp3+ T cells, a comparable decrease was observed, with a 31–48% decreased in spleen and PLN. Similar results were obtained in ICOS−/− and ICOSL−/− NOD mice as compared with their respective littermates (Supporting Information Fig. 1). Whatever the strains considered, no decrease was seen in ALN.

We next evaluated the ability of Treg from BDC2.5 ICOS−/− NOD mice, as compared with control cells, to inhibit the proliferative response of Teff cells to anti-CD3 Ab or RVRPLWVRME mimotope peptide. As shown in Fig. 4A, no defective inhibitory capacity of CD4+CD25+ Treg from BDC2.5 ICOS−/− NOD mice was observed when cells were stimulated with anti-CD3 Ab. In contrast, CD4+CD25+ Treg from BDC2.5 ICOS−/− NOD mice were less efficient in suppressing the proliferative response of CD4+CD25 Teff cells from BDC2.5 ICOS+/+ NOD mice to stimulation by the RVRPLWVRME peptide than BDC2.5 CD4+CD25+ Treg from BDC2.5 ICOS+/+ NOD mice (Fig. 4B). These results were confirmed at various CD4+CD25+/CD4+CD25 T-cell ratios. Taken together, our data suggest that qualitative and quantitative defects in ICOS-dependent Treg drive the increased incidence of T1D in transgenic BDC2.5 ICOS−/− NOD mice.

Figure 4.

Treg from BDC2.5 ICOS−/− are less efficient in suppressing T-cell proliferation in response to peptide stimulation. (A, B) In total, 5×104 CD4+CD25CFSE+ Teff cells (Teff cells) from 8-wk-old BDC2.5 ICOS+/+ NOD mice (n=3) were cultured in the presence of CD4+CD25+ Treg from 8-wk-old BDC2.5 ICOS+/+ (♦ BDC2.5 ICOS+/+ donor Treg, n=3) or BDC2.5 ICOS−/− (◊ BDC2.5 ICOS−/− donor Treg, n=3) NOD mice at indicated ratio (Treg/Teff). Cells were stimulated by anti-CD3 Ab (1 μg/mL, (A)) or by RVRPLWVRME peptide (60 μg/mL, (B)). Proliferation of CD4+CD25 Teff cells was assessed after 72h by CFSE dilution. *p<0.05 (Student's t-test). Data are representative of three independent experiments.

A low frequency of ICOS-independent diabetogenic precursors contributes to diabetes protection in ICOS−/− NOD mice

The development of accelerated T1D in BDC2.5 ICOS−/− NOD mice points to diabetogenic BDC2.5 Teff cells as ICOS-independent. The difference between polyclonal T cells in conventional NOD mice and BDC2.5 T cells in BDC2.5 transgenics is likely in the frequency of corresponding precursors. To test this hypothesis, we performed transfer experiments in which an increasing ratio of BDC2.5 monoclonal Teff cells/polyclonal Teff cells was transferred into ICOSL−/− NOD.scid recipients. As shown in Fig. 5, 2×105 CD4+CD25 Teff cells from ICOS−/− NOD mice failed to transfer T1D, as it was the case when 1×103 CD4+CD25 Teff cells from BDC2.5 ICOS−/− mice and 1.99×105 CD4+CD25 polyclonal Teff cells from conventional ICOS−/− NOD mice were cotransfered into ICOSL−/− NOD.scid recipients (ratio 1:100). In contrast, co-injection of 5×103 CD4+CD25 Teff cells from BDC2.5 ICOS−/− NOD mice with 1.95×105 CD4+CD25 Teff cells from ICOS−/− NOD mice (ratio 1:40) was sufficient to efficiently transfer T1D in ICOSL−/− NOD.scid recipients. The frequency of potent diabetogenic Teff cells thus explains the opposite T1D phenotype observed in conventional ICOS−/− NOD mice and BDC2.5 ICOS−/− NOD mice.

Figure 5.

Dependence of NOD diabetogenic T cells on the ICOS pathway is determined by frequency of precursors. About 8-wk-old ICOSL−/− NOD.scid mice were injected with CD4+CD25 Teff cells from 14-wk-old ICOS−/− NOD mice (ICOS−/− donor Teff) and/or from BDC2.5 ICOS−/− NOD mice (BDC2.5 ICOS−/− donor Teff) at indicated ratio (• 2×105 ICOS−/− donor Teff, n=3; ▵ 1.99×105 ICOS−/− donor Teff and 1×103 BDC2.5 ICOS−/− donor Teff, n=4; ▿1.95×105 ICOS−/− donor Teff and 5×103 BDC2.5 ICOS−/− donor Teff, n=4; ♦ 1.90×105 ICOS−/− donor Teff and 1×104 BDC2.5 ICOS−/− donor Teff, n=4; ○ 2×105 BDC2.5 ICOS−/− donor Teff, n=3). *p<0.05 as compared with mice injected with 2×105 ICOS−/− donor Teff (Logrank Test). Data are representative of three independent experiments.

Development of neuromuscular autoimmunity in ICOS−/− and ICOSL−/− NOD mice

Beginning at 20 wk of age, ICOS−/− and ICOSL−/− NOD mice exhibited a symmetrical, rapidly progressing hind leg paralysis with a higher prevalence in female (100 and 72.9%, respectively at 40 wk of age) than in males (20 and 26.9%, respectively, at 40 wk of age) (Fig. 6A). The disease was not observed in corresponding control NOD mice. ICOS−/− NOD mice showing evidence for paralysis were analyzed for pathological alterations in the nervous system and muscle (Fig. 6B). Inflammatory infiltrates seen in muscles, peripheral nerves, sensory ganglia, brain, and spinal cord were composed of T cell, macrophages, and a variable number of granulocytes. Overall, the picture was that of a T-cell-mediated inflammatory reaction, potentially of autoimmune origin. By contrast, no histological lesions were seen in control ICOS+/+ NOD mice in the corresponding organs (Fig. 6B).

Figure 6.

Muscle and neural autoimmunity in ICOS−/− NOD mice. (A) Spontaneous incidence of neuropathy in male (▴, n=10) and female (▪, n=10) ICOS−/− NOD mice and in male (▵, n=15) and female (□, n=15) ICOSL−/− NOD mice. Mice were checked weekly for clinical signs of neuropathy, consisting chronologically in hind legs weakness, ataxia, and sensitive troubles. Data are representative of two independent experiments. (B) Histology of spinal ganglia and muscles from 40-wk-old ICOS+/+ or ICOS−/− NOD females. As compared with age-matched female ICOS+/+ NOD mice, female ICOS−/− NOD mice show inflammation and infiltrating T cells (left column for hematoxylin and eosin staining, 90× magnification; right column for anti-CD3 immunocytochemistry staining, 90× magnification) in spinal ganglia (upper panel) and in muscle (lower panel). (C) Briefly, 8-wk-old NOD.scid mice (n=5 per group) were injected with 10×106 stimulated total spleen cells (▵, ICOS−/− donor total cells), 5×106 purified stimulated T cells including CD4+ T cells and CD8+ T cells (□, ICOS−/− donor T cells), 5×106 stimulated splenocytes depleted of CD4+ and CD8+ T cells (○, ICOS−/− donor non-T cells) from 40-wk-old ICOS−/− NOD mice. Data are representative of three independent experiments. (D) The CD28/B7 pathway is not altered in the ICOS−/− NOD mice. Flow cytometric analysis of B7.1 and B7.2 expression on B cells from spleen of 8-wk-old ICOS+/+ or ICOS−/− NOD mice before and after in vitro stimulation by LPS. Plots show representative staining of B7.1 or B7.2 intensity gated on CD19+ living cells. Data are representative of three independent experiments.

To assess whether muscular and neurological lesions were immune-mediated, purified T cells including CD4+ T cells and CD8+ T cells, splenocytes depleted of CD4+ and CD8+ T cells or total splenocytes from 40-wk-old ICOS−/− or ICOS+/+ NOD mice were first activated in vitro, then adoptively transferred into conventional NOD.scid recipients (Fig. 6C). Muscular and neurological symptoms, consisting in hind leg weakness and ataxia, were observed only in recipients of total splenocytes or purified T cells including CD4+ and CD8+ T cells from ICOS−/− NOD mice. Clinical paralysis started 4–6 wk post-transfer. In contrast, none of the NOD.scid recipients that received total splenocytes or purified T cells including CD4+ and CD8+ T cells from ICOS−/− NOD mice developed T1D. Clinical T1D was observed only in recipient mice transferred with corresponding T cells from ICOS+/+ NOD mice (data not shown). Neither T1D nor neuromuscular disease was observed in mice transferred by splenocytes depleted of both CD4+ and CD8+ T cells from ICOS+/+ or ICOS−/− NOD mice. Histological analyses of recipient mice were performed 30 days after transfer. Although no infiltrates were found in CNS, the same pathological alterations as those seen in ICOS−/− mice developing a spontaneous neuromuscular disease were observed in the peripheral nervous system (PNS) and in muscles (data not shown).

Since a similar neuropathy phenotype was previously reported in the B7.2−/− NOD mice, and since the ICOS/ICOSL costimulation pathway belongs to the same costimulation superfamily as CD28/B7, we checked the expression of CD28, B7, and CTLA-4. No significant difference in the expression of CD28 and CTLA-4 on T cells was seen in ICOS−/− NOD mice as compared with ICOS+/+ NOD mice (data not shown). As shown in Fig. 6D, expression of B7.1 and B7.2 on B cells remains unchanged in both strains before and after 24 h stimulation by LPS. These data demonstrate that the neuropathy observed in ICOS−/− NOD mice does not depend on secondary modifications of the CD28/B7 pathway.

Discussion

Given the role of costimulation in T-cell responses to antigens, attempts to modulate signaling through costimulatory pathways are expected to modify the development of autoimmunity in models in which T cells are central in the autoimmune process 13, 14, 23. Understanding the role of the ICOS/ICOSL costimulatory pathway in processes that lead to autoimmune destruction of pancreatic β cells in T1D is thus a key issue and the first step toward potential therapeutic applications. In this study, using a genetic deletion approach, we have demonstrated that the ICOS/ICOSL costimulation pathway is required for the development of T1D in the NOD mouse. We observed that ICOS−/− and, for the first time, ICOSL−/− NOD mice are both protected from developing T1D. Since the Icos gene and the Icosl gene locate on different chromosomes, our data indicate that protection against T1D directly relates with deficient ICOS/ICOSL pathway and is not due to the effect of a distinct Idd gene located in the vicinity of Icos, such as Cd286, 7 or Ctla48. Moreover, the inefficiency of cyclophosphamide injections to induce acute T1D, the striking protection from insulitis in 40-wk-old ICOS−/−, and ICOSL−/− NOD mice, the failure to transfer T1D into NOD.scid mice after adoptive transfer of splenocytes from ICOS−/− NOD mice suggests that the ICOS/ICOSL pathway is necessary in the early steps that lead to the failure of immune tolerance to β cells.

Interestingly, as compared with the levels of expression within spleen, PLN, and ALN, a high percentage of islet-infiltrating CD4+ T cells express ICOS in conventional 8-wk-old NOD mice. More surprisingly, no peak of ICOS expression was ever observed in PLN along T1D development. However, it is possible that we failed to capture an ICOS expression peak since activated T cells rapidly migrate into the islets where further increase in ICOS expression is likely to take place. We studied ICOSL expression in pancreas in order to further precise the role of the ICOS/ICOSL pathway in the islets. In situ, ICOSL expression was restricted to conventional APC. No ICOSL expression was observed on β cells. High ICOS expression by islet T cells is thus possibly secondary to the local expansion of diabetogenic T cells within the islets independently of ICOSL expression by β cells.

Although we cannot exclude that the ICOS/ICOSL pathway is necessary for activation and homeostasis of diabetogenic T cells, we have to take into account the role of ICOS and ICOSL in T-cell-dependent B-cell responses 22, 24–26. The ICOS/ICOSL pathway is indeed critical for isotypic commutation and somatic mutation in B cells. Defective Icos or Icosl expression in NOD mice is expected to disrupt interactions between T and B cells and prevent the failure of immune tolerance to β cells initiated by B lymphocytes. B cells are essential APC during T1D development 27–29. This is consistent with the diminished production of auto-Ab that has been previously reported in ICOS−/− NOD mice 20.

The increased incidence of T1D seen in BDC2.5 ICOS−/− NOD mice suggests that the ICOS/ICOSL pathway is critical for Treg. In ICOS−/− and BDC2.5 ICOS−/− NOD mice, a significant decrease in percent and absolute number of CD4+Foxp3+ T cells is seen in spleen and PLN although not in ALN. This difference may reflect the local immune activation process that leads to T1D. Interestingly, in vitro, CD4+CD25+ Treg from BDC2.5 ICOS−/− NOD mice were significantly less efficient than control Treg in suppressing the T-cell response to a mimitope peptide but not to anti-CD3 Ab. These observations suggest that the ICOS/ICOSL pathway is critical for the function of both Teff cells and Treg, but that the dependence of T cells on ICOS varies depending on the T-cell clone considered and the strength of signals involved in T-cell activation.

Accelerated T1D in ICOS−/− BDC2.5 NOD mice indicates that all diabetogenic Teff cells do not depend on the ICOS/ICOSL costimulation pathway. Affinity of the TCR for peptide–major histocompatibility complexes is expected to inversely correlate with dependence of T-cell activation on ICOS costimulation. Addressing the affinity issue would require studying transgenic mice carrying different TCR specific for the same peptide–major histocompatibility complexes. However, transfer experiments using decreasing dilution of BDC2.5 diabetogenic Teff cells in a polyclonal NOD Teff cells pool allowed showing that the frequency of ICOS-independent diabetogenic Teff explains the opposite T1D phenotype observed in conventional ICOS−/− NOD mice and BDC2.5 ICOS−/− NOD mice.

For the first time, we report, in both ICOS−/− and ICOSL−/− NOD mice, a deviation of autoimmunity from the islets toward muscles, peripheral nerves, sensory ganglia, and, to a lesser extent, the CNS. Diffuse inflammatory infiltrates composed of T cells, macrophages, and variable numbers of granulocytes are observed, similar to those seen in autoimmune encephalitis, neuritis, and myositis. The observation of muscle and peripheral nerve autoimmunity is not unexpected if one considers that, beyond spontaneously developing T1D, NOD mice are susceptible to many other autoimmune diseases. These include autoimmune sialadenitis 30, autoimmune thyroiditis 31, hemolytic anemia 32, and prostatitis in male mice 33. Accordingly, the NOD-derived mouse has been considered as a model of primary Sjogren syndrome, Guillain-Barré syndrome and multiple sclerosis 34–36.

A shift from T1D to other autoimmune diseases has been evidenced in NOD mice in which a costimulatory molecule that is distinct from ICOSL, B7.2, was disrupted 34. Autoimmunity in B7.2−/− NOD mice is directed against myelin protein zero and is mainly directed against the PNS 37. A critical point is that the CD28/B7 pathway remains unaltered in ICOS−/− NOD mice as compared with control mice. Moreover, autoimmune deviation in ICOS−/− NOD mice extends to muscles, sensory ganglia and to a lower extent the CNS while it is only restricted to PNS in B7.2−/− NOD mice. Our observation, along with the previous report in the B7.2−/− NOD model, points to local interactions of T cells with other cell partners as a key factor in maintenance or disruption of organ-specific immune tolerance. Further experiments will be required to address whether Teff cells that drive muscle and/or neurological autoimmunity in ICOS−/− NOD mice can be downregulated by Treg from ICOS+/+ NOD mice that are likely to be ICOS-independent.

In conclusion, we bring evidence that ICOS/ICOSL interaction controls in part the activation of diabetogenic Teff cells and influences the balance between Treg and diabetogenic Teff cells. This likely explains the discordant observations made in conventional and transgenic BDC2.5 NOD mice. Furthermore, although ICOS−/− or ICOSL−/− NOD mice are protected from T1D, they develop autoimmunity against neural and muscular tissues, indicating that ICOS–ICOSL interactions are important in polarizing the NOD autoimmune response.

Materials and methods

Mice

NOD, ICOS−/− NOD, ICOSL−/− NOD, NOD.scid, ICOSL−/− NOD.scid, BDC2.5 NOD and BDC2.5 ICOS−/− NOD mice were bred and housed in our facilities under specific pathogen-free conditions. ICOS−/− NOD mice were generated by crossing NOD mice with ICOS−/−22 and backcrossing for nine generations to NOD mice. ICOSL−/− NOD mice were generated by backcrossing ICOSL−/−38 to NOD mice for 16 generations. The prevalence of T1D in our NOD colony reaches 10% in males and 75% in females by 6 month of age. Animal studies were approved by institutional review.

Diabetes assessments

T1D was assessed by monitoring mice for glycosuria using test strips from 4 wk on twice a week for spontaneous incidence or three times a week, 10 days after the beginning of adoptive transfer experiments. Blood glucose was measured (Euroflash, Lifescan, Milpitas, CA) when glycosuria was detected. Mice were diagnosed as diabetic after two consecutive glycemia over 200 mg/dL. The onset of T1D was dated from the first of the sequential glycemia measurements.

Histological analysis

To evaluate inflammation in nervous tissue and muscle, animals were perfused, fixed by 4% formaldehyde solution (Electron Microscopy Science, Hatfield, PA), then spinal ganglia and pieces of different skeletal muscles were dissected and embedded into paraffin. Briefly, 3 m sections were stained with hematoxylin and eosin. Immunocytochemistry was performed with primary Ab against T cells (CD3; MCA 1477; Serotec, GB) using a biotin/avidin/peroxidase technique.

Flow cytometric analysis

Ab

The following mAb were used (hybridoma clone indicated in parentheses): PerCP-anti-CD4 (RM4-5), PE-anti-ICOSL (HK5.3) from BD Bioscience (San Jose, CA); FITC-anti-B7.1 (16–10A1), PE-anti-B7.2 (GL1), APC-anti-CD19 (MB19-1), Biotine-anti-CD278/ICOS (15F9), PE-anti-Foxp3 (FJK-16s) from eBioscience (San Diego, CA). β cells were stained with bovin anti-insulin Ab (Sigma-Aldrich, St. Louis, MO) then with Biotin-SP-AffiniPure Goat Anti-Guinea Pig IgG (Jakson Immunoresearch Laboratories, West Grove, PA). Appropriate isotype control mAb were included.

Immunofluorescence staining

Single-cell suspensions were pelleted in 96-well plates, preincubated with 10 μg/mL 2.4G2 for 15 min at 4°C and washed twice in order to minimize unspecific binding. Cells were then stained for 30 min at 4°C with optimal concentrations of reagents in 20 μL of PBS supplemented with 2% FBS and 5 mM sodium azide. Cells were washed twice, stained again if necessary, or resuspended in PBS containing 1% formaldehyde. Intracellular Foxp3 staining was performed according to the manufacturer's instructions (Foxp3 anti-mouse/rat Foxp3 staining set; eBioscience). Flow cytometric analysis was performed using a FACSCalibur, FlowJo softwares (Tree Star, Ashland, OR).

Cell purifications

CD4+CD25 and CD4+CD25+ T cells were purified through a first step of CD4+ T-cell enrichment using mouse CD4 Selection Kit (Invitrogen Dynal AS, Carlsbad, CA) according to the manufacturer's instructions. Cells were then separated by MACS sorting, using LS columns, biotin-anti-CD25 (7D4, BD Bioscience) Ab, and streptavidin beads according to the manufacturer's instructions (Miltenyi Biotec, Bergisch Gladbach, Germany): CD4+CD25+ T cells were recovered from the “positive” fraction, whereas CD4+CD25 T cells were in the flow through. Purified T cells including CD4+ T cells and CD8+ T cells, splenocytes depleted of CD4+ and CD8+ T cells were directly separated by MACS sorting as described above using biotin-anti-CD8 and biotin-anti-CD4 Ab prepared in the laboratory. β cells were prepared using collagenase as described previously 39 with slight modifications and further digestion in enzyme-free cell-dissociation buffer (Invitrogen).

In vitro suppression assay

CD4+CD25+ and CD4+CD25 T cells were sorted as described above from pooled spleen cells of three 8-wk-old BDC2.5 ICOS+/+ or BDC2.5 ICOS−/− NOD mice. CD4+CD25 Teff cells from BDC2.5 ICOS+/+ NOD mice were labeled by incubation with 5 μM CFSE (Molecular Probes, Invitrogen) for 15 min at 37°C. Cells were then incubated in triplicate wells in 96-plates at 5×104/well with different ratios of unlabeled CD4+CD25+ Treg from BDC2.5 ICOS+/+ or BDC2.5 ICOS−/− NOD mice. Cells were stimulated in the presence of anti-CD3 (1 μg/mL, eBioscience) or RVRPLWVRME peptide (60 μg/mL) and 1×105 irradiated CD4+ and CD8+-depleted splenocytes from BDC2.5 ICOS−/− NOD mice. Proliferation of CFSE-labeled CD4+CD25 Teff cells was assessed after 72h by flow cytometry.

Adoptive cell transfers

CD4+CD25 Teff cells, purified T cells including CD4+ T cells and CD8+ T cells, splenocytes depleted of CD4+, and CD8+ T cells from 8, 14, or 40-wk-old either ICOS+/+, ICOS−/−, BDC2.5 ICOS+/+, BDC2.5 ICOS−/− NOD mice were sorted as described above. When indicated, cells were stimulated with 1 μg/mL anti-CD3 (BD Bioscience) and anti-CD28 (BD Bioscience) for 48 h. IL2 (20 U/mL, R&D Systems, Minneapolis, MN) was added to the culture of the splenocytes depleted of CD4+ and CD8+ T cells. Total spleen cells or purified T cells, stimulated or not, were then adoptively transferred as indicated in figure legends i.v. into 8-wk-old ICOSL+/+ or ICOSL−/− NOD.scid mice as described previously 40.

Ten days after the transfer, mice were evaluated three times a week, for clinical T1D or neuropathy (consisting in hind legs weakness, ataxia) during 15 wk or sacrificed 30 days after transfer, perfused, and fixed by 4% formaldehyde solution (Electron Microscopy Science) in order to perform histological analysis.

Statistical analysis

Appropriate statistical tests were performed on all data as indicated in figure legends using GraphPad Prism Version 4.0b software (GraphPad Software, La Jola, CA).

Acknowledgements

A. T. was supported by an “Avenir” INSERM contract. E. R. was supported by an FRM contract. C. B. was supported by INSERM contract.

Conflict of interest: The authors declare no financial or commercial conflict of interest.

Ancillary