Main

Multiple sclerosis is an inflammatory, demyelinating disease of the central nervous system (CNS) that affects more than one million people worldwide. It is believed to be an autoimmune disease in which exposure of genetically predisposed people to environmental factors triggers a breakdown in T cell tolerance to myelin antigens. The specific types of myelin-specific T cells that contribute to the pathogenesis of multiple sclerosis are not known. Most studies have focused on the pathogenic role of myelin-specific CD4+ T cells because of the relatively strong association of susceptibility to multiple sclerosis with major histocompatibility complex (MHC) class II alleles. In addition, CD4+ T cells are the main effector T cells in experimental autoimmune encephalomyelitis (EAE), a widely used animal model of multiple sclerosis. However, there is increasing recognition of the potential importance of CD8+ T cells in the pathogenesis of multiple sclerosis. CD8+ T cells typically outnumber CD4+ T cells in acute and chronic lesions of patients with multiple sclerosis, and the CD8+ T cell subset shows more evidence of antigen-driven activation than do CD4+ T cells in the CNS and blood of patients with multiple sclerosis1,2. The frequency of neuroantigen-specific CD8+ T cells, but not CD4+ T cells, in the CNS is also higher in patients with multiple sclerosis than in healthy controls3. Furthermore, depletion of CD4+ T cells is not beneficial to patients with multiple sclerosis, whereas depletion of a broader spectrum of leukocytes, including both CD4+ T cells and CD8+ T cells, diminishes lesion formation and relapses4. Together these observations support the idea of a role for both CD4+ myelin-specific T cells and CD8+ myelin-specific T cells in the pathogenesis of multiple sclerosis.

The conditions that lead to a loss of tolerance in either myelin-specific CD4+ or CD8+ T cells are not known. Genome-wide association studies have identified many multiple sclerosis–susceptibility alleles, each of which (apart from HLA-DR2) seems to contribute only slightly to the risk of developing multiple sclerosis5. This genetic complexity, together with the variability in the pathology, symptoms and clinical course of multiple sclerosis, suggest the possibility of multiple disease-initiating pathways. Disease heterogeneity may account for the difficulty in identifying environmental triggers of multiple sclerosis. Although viral infections have long been proposed to initiate the disease process6,7,8,9, linkage of a particular virus to the pathogenesis of multiple sclerosis has not yet been achieved. Association with a specific infection is particularly difficult for a multifactorial disease such as multiple sclerosis because a ubiquitous infection may trigger disease in only a small fraction of infected people depending on the diverse interactions of their particular susceptibility alleles with the environment.

Few animal models exist in which infectious triggers of CNS autoimmunity can be investigated. Infection with Theilers murine encephalomyelitis virus (TMEV), another model of multiple sclerosis, has been shown to induce CNS autoimmunity by causing bystander activation of myelin-specific CD4+ T cells10. However, no model has been described in which an infectious agent abrogates tolerance in myelin-specific CD8+ T cells. Here we used a MHC class I–restricted TCR-transgenic model that generates CD8+ T cells specific for myelin basic protein (MBP) to investigate conditions that break CD8+ T cell tolerance and induce CNS autoimmunity. Two TCR-transgenic models expressing distinct TCRs specific for residues 79–87 of MBP (MBP(79–87)) presented by the H-2Kk MHC molecule have been generated11. Mice expressing a transgenic TCR composed of α-chain variable region 8 (Vα8) and Vβ6 (8.6 mice) show both central and peripheral tolerance, consistent with the constitutive presentation of MBP in lymphoid and other tissues. In contrast, T cells expressing a transgenic TCR composed of Vα8 and Vβ8 (8.8 mice) escape central and peripheral tolerance, although they proliferate vigorously in response to MBP(79–87) peptide in vitro. This split tolerance has also been observed for several MHC class II–restricted MBP-specific TCR-transgenic models in which the low avidity of the interactions between the MBP-specific T cells and their ligand prevents responses to endogenous MBP in vivo, although the T cells can respond to MBP peptide in vitro. Tolerance in MHC class II–restricted TCR-transgenic models is broken at some stochastic frequency, as spontaneous EAE can occur, especially in the absence of regulatory T cells12,13. Disease is also easily induced in MHC class II–restricted TCR-transgenic models by immunization protocols that induce CD4+ T cell–mediated EAE in wild-type mice. The tolerance maintained by CD8+ T cells in 8.8 mice differs from that of MBP-specific CD4+ T cells in that 8.8 T cells show high avidity for their ligand and seem to remove the ligand from antigen-presenting cells (APCs) without triggering T cell activation11. This observation suggests that the circumstances that lead to loss of tolerance may differ for CD8+ MBP-specific T cells and CD4+ MBP-specific T cells.

Our studies here show that many conditions that induce disease in MHC class II–restricted myelin-specific TCR-transgenic models failed to break tolerance in CD8+ 8.8 T cells. CNS autoimmunity in 8.8 mice was triggered by infection with a recombinant vaccinia virus encoding MBP (Vac-MBP), consistent with the ability of viruses to efficiently prime MHC class I–restricted T cells. Unexpectedly, infection with wild-type vaccinia virus also triggered CNS autoimmunity as efficiently as infection with Vac-MBP despite a lack of cross-reactivity between the 8.8 TCR and viral epitopes. Disease induction by wild-type virus required the expression of endogenous TCR chains on the 8.8 T cells. Our results demonstrate a role for dual TCRs in the initiation of CNS autoimmune disease and suggest a previously unknown mechanism by which a ubiquitous viral infection may trigger disease in only a subset of infected people.

Results

Tolerance differs in CD8+ and CD4+ MBP-specific T cells

In contrast to MHC class II–restricted MBP-specific TCR-transgenic models, MBP-specific 8.8 mice did not develop spontaneous EAE, even on the recombination-activating gene 2–deficient (Rag2−/−) background (0 of 198 Rag2+/+ 8.8 mice and 0 of 24 Rag2−/− 8.8 mice observed for more than 12 weeks). This result indicated that regulatory T cells, which were absent from the Rag2−/− 8.8 mice (data not shown), were not required to prevent 8.8 T cells from responding to endogenous MBP. We investigated the susceptibility of 8.8 mice to active disease induction by using a protocol that efficiently induces EAE in MHC class II–restricted MBP-specific TCR-transgenic models and has also been shown to induce autoimmune disease in a published MHC class I–restricted TCR-transgenic humanized mouse model of multiple sclerosis in which the transgenic CD8+ T cells recognize a MHC class I–restricted epitope of proteolipid protein (PLP)14. We observed no neurological signs in 8.8 mice immunized with MBP(79–87) in complete Freund's adjuvant with or without injection of pertussis toxin (Supplementary Table 1). To assess whether peptide immunization is an efficient protocol for activating CD8+ 8.8 T cells, we labeled 8.8 and 8.6 T cells with the cytosolic dye CFSE, transferred them into Mbp−/− mice that had been previously immunized with MBP(79–87) in complete Freund's adjuvant and analyzed T cell proliferation 3 d later. Although both 8.8 and 8.6 T cells proliferated equally well after stimulation with MBP(79–87) in vitro, 8.8 T cells barely proliferated in vivo in response to adjuvant-activated APCs presenting exogenous MBP(79–87), whereas 8.6 T cells proliferated strongly (Supplementary Fig. 1). To determine if 8.8 T cell tolerance could be abrogated by strong, widespread activation of the APCs that present endogenous MBP throughout the animal, we administered lipopolysaccharide (LPS) and agonistic antibody to CD40 (anti-CD40) to 8.8 mice. Neither reagent, alone or in combination, induced disease in 8.8 mice. Likewise, we observed no disease in 8.8 mice treated with polyinosinic-polycytidylic acid (Supplementary Table 1). However, we observed weight loss and mild neurological signs in 8.8 mice when we injected MBP(79–87) simultaneously with both LPS and anti-CD40 (Supplementary Fig. 2 and Supplementary Table 1). Injection of MBP(79–87) alone had no effect. These results suggest that both strongly activating APCs in multiple tissues and increasing the concentration of ligand above the amount generated from endogenous MBP are needed to break 8.8 T cell tolerance in vivo.

Viral infection triggers autoimmunity in 8.8 mice

The conditions for breaking 8.8 T cell tolerance described above suggested that CD8+ T cell–mediated autoimmunity might be triggered by a viral infection that causes both widespread APC activation and generates de novo expression of a self antigen mimic. Consistent with that hypothesis, we found that 8.8 mice showed 100% incidence of autoimmune disease after infection with Vac-MBP. Unexpectedly, we also found that 8.8 mice infected with wild-type vaccinia virus showed the same incidence and severity of disease (Fig. 1a and Supplementary Table 1). In both groups of mice, the disease was characterized by weight loss and clinical signs such as ataxia, knuckling, difficulty walking and tail weakness. The clinical course of disease is presented here as increasing weight loss as this is the most quantitative measure of disease progression; however, immunochemical analysis demonstrated infiltration of CD8+ T cells and F4/80+ macrophages and activated microglia in both the brain and spinal cord of 8.8 mice infected with wild-type vaccinia virus, as expected for an autoimmune disease targeting the CNS (Supplementary Fig. 3). The disease progressed rapidly and we killed most mice 9 d after infection. In some experiments, mice survived this acute disease and developed chronic neurological symptoms such as walking difficulty and tail weakness (Supplementary Video 1). Analysis of viral titers in lymphoid organs and the CNS after infection showed that 8.8 mice cleared vaccinia virus as efficiently as wild-type mice did (data not shown), which indicated that the disease was not due to poor viral clearance in the TCR-transgenic mice. The idea of effective viral clearance in 8.8 TCR-transgenic mice was supported by the absence of clinical signs in Mbp−/− 8.8 mice infected with wild-type vaccinia virus (Fig. 1a and Supplementary Table 1). Infection with wild-type vaccinia virus activated a population of 8.8 T cells in vivo identified by lower fluorescence intensity of staining for MBP–H-2Kk tetramer and CD62L and higher CD44 expression than that of uninfected mice (Fig. 1b). These cells acquired effector function, as MBP-pulsed splenocytes were specifically lysed when transferred into vaccinia virus–infected 8.8 mice but not when transferred into naive 8.8 mice (Fig. 1c). We observed accumulation of 8.8 T cells in the CNS of vaccinia virus–infected Mbp+/+ 8.8 mice but not Mbp−/− 8.8 mice, and this correlated with neurological symptoms (Fig. 1d). These data indicate that infection with wild-type vaccinia virus breaks 8.8 T cell tolerance and promotes an autoimmune response directed against endogenous MBP in the CNS.

Figure 1: Infection with wild-type vaccinia virus induces autoimmune disease in 8.8 mice.
figure 1

(a) Weight change in Mbp+/+ 8.8, Mbp−/− 8.8 and Mbp+/+ wild-type (WT) mice (n = 7–11 mice per group) infected on day 0 with Vac-MBP or wild-type vaccinia virus (WT Vac), presented relative to weight on day 0. P < 0.001, Mbp+/+ 8.8 versus Mbp−/− 8.8, and P < 0.0001, Mbp+/+ 8.8 versus wild-type mice, after infection with wild-type vaccinia virus (Student's t-test). (b) Flow cytometry of splenocytes obtained from naive or infected 8.8 mice (7 d after infection with wild-type vaccinia virus), stained with MBP(79–87)–H-2Kk tetramer (MBP tetramer) and anti-CD8, anti-CD44 and anti-CD62L and gated on CD8+ cells. (c) Flow cytometry of CFSE-labeled cells from naive or infected 8.8 mice (7 d after infection with wild-type vaccinia virus), injected with equal numbers of wild-type splenocytes pulsed with MBP(79–87) (CSFEbright) and unpulsed splenocytes (CFSEdim) and analyzed 20 h later. (d) Flow cytometry of mononuclear CNS cells isolated from Mbp+/+ and Mbp−/− 8.8 mice 7 d after infection with wild-type vaccinia virus, then stained with MBP(79–87)–H-2Kk tetramer and anti-CD8. Numbers in quadrants (b,d) indicate percent cells in each; numbers above bracketed lines (c) indicate percent CFSEdim cells (left) or CSFEbright cells (right). Data were compiled from three independent experiments (a; mean ± s.e.m.) or are representative of two experiments (bd).

Viral infections have been proposed to trigger autoimmune disease via several mechanisms. Bystander activation of self-reactive T cells might occur if a viral infection caused the release of sequestered autoantigens into an inflammatory milieu. To test that possibility, we infected Mbp−/− 8.8 mice with wild-type vaccinia virus and analyzed T cells for activation markers and their ability to lyse MBP-pulsed splenocytes in vivo. As observed for Mbp+/+ 8.8 mice, infection with vaccinia virus induced a population of activated 8.8 T cells in Mbp−/− mice that specifically lysed MBP-pulsed target cells in vivo (Fig. 2). These results indicate that infection with vaccinia virus does not activate 8.8 T cells via a bystander mechanism, because the 8.8 T cells were activated in the absence of endogenous MBP.

Figure 2: Wild-type vaccinia virus does not stimulate 8.8 T cells via bystander activation.
figure 2

(a) Flow cytometry of splenocytes from Mbp−/− 8.8 mice, either naive or infected 7 d earlier with wild-type vaccinia virus, stained with MBP(79–87)–H-2Kk tetramer and anti-CD8, anti-CD44 and anti-CD62L and gated on CD8+ cells. Numbers in quadrants indicate percent cells in each. Data are representative of two experiments. (b) Flow cytometry of CFSE-labeled cells from naive Mbp−/− 8.8 mice and Mbp−/− 8.8 mice infected with wild-type vaccinia virus (7 d after infection), injected with equal numbers of wild-type splenocytes pulsed with MBP(79–87) (CFSE bright) and unpulsed splenocytes (CFSE dim) and analyzed 20 h later. Numbers above bracketed lines indicate percent CFSEdim cells (left) or CSFEbright cells (right). Data are representative of three experiments.

Molecular mimicry could also account for the ability of wild-type vaccinia virus to induce disease in 8.8 mice if the 8.8 TCR showed cross-reactivity to a viral antigen. In vitro experiments did not support this possibility, as 8.8 T cells proliferated in response to splenocytes infected with Vac-MBP but not those infected with wild-type vaccinia virus (data not shown). To explore this possibility in vivo, we transferred genetically marked, wild-type splenocytes into Rag2−/− 8.8 mice before infecting the mice with wild-type vaccinia virus to provide the B cells and nontransgenic T cells needed for viral clearance. At 7 d after infection, we stimulated splenocytes from the infected mice in vitro with vaccinia virus–infected target cells and analyzed T cell responses by intracellular staining for interferon-γ (IFN-γ; Fig. 3). We detected IFN-γ secretion only in the CD8+ T cell population derived from nontransgenic donor splenocytes and not in the host Rag2−/− 8.8 T cells, which demonstrated the inability of the 8.8 TCR to recognize viral antigens.

Figure 3: The 8.8 TCR is not cross-reactive to wild-type vaccinia virus epitopes.
figure 3

Flow cytometry of splenocytes from Thy-1.2+ Rag2−/− 8.8 mice given nontransgenic splenocytes (2.5 × 106) from Thy-1.1+ C3HeB/Fej mice and then infected with wild-type vaccinia virus 2 weeks later; cells collected 7 d after infection were stimulated in vitro for 18 h with either vaccinia virus–infected Thy-1.1+ splenocytes (Vac targets) or MBP(79–87)-pulsed Thy-1.1+ splenocytes (MBP targets) and stained with anti-Thy-1.2 and anti-CD8 (left), then were made permeable and stained with anti-IFN-γ (right). Numbers adjacent to outlined areas (left) indicate percent Thy-1.2+CD8+ cells (top) or Thy-1.2CD8+ cells (bottom); numbers in quadrants (right) indicate percent IFN-γ+CD8+ cells (top right) or IFN-γCD8+ cells (bottom right). Data are representative of two experiments.

The 8.8 T cells are activated via endogenous TCR chains

The lack of IFN-γ production by Rag2−/− 8.8 T cells in response to vaccinia virus–infected target cells suggested that Rag2−/− 8.8 T cells differed from Rag2+/+ 8.8 T cells in that they were not activated during infection with vaccinia virus. Therefore, we sought to determine if Rag2−/− 8.8 mice were susceptible to autoimmune disease induced by infection with wild-type vaccinia virus. We transferred wild-type splenocytes into Rag2+/+ and Rag2−/− 8.8 mice before infecting the recipients with wild-type vaccinia virus, and then monitored them for clinical signs. Although all of the control Rag2+/+ 8.8 mice developed autoimmune disease, none of the Rag2−/− 8.8 recipients developed disease (Fig. 4a and Supplementary Table 1). Consistent with the lack of disease induction, Rag2−/− 8.8 T cells did not acquire effector function as a result of infection with wild-type vaccinia virus, as MBP-pulsed splenocytes were lysed only after transfer into infected Rag2+/+ 8.8 recipients and not after transfer into infected Rag2−/− 8.8 recipients (Fig. 4b). This result, together with the fact that the Rag2−/− 8.8 mice contained nontransgenic T cells and B cells before infection, indicates the involvement of an intrinsic difference in Rag2−/− versus Rag2+/+ 8.8 T cells in conferring susceptibility to virus-induced autoimmunity.

Figure 4: Activation of Rag2+/+ 8.8 T cells by wild-type vaccinia virus requires expression of endogenous TCR chains.
figure 4

(a) Weight change in Rag2+/+ and Rag2−/− 8.8 mice (n = 5 mice per group) given splenocytes (2.5 × 106) from wild-type mice and then infected with wild-type vaccinia virus 2 weeks later. P < 0.0001 (Student's t-test). (b) Lysis of MBP peptide–pulsed splenocytes in vivo by Rag2+/+ and Rag2−/− 8.8 mice infected with wild-type vaccinia virus, assessed as described in Figure 1c. Data are representative of four experiments (mean ± s.e.m. in a).

Peripheral T cells in Rag2+/+ 8.8 mice are skewed toward the CD8+ subset; however, some CD4+ Vα8+Vβ8+ T cells also develop. In contrast, T cells in Rag2−/− 8.8 mice are all CD4CD8+. To determine if the inability of wild-type vaccinia virus to induce autoimmunity in Rag2−/− 8.8 mice was due to the loss of CD4+ 8.8 T cells, we treated Rag2+/+ 8.8 mice with depleting antibody to CD4 before infecting them with vaccinia virus. We found that 8.8 mice depleted of CD4+ T cells showed a delayed onset and somewhat milder disease, but the incidence of autoimmunity was not lower in mice lacking CD4+ T cells (Supplementary Fig. 4a and Supplementary Table 1). The lower disease severity in CD4+ T cell–depleted, vaccinia virus–infected 8.8 mice could have been due to a loss of pathogenic CD4+ 8.8 T cells or to a loss of help provided by CD4+ T cells to activated CD8+ 8.8 T cells. The latter possibility was supported by the finding that adoptive transfer of 8.8 T cells stimulated with MBP(79–87) in vitro induced autoimmunity only when interleukin 2 (IL-2) was administered after T cell transfer (data not shown). To determine if CD4+ 8.8 T cells were pathogenic, we purified CD4+ T cells and CD8+ T cells from Rag2+/+ 8.8 mice, stimulated them in vitro with MBP(79–87) and adoptively transferred them into naive recipients, accompanied by injections of IL-2. Although all recipients of CD8+ 8.8 T cells succumbed to autoimmunity, none of the recipients of CD4+ 8.8 T cells showed weight loss or clinical signs (Supplementary Fig. 4b). These results indicate that the ability of vaccinia virus to induce disease in Rag2+/+ 8.8 mice does not depend on the presence of CD4+ 8.8 T cells.

The other main difference between Rag2+/+ and Rag2−/− 8.8 T cells is the potential for Rag2+/+ 8.8 T cells to express endogenous TCR chains because of the incomplete allelic exclusion of rearrangements of Tcra and Tcrb. Indeed, expression of dual TCRs containing endogenous TCR β-chains paired with transgenic TCR α-chains or endogenous TCR α-chains paired with transgenic TCR β-chains has been observed on peripheral T cells in several TCR-transgenic models15,16,17,18. The possibility that expression of endogenous TCR chains on 8.8 T cells is required for susceptibility to wild-type virus–induced autoimmune disease suggested the hypothesis that vaccinia virus breaks 8.8 T cell tolerance by triggering T cell activation via a virus-specific TCR coexpressed with the MBP-specific TCR on 8.8 T cells. If this mechanism is correct, infected 8.8 mice should be more enriched for CD8+ T cells coexpressing the 8.8 TCR with particular endogenous TCR α- and/or β-chains that confer specificity to viral antigens than uninfected 8.8 mice. To test that hypothesis, we compared the expression of a panel of TCR Vβ chains on CD8+ T cells isolated from naive Rag2+/+ 8.8 mice and Rag2+/+ 8.8 mice infected with wild-type vaccinia virus. Although none of the antibodies specific for endogenous Vβ chains detected populations of more than 1% of CD8+ T cells in uninfected 8.8 mice (data not shown), T cell populations coexpressing Vβ8 and Vβ6 expanded after infection with vaccinia virus (Fig. 5a). The absolute number of Vβ8+Vβ6+ T cells in the spleen was over 40-fold more in infected 8.8 mice (n = 4) than in naive 8.8 mice (n = 8; P = 0.004). In contrast to the naive phenotype of Vβ8hiVβ6 T cells, Vβ8+Vβ6+ T cells had an activated phenotype in infected mice (Fig. 5b). To investigate the antigen specificity associated with Vβ6 expression, we isolated T cells from vaccinia virus–infected 8.8 mice, restimulated the cells in vitro with vaccinia virus–infected splenocytes or MBP-pulsed splenocytes and analyzed IFN-γ production and Vβ6 expression. More than 50% of the T cells that produced IFN-γ in response to splenocytes infected with wild-type vaccinia virus expressed Vβ6, whereas only 9.6% of the T cells that produced IFN-γ in response to MBP-pulsed splenocytes were Vβ6+ (Fig. 5c), which indicated that coexpression of Vβ6+ 'preferentially' conferred specificity for viral antigens. To determine if 8.8 T cell populations coexpressing particular Vα chains also expanded in response to infection with vaccinia virus, we sorted T cells with either a naive or activated phenotype from infected 8.8 mice and analyzed their expression of various Vα chains by real-time PCR. Activated T cells isolated from infected mice had enriched expression of Vα11, Vα13 and Vα14 compared with that of naive T cells isolated from infected mice (Supplementary Fig. 5). Together these data indicate that T cell populations coexpressing the 8.8 TCR and specific endogenous TCR chains are expanded in response to infection with vaccinia virus and that activation of these T cells correlated with loss of 8.8 T cell tolerance and the induction of autoimmune disease. This mechanism allows the prediction that other viral infections would trigger autoimmunity in 8.8 mice, which was supported by our finding that infection with adenovirus also induced autoimmunity in Rag2+/+ 8.8 mice (Supplementary Table 1).

Figure 5: Infection of 8.8 mice with wild-type vaccinia virus 'preferentially' expands CD8+ Vβ6+Vβ8+ T cell populations that respond to vaccinia virus epitopes.
figure 5

(a) Flow cytometry analysis of the expression of Vβ6 and Vβ8 by splenocytes collected from naive or infected Rag2+/+ 8.8 mice (7 d after infection with wild-type vaccinia virus), gated on CD8+ cells. Far left, staining of splenocytes from naive Rag2+/+ 8.8 mice with isotype-matched control antibodies for anti-Vβ6 and anti-Vβ8. (b) Expression of CD44 and CD62L by the CD8+ Vβ6Vβ8+ and CD8+ Vβ6+Vβ8+ cells identified in the vaccinia virus–infected mice in a. (c) Flow cytometry analysis of the expression of Vβ6 and IFN-γ by splenocytes obtained from the infected mice in a, cultured in vitro overnight with unmanipulated (Media), vaccinia virus–infected (Vac targets) or MBP peptide-pulsed (MBP targets) splenocytes from Thy-1.1 C3HeB/Fej mice, and gated on CD8+Thy-1.2+ cells. Numbers in quadrants indicate percent cells in each. Data are representative of five experiments with more than ten mice (a,b) or are representative of three experiments (c).

Vac-MBP activates 8.8 T cells via the MBP-specific TCR

Our experiments in which we administered MBP(79–87), LPS and anti-CD40 in vivo demonstrated that the 8.8 T cells can be activated via the MBP-specific TCR if the APCs are strongly activated and the dose of MBP increases above endogenous amounts. Therefore, we sought to determine whether infection with Vac-MBP could induce disease in Rag2−/− 8.8 mice. Rag2−/− 8.8 T cells proliferated in vivo in response to Vac-MBP-infected cells but not in response to cells infected with wild-type vaccinia virus (Fig. 6a). In contrast to our results obtained after infection with wild-type vaccinia virus, Rag2−/− 8.8 mice that received the wild-type splenocytes needed to clear the virus succumbed to autoimmune disease after infection with Vac-MBP (Fig. 6b and Supplementary Table 1). This result demonstrates that viral infection can break 8.8 T cell tolerance via signaling through the MBP-specific TCR if the infection causes both higher expression of MBP and widespread APC activation.

Figure 6: Vac-MBP activates Rag2−/− 8.8 T cells to induce autoimmunity but wild-type vaccinia virus does not.
figure 6

(a) CFSE dilution by CD8+Thy-1.2+Vα8+Vβ8+ cells among CFSE-labeled Rag2−/− 8.8 splenocytes (2 × 106) transferred into Mbp−/− mice that were left uninfected (gray lines) or infected 1 d earlier with either wild-type vaccinia virus or Vac-MBP (black lines); splenocytes collected 3 d later were stained with anti-CD8, anti-Thy-1.2, anti-Vα8 and anti-Vβ8. Data are representative of two experiments. (b) Weight change of Rag2+/+ and Rag2−/− 8.8 mice reconstituted with 2.5 × 106 splenocytes from naive wild-type mice and infected with Vac-MBP virus 2 weeks later (monitored as described in Fig. 1a). Numbers in parentheses represent disease incidence (diseased mice / total mice). Data are representative of two experiments (mean ± s.e.m.).

Discussion

In the present study, we have demonstrated that viral infection triggered CD8+ T cell–mediated autoimmune disease in the CNS by two distinct mechanisms. Infection with Vac-MBP induced disease via a 'molecular identity' mechanism in which the virus encoded an epitope recognized directly by the MBP-specific TCR. In contrast, infection with wild-type vaccinia virus broke tolerance in T cells that expressed dual TCRs due to incomplete allelic exclusion of the Tcra or Tcrb locus. Activation via a coexpressed virus-specific TCR overcame the lack of response of the MBP-specific 8.8 TCR to endogenous MBP such that the T cell was able to respond to both the viral epitope and endogenous MBP.

Viral infection has long been postulated to be an environmental factor that contributes to the etiology of multiple sclerosis. Although several different viruses have been linked to this over the years19,20, no specific virus has been confirmed as being a causative agent in the pathogenesis of multiple sclerosis. Because it is a multifactorial disease5, it is possible that multiple viruses influence susceptibility to multiple sclerosis, and the ability of any particular virus to contribute to the pathogenesis of this disease may be dependent on the repertoire of susceptibility alleles each person carries and their exposure to other predisposing environmental factors. Alternatively, multiple sclerosis may be triggered by a common infection that initiates disease in only a small fraction of infected people, as suggested by the geographical distribution of multiple sclerosis and the change in risk observed in migrants. In particular, a large body of evidence accumulated over the past two decades indicates that the human herpes virus Epstein-Barr virus (EBV) is a risk factor in the development of multiple sclerosis that operates independently of the risk contributed by the MHC DR15 allele20,21,22,23. Data from many studies have shown that the risk of developing multiple sclerosis is 15 times higher in EBV+ people than in EBV people and is two- to threefold higher in people with a history of infectious mononucleosis than in people who experienced asymptomatic infection24,25.

Animal models of multiple sclerosis induced by viral infection have identified some mechanisms by which viruses could trigger CNS autoimmunity. Infection with murine hepatitis virus induces a chronic, demyelinating disease that depends only on the activity of virus-specific T cells rather than on the emergence of myelin-specific T cells during the course of infection26. In contrast, TMEV induces CNS autoimmune disease in susceptible mouse strains via bystander activation of myelin antigen–specific CD4+ T cells. Bystander activation is facilitated by myelin damage that occurs during the initial clearance of virus by CD8+ T cells, which results in presentation of myelin epitopes by APCs to CD4+ myelin antigen–specific T cells that were nonspecifically recruited to the CNS. This phenomenon of epitope spreading from viral antigen–specific CD8+ T cells to self-reactive, myelin-specific CD4+ T cells results in a chronic disease that resembles multiple sclerosis27. A different mechanism has been demonstrated for recombinant TMEV encoding either a peptide derived from Haemophilus influenzae that shares six of 13 amino acids with a peptide from PLP(139–151) or a peptide derived from murine hepatitis virus sharing only three amino acids with PLP(139–151)28,2928,29. In these cases, the molecular mimicry between H. influenzae or murine hepatitis virus peptide and PLP peptide is sufficient to prime PLP(139–151)-specific CD4+ T cells, which then initiate chronic disease. Autoreactive CD8+ T cell clones have also been isolated from mice infected with the DA strain of TMEV, and these clones induce CNS pathology after adoptive transfer into uninfected mice30. However, the self antigen recognized by these CD8+ T cell clones has not been identified. Infection with Semliki Forest virus also induces an inflammatory, demyelinating disease in the CNS in which both virus-specific T cells and antibodies are generated that cross-react with myelin epitopes31.

In contrast to those models, our studies have demonstrated mechanisms by which viral infection broke tolerance directly in CD8+ MBP-specific T cells. Vac-MBP triggered autoimmunity in Rag2−/− 8.8 mice by a mechanism analogous to molecular mimicry in that an epitope encoded by the virus was specifically recognized by the MBP-specific TCR. The context of viral infection was important in breaking the tolerance normally maintained in vivo when the 8.8 TCR engages endogenous MBP ligand, as immunization with MBP(79–87) in complete Freund's adjuvant was not sufficient to induce disease in 8.8 mice. This result differs from findings in MHC class I–restricted PLP-specific TCR-transgenic mice in which neurological signs could be induced by immunization with PLP peptide14. The clinical signs induced by this immunization were very mild and CD4+ T cell activity was required for further disease progression and relapses. The differences in the requirement for CD4+ T cells and the severity of disease in that model relative to that of our model may reflect differences in the tolerance mechanisms that allow the CD8+ PLP-specific and MBP-specific 8.8 T cells to circulate in the periphery as naive T cells. Furthermore, CD4+ T cell help is not needed to activate naive CD8+ T cells during viral infection. The administration of LPS and agonistic anti-CD40 also did not induce disease in 8.8 mice, even though this method of activating APCs presenting endogenous MBP breaks tolerance in some CD4+ MBP-specific TCR-transgenic T cells32. Mild disease was induced by coadministration of MBP(79–87) with LPS and agonistic anti-CD40, which indicated that the epitope stripping that normally maintains 8.8 T cell tolerance can be overcome when there is both widespread activation of APCs and an increase in the concentration of the MBP ligand. These conditions were achieved more efficiently by infection with Vac-MBP, which accounted for the greater ability of Vac-MBP infection to induce disease in 8.8 mice than coadministration of MBP(79–87) with LPS plus anti-CD40.

Unexpectedly, we found that infection of Rag2+/+ 8.8 mice with wild-type vaccinia virus also induced autoimmunity. In contrast to infection with TMEV, infection with this virus does not induce CNS autoimmunity by bystander activation or by molecular mimicry. Notably, Rag2−/− 8.8 mice infected with wild-type vaccinia virus were not susceptible to autoimmune disease. The lack of disease in Rag2−/− 8.8 mice did not reflect a requirement for the CD4+ 8.8 T cells that are eliminated on the Rag2−/− background. Instead, our data have shown that expression of endogenous TCR chains on 8.8 T cells was required for wild-type vaccinia virus to induce disease in 8.8 mice. Consistent with the hypothesis that 8.8 T cells are activated via a viral antigen–specific TCR coexpressed with the 8.8 TCR, a population of T cells coexpressing the 8.8 TCR and Vβ6 expanded after vaccinia virus infection, and this population was specifically enriched for T cells that responded to viral antigens. CD4+ Vβ6+ T cell populations did not expand after infection of 8.8 mice (data not shown), which indicated that the virus does not function as a superantigen that activates T cells expressing particular Vβ chains independently of antigen specificity. After infection with vaccinia virus, the activated T cell population showed greater enrichment for T cells expressing Vα11, Vα13 and Vα14 than did the nonactivated T cell population. We conclude from these data that wild-type vaccinia virus breaks 8.8 T cell tolerance by triggering T cell activation via virus-specific TCRs that are coexpressed with the 8.8 MBP-specific TCR.

The ability to activate 8.8 T cells via a second TCR indicated that the effector functions of 8.8 T cells were intact despite their lack of response to endogenous MBP. Similarly, in mice engineered to express two transgenic TCRs, one of which induces strong anergy in vivo after interaction with a neo self antigen, stimulation via the second transgenic TCR activates the anergic T cells33. Our model differed from that in that the 8.8 T cells were specific for a true self antigen linked to the pathogenesis of multiple sclerosis and were not subjected to clonal deletion or anergy in vivo11.

Although our model used TCR-transgenic mice, T cells expressing dual TCRs exist in both mice and humans. Dual-TCR T cells usually express one TCR β-chain and two α-chains because allelic exclusion is less complete for the Tcra locus. In mice, the percentage of T cells reported to express two TCR α-chains varies from 2% to 15% (refs. 34,35,36), and a 33% frequency has been reported for humans37. The frequency of T cells expressing two TCR β-chains is estimated to be 1% in humans and 3% in mice, with the frequency increasing with age16,38,39. Although our data did not distinguish the specific pairing of TCR chains that compose virus-specific TCRs in 8.8 mice, it is more likely that a second TCR is generated in 8.8 T cells by pairing of an endogenous α-chain with the transgenic β-chain or pairing of an endogenous β-chain with the transgenic α-chain rather than via a failure of allelic exclusion at both the Tcra and Tcrb loci in individual transgenic T cells.

Peripheral T cells expressing dual TCRs have been shown to be beneficial by expanding the repertoire of T cells that respond to foreign antigens40. Since the first demonstration of T cells expressing two α-chains on the cell surface, however, most studies have focused on the hypothesis that expression of dual TCRs may promote autoimmunity and alloreactivity18,41. Support for the idea of a substantial contribution to alloreactivity has been provided by studies demonstrating that dual-TCR T cells have a dominant role in graft-versus-host disease42. In contrast, studies of animal models of autoimmune disease, including collagen-induced arthritis, EAE and diabetes, have failed to show any role for dual-TCR T cells in autoimmunity35,43,44. The frequency of spontaneous EAE in MHC class II–restricted MBP-specific TCR-transgenic mice increases when microbial exposure in the environment increases12,45; however, a role for dual TCRs responding to environmental antigens could not be established in the model in those studies12,45 because the loss of regulatory T cells on the Rag2−/− background considerably enhances the incidence of spontaneous EAE46. Thus, the studies reported here are the first to our knowledge to show a mechanism for triggering autoimmune disease that depends on the expression of dual TCRs.

Our findings suggest a new perspective on the proposed virus-induced etiology of multiple sclerosis that is consistent with the inability to detect infectious virus in the CNS. As the frequency of T cells coexpressing a myelin-specific TCR and a virus-specific TCR in the peripheral T cell repertoire should be low and probably varies among people, this mechanism may represent one way a common infection can trigger autoimmunity in a small subset of genetically predisposed people. The cumulative data on the connection between EBV infection and multiple sclerosis are consistent with this hypothesis. Despite the association of multiple sclerosis with higher serum titers of antibodies specific for EBV-encoded nuclear antigen 1 (ref. 29) and with a greater frequency of EBV-specific CD4+ and CD8+ T cells21, the evidence for an association of lytic EBV replication with multiple sclerosis is controversial. A published study has reported an incidence of almost 100% of EBV infection in CNS B cells from patients with multiple sclerosis, as well as viral reactivation in B cells present in CNS follicles accompanied by accumulation of activated CD8+ T cells47, which suggests that reactivation of EBV is a key factor in multiple sclerosis pathogenesis. However, the detection of EBV DNA in the CNS of patients with multiple sclerosis was not reproduced in another study using highly sensitive techniques48, which suggests that the function of EBV as a risk factor for multiple sclerosis is not dependent on reactivation of the virus during autoimmune disease. That idea is consistent with our conclusion that viral infection can activate T cells expressing dual TCRs, one specific for a viral epitope and one specific for a myelin epitope, which then drive the autoimmune process independently of an ongoing immune response to the pathogen. The low probability of this event may account not only for the low incidence of autoimmunity among people infected with one common pathogen but also for the higher antibody titers for several other common viruses in patients with multiple sclerosis21.

Methods

Mice.

MBP(79–87)-specific TCR-transgenic 8.8 and 8.6 mice on the Mbp+/+, Mbp−/− and Rag2−/− backgrounds have been described11. Thy-1.1 C3HeB/FeJ mice were generated by backcrossing of the allele encoding Thy-1.1 onto the C3HeB/FeJ background for 12 generations. All mice were bred and maintained in a specific pathogen–free facility at the University of Washington. Mice used for EAE induction were female mice between 8 and 12 weeks old. All procedures have been approved by the Institutional Animal Care and Use Committee at the University of Washington.

EAE induction by infection with vaccinia virus.

Wild-type vaccinia virus (New York City Board of Health) and Vac-MBP were obtained from Therion Biologics. Vaccinia virus was grown in HeLa human cervical cancer cells and was titered in BSC-40 monkey kidney cells. Mice were injected intraperitoneally with 1 × 106 to 5 × 106 plaque-forming units of wild-type vaccinia virus or Vac-MBP. Rag2−/− 8.8 mice were injected intravenously with 2.5 × 106 naive splenocytes from wild-type mice 2 weeks before being infected with virus. Mice were weighed daily and were killed when they lost more than 20% of their original body weight. Neurological symptoms, such as ataxia, knuckling, hypersensivity or difficulty in walking, usually appeared 6 d after infection.

Flow cytometry.

Splenocytes from naive mice or mice infected 7 d earlier with vaccinia virus, and CNS mononuclear cells, were isolated from perfused mice as described49 and were stained with various combinations of anti-CD8 (53-6.7), anti-Thy-1.2 (30-H12), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-Vβ2 (B20.6), anti-Vβ4 (KT4), anti-Vβ5 (MR9-4), anti-Vβ6 (RR4-7), anti-Vβ7 (TR310), anti-Vβ8 (MR5-2), anti-Vβ9 (MR10-2), anti-Vβ10 (B21.5), anti-Vβ11 (RR3-15) and anti-Vβ14 (14-2; all from BD Biosciences) or MBP(79–87)–H-2Kk tetramer (generated in-house with a construct encoding H-2Kk provided by the National Institutes of Allergy and Infectious Diseases Tetramer Core Facility and then conjugated to phycoerythrin). Cells were analyzed on a FACScan, FACScanto or LSR II (BD Biosciences).

In vivo cytotoxic T lymphocyte killing assay.

Splenocytes from wild-type mice were incubated for 1–2 h at 37 °C with or without 20 μM MBP(79–87) in RPMI-1640 medium (HyClone) supplemented with 2 mM L-glutamine (Invitrogen), 1 mM sodium pyruvate (Invitrogen), 0.1 mM nonessential amino acids (Invitrogen), 100 U/ml of penicillin (Invitrogen), 100 μg/ml of streptomycin (Invitrogen), 50 μM β-mercaptoethanol and 10% (vol/vol) heat-inactivated FBS (Hyclone). Splenocytes were then washed with PBS and incubated for 10 min at 37 °C in 5 μM (peptide-pulsed) or 1.2 μM (unpulsed) CFSE (carboxyfluorescein diacetate succinimidyl ester). After being washed three times with PBS, 1 × 107 peptide-pulsed and unpulsed splenocytes were mixed together and injected intravenously into mice. Mice were killed 20 h later. Spleens were collected and CFSE-labeled cells were analyzed by flow cytometry.

Intracellular IFN-γ staining.

Splenocytes (1 × 106) from infected mice were incubated overnight at 37 °C in 96-well U-shape plates with 1 × 106 naive Thy-1.1 splenocytes, naive Thy-1.1 splenocytes plus 5 μM MBP(79–87), or Thy-1.1 splenocytes infected with vaccinia virus, followed by additional culture for 5 h with GolgiPlug (1 μl/ml; BD Biosciences). Cells were stained for CD8, Thy-1.2, Vβ8 and Vβ6 (Fig. 5 only). Cells were fixed and made permeable (Cytofix/Cytoperm kit) and were subsequently stained with anti-IFN-γ (XMG1.2) or rat IgG1 isotype-matched antibody (R3-34; all from BD Biosciences). Samples were washed and then were fixed in PBS containing 1% (vol/vol) paraformaldehyde and analyzed on a FACSCalibur or FACSCanto (BD Biosciences).

Quantitative RT-PCR.

Total RNA was extracted from sorted cells with the RNeasy Mini kit (Qiagen), and first-strand cDNA was synthesized with SuperScript II according to the manufacturer's directions (Invitrogen). Quantitative PCR was done on an ABI 7300 Real Time PCR System (Applied Biosystems) with Power SYBR Green PCR Master Mix (Applied Biosystems). Primers for Tcra Vα and a primer for an α-chain constant segment have been described50. Reactions were run in duplicate, and results were normalized to those of the internal β-actin control.

Statistical analysis.

All P values were calculated with a two-tailed Student's t-test.