Autoreactive CD8⁺ T cells in multiple sclerosis: a new target for therapy?

Abstract

Multiple sclerosis afflicts more than 1 million individuals worldwide and is widely considered to be an autoimmune disease. Traditionally, CD4⁺ T helper cells have almost exclusively been held responsible for its immunopathogenesis, partly because certain MHC class II alleles clearly predispose for developing multiple sclerosis and also, because of their importance in inducing experimental autoimmune encephalomyelitis (EAE), the animal model for multiple sclerosis. However, several strategies that target CD4⁺ T cells beneficially in EAE have failed to ameliorate disease activity in multiple sclerosis, and some have even triggered exacerbations. Recently, the potential importance of CD8⁺ T cells has begun to emerge. Physiologically, CD8⁺ T cells are essential for detecting and eliminating abnormal cells, whether infected or neoplastic. In multiple sclerosis, genetic associations with MHC class I alleles have now been established, and CD8⁺ as well as CD4⁺ T cells have been found to invade and clonally expand in inflammatory central nervous system plaques. Recent animal models induced by CD8⁺ T cells show interesting similarities to multiple sclerosis, in particular, in lesion distribution (more inflammation in the brain relative to the spinal cord), although not all of the features of the human disease are recapitulated. Here we outline the arguments for a possible role for CD8⁺ T cells, a lymphocyte subset that has long been underrated in multiple sclerosis and should now be considered in new therapeutic approaches.

multiple sclerosis, experimental autoimmune encephalomyelitis, T cells, CD8, autoimmunity

APC = antigen-presenting cells, β2m = β-2-microglobulin, CNS = central nervous system, CSF = cerebrospinal fluid, DC = dendritic cell, EAE = experimental autoimmune encephalomyelitis, EBV = Epstein–Barr virus, GA = glatiramer acetate, HLA = human leukocyte antigen, IFN-γ = interferon-γ, IL = interleukin, LN = lymph node, MBP = myelin basic protein, MAG = myelin associated glycoprotein, MHC = major histocompatibility complex, MOG = myelin oligodendrocyte glycoprotein, MRI = magnetic resonance image(s), PLP = proteolipid protein, PPMS = primary progressive multiple sclerosis, RRMS = relapsing–remitting multiple sclerosis, SPMS = secondary progressive multiple sclerosis, TCR = T cell receptor, T_H = CD4⁺ T helper cell, TNF-α = tumour necrosis factor-α

Introduction

Multiple sclerosis is the commonest autoimmune disorder of the central nervous system (CNS), afflicting about 400 000 Europeans and 350 000 people in North America (Sospedra and Martin, 2005; Steinman, 1996). Women are affected more often than men (∼1.6 : 1), and multiple sclerosis usually strikes between puberty and the menopause, as in many other autoimmune disorders (Keegan and Noseworthy, 2002). So far, no curative treatment is available. Approximately 85% of patients present with a relapsing–remitting multiple sclerosis disease course (RRMS), with discrete attacks (exacerbations) of neurological deficits, interrupted by periods of relative clinical stability; after ∼10 years, a secondary progressive multiple sclerosis phase (SPMS) ensues, with persistent and advancing deficits in most cases. Another 10–15% of usually somewhat older patients show an insidious onset and a primary progressive multiple sclerosis course (PPMS) with few, if any, exacerbations and no remissions (Thompson et al., 1997). In general, symptoms typically include hyperreflexia, ataxia, spasticity, visual and sensory impairment, bladder and bowel disturbances and fatigue (Noseworthy et al., 2000).

The cause of multiple sclerosis is still unknown and disease pathways are poorly understood: focal white matter infiltration by T and B lymphocytes and plasma cells in the brain and spinal cord strongly suggests an autoimmune pathogenesis (Traugott et al., 1983; Hauser et al., 1986; Hohlfeld and Wekerle, 2001). It is generally believed that T lymphocytes react against myelin components and activate microglia and macrophages, leading to damaged myelin sheaths with impaired nerve conduction (Hohlfeld and Wekerle, 2001; Steinman et al., 2002). Oligoclonally expanded T cells are found in multiple sclerosis lesions, but their actions and primary target antigens are still largely unknown (Oksenberg et al., 1990; Babbe et al., 2000). Moreover, we still do not know what factors trigger the autoreactive T cells to migrate into the CNS and mediate disease, though environmental factors, such as infections have long been suspected (Sibley et al., 1985; Marrie, 2004).

Although most T cell responses are beneficial, some have the potential to cause immunopathology in certain situations, and need to be restrained by protective and regulatory processes. T lymphocytes can be divided into CD4⁺ and CD8⁺ expressing T cells. In general, CD4⁺ T helper cells (T_H) recognize peptides that are presented by major histocompatibility complex (MHC) class II molecules on specialized antigen-presenting cells (APC) and are usually derived from antigens ingested by the APC. Once activated, naïve T_H cells differentiate into functional subsets; the so-called pro-inflammatory T_H1 cells which predominantly produce interferon (IFN)-γ and tumour necrosis factor (TNF)-α and protect against intracellular pathogens, whereas T_H2 cells produce interleukin (IL)-4, IL-5 and IL-13, and help to combat extracellular infections (Seder and Ahmed, 2003).

By contrast, CD8⁺ T cells mostly recognize peptides from endogenously synthesized antigens presented by MHC class I molecules which, in general, are expressed by all nucleated cells. They mediate effector functions through production of cytokines, such as IFN-γ and TNF-α, and/or by direct cytotoxicity. Most of our understanding about their function comes from analysis of viral infections in mice (Wong and Pamer, 2003). These typically lead to expansions of CD8⁺ T cell clones which eliminate infected cells (van Lier et al., 2003). After resolution of the infection, the clonal expansions contract through apoptosis. About 10% of the responding cells persist as memory cells (Murali-Krishna et al., 1998) and confer protection against reinfection, but can be deleterious if they cross-react with autoantigens. The initial priming of CD8⁺ T cells, the size of the expansions and the formation of memory cells are largely controlled by interactions with APCs (Murali-Krishna et al., 1998; Badovinac et al., 2002). However, it is still controversial whether T_H cell help is required for initiating CD8⁺ T cell responses (Bevan, 2004), although it clearly is essential for CD8⁺ memory generation (Sun et al., 2004). In addition, in certain viral infections (e.g. herpes simplex virus-1) and in situations with only mild inflammatory reactions (i.e. autoimmune responses), ‘licensing’ of APCs by T_H cells is needed to prime CD8⁺ T cell responses (Smith et al., 2004).

In multiple sclerosis, CD4⁺ T cells have long been presumed to play an almost exclusive role throughout the disease (Steinman, 1996), mainly because of strong evidence from genetic surveys in multiple sclerosis patients and cellular studies in animal models. First, the MHC class II allele human leukocyte antigen (HLA)-DR2 is an independent susceptibility factor for multiple sclerosis, conferring a relative risk of ∼4 (Olerup and Hillert, 1991). Furthermore, T_H cell receptors specific for myelin basic protein (MBP) have been found in multiple sclerosis brains (Oksenberg et al., 1993), and some multiple sclerosis patients show an immunodominant MBP peptide complexed with HLA-DR2 on APCs at sites of demyelination (Krogsgaard et al., 2000). Further evidence comes from the animal model for multiple sclerosis, experimental autoimmune encephalomyelitis (EAE), in which mice are immunized with whole spinal cord, myelin proteins or their encephalitogenic epitopes, usually in complete Freund's adjuvant. In some variants, T_H cells are clearly sufficient to induce EAE (Madsen et al., 1999), but microglia and macrophages also play key roles in the effector processes (Heppner et al., 2005). Starting in the early 1980s, pathogenic T_H cells from animals with EAE have been isolated and propagated as oligoclonal lines. They initiated disease after transfer into naïve healthy animals (Ben-Nun et al., 1981; Zamvil et al., 1985a, b), reinforcing the presumption that CD4⁺ T cells are almost exclusively responsible for initiating and maintaining the pathogenic immune response in multiple sclerosis.

This view was severely challenged when depletion of CD4⁺ T cells in multiple sclerosis patients caused no improvement in relapse rates or in inflammatory activity in magnetic resonance images (MRI) (van Oosten et al., 1997; Hohlfeld and Wiendl, 2001). Notably, however, global depletion of CD8⁺ as well as CD4⁺ T cells with an anti-CD52 monoclonal antibody (Campath-1H) led to reductions in relapses and new lesions, though with little improvement in long-term neurological deficits (Coles et al., 1999; Paolillo et al., 1999). Recently, therefore, CD8⁺ T cells have received increasing attention as possible effectors in multiple sclerosis (Steinman, 2001; Hemmer et al., 2002; Neumann et al., 2002; Lassmann and Ransohoff, 2004).

Here we review the evidence for contributions from CD8⁺ T cells in multiple sclerosis and EAE, and challenge the view that multiple sclerosis is an exclusively T_H cell-mediated disease. We propose a two-step model in which CD4⁺ T cells are important in initiating multiple sclerosis, whereas CD8⁺ T cells are the predominant cell type causing damage in the CNS during the relapses and possibly also in the chronic phase. Several prerequisites have to be met to establish a likely role for CD8⁺ T cells in the pathogenesis of multiple sclerosis; one of the first could be a genetic association with MHC class I alleles.

HLA class I associations in multiple sclerosis

The first studies on HLA associations in multiple sclerosis date from the early 1970s. They showed an association with the MHC class I antigen HLA-A3 (Naito et al., 1972), or possibly with the HLA-B7 antigen with which it was found to be in linkage disequilibrium (i.e. preferential co-inheritance of linked alleles, such as A3 and B7, en bloc) (Jersild et al., 1972). Afterwards, the MHC class II molecules HLA-DR15 and HLA-DR51 (encoded by HLA-DRB1*1501 and HLA-DRB5*0101 alleles in the HLA-DR2 haplotype) also became strong candidates because the linkage disequilibrium extends as far as the MHC class II regions in this multiple sclerosis-associated haplotype (Fogdell et al., 1995).

Recently, with more refined methods in larger series, these provisional results have been clarified. HLA-A3 (A*0301) roughly doubles the risk of developing multiple sclerosis, and does so independent of the HLA-DR2 alleles (DRB1*1501 and DRB5*0101) (Fogdell-Hahn et al., 2000; Harbo et al., 2004). Although both studies again show an association with HLA-B7 (B*0702), it is clearly secondary to the stronger association with the HLA-DR2 alleles owing to linkage disequilibrium and is independent of HLA-A3. As a result, subjects with both HLA-A*0301 and the HLA-DR2 alleles have a more than additive risk of developing multiple sclerosis (Table 1). In sharp contrast, the HLA-A2 allele (A*0201) confers some protection against multiple sclerosis, approximately halving the relative risk (Fogdell-Hahn et al., 2000; Harbo et al., 2004). Two other studies corroborate these findings; the first reported an MHC class I association independent of MHC class II, most likely with HLA-A*0301, and the risk conferred by it and the linked MHC class II alleles acted in concert (Rubio et al., 2002). In the second study, an unidentified HLA class I allele was associated with a protective effect on multiple sclerosis (Boon et al., 2001).

Table 1

HLA associations in multiple sclerosis (Fogdell-Hahn et al., 2000; Harbo et al., 2004)

HLA haplotype	Odds ratio
HLA-A*0201	0.52–0.7
HLA-A*0301	1.9–2.1
HLA-DR2, HLA-DQB1*06	2.9–3.6
HLA-DR2, HLA-DQB1*06 + HLA-A*0201	1.5
HLA-DR2, HLA-DQB1*06 + HLA-A*0301	5.2–6.8
HLA-A*0201 + HLA-A*0301	1.0

Thus, multiple sclerosis is positively and independently associated with HLA-A*0301 and HLA-DR2 (DRB1*1501, DRB5*0101). The relative risk increases profoundly in HLA-A3 and HLA-DR2 positive individuals. This association can be partly overridden by the protective allele HLA-A*0201. The earlier reported association with HLA-B7 is not a true association, but is attributable to a linkage disequilibrium with the HLA-DR2 alleles.

Extensive linkage disequilibrium within the MHC region makes it difficult to establish which allele within the HLA-DR2 haplotype confers the strongest susceptibility to multiple sclerosis. Thus, in most populations the DRB1*1501 and DRB5*0101 alleles are almost inseparable from a third MHC class II allele, DQB1*0602, that encodes the β-chain of the DQ6 molecule. However, in African Americans, the DRB1*1501 and DRB5*0101 alleles display a lower degree of linkage disequilibrium with the DQB1*0602 allele than in Caucasians. In an African American cohort DRB1*15 chromosomes carrying non-DQB1*0602 alleles are significantly increased in multiple sclerosis patients, whereas chromosomes carrying the DQB1*0602 allele with other DRB1 alleles (i.e. not DRB1*15 alleles) are not. This study indicates that DRB1*1501, rather than DQB1*0602, constitutes the principal multiple sclerosis disease-risk gene but did not address the role of the DRB5*0101 allele (Oksenberg et al., 2004). Additional studies on the role of MHC class I molecules in their association with multiple sclerosis in other ethnic groups would be desirable to undertake.

Furthermore, the influence of the MHC class I region in EAE has been addressed using rat models. In these studies, the MHC class I allele A^u exerted a protective influence, presumably by recruiting MBP-peptide specific, TGF-β producing CD8⁺ T cells (Mustafa et al., 1994; Issazadeh et al., 1997).

The pathogenetic relevance of this MHC class I association would be strengthened if these molecules proved to be expressed in the target tissue of the autoimmune attack, the CNS.

MHC class I expression in the brain

The CNS was long considered to be immune-privileged, protected against T cell-mediated damage partly by the absence or sparsity of MHC class I molecules (Joly et al., 1991; Rall et al., 1995). In fact, all glial and neuronal cells can express MHC class I after stimulation with appropriate cytokines in vitro, especially IFN-γ (Neumann et al., 2002). Furthermore, human oligodendrocytes expressing MHC class I molecules can be lysed effectively by blood-derived CD8⁺ T cells that are alloreactive or recognize MHC class I-presented peptides (Ruijs et al., 1990; Jurewicz et al., 1998). In normal brain, constitutive expression was found on endothelial cells, perivascular macrophages and some microglial cells (Hoftberger et al., 2004). In vivo models show MHC class I expression in areas of inflammatory activity (Neumann et al., 1995; Horwitz et al., 1999; Medana et al., 2000). Notably, MHC class I is clearly upregulated in multiple sclerosis lesions, primarily on endothelial and microglial cells, in several studies (Hayashi et al., 1988; Ransohoff and Estes, 1991; Gobin et al., 2001). A recent study on normal brain tissue and multiple sclerosis brains at various disease stages, using newly available antibodies and refined techniques, showed that all cell types in the human CNS can express MHC class I antigens at certain disease stages (Hoftberger et al., 2004). Peri-plaque white matter and inactive demyelinated plaques in multiple sclerosis brains showed upregulation of MHC class I mainly on microglial cells. In contrast, MHC class I expression in active lesions in acute or chronic multiple sclerosis was evident on all cell types including microglial cells, astrocytes, oligodendrocytes and neurons (Hoftberger et al., 2004). Therefore, all neuroectodermal cell types in the CNS are potential targets of a cytotoxic CD8⁺ T cell attack. However, electrically active and intact neurons usually suppress the induction of MHC class I expression in their glial environment and on their own surface (Neumann et al., 1998). Loss of integrity and electrical activity allows MHC class I induction by pro-inflammatory cytokines.

MHC class I molecules have also been linked with mechanisms that are not traditionally considered to be immunological (Corriveau et al., 1998). Recently, MHC class I molecules have been described to be important for retraction of synaptic connections between neurons that occur during development (Huh et al., 2000), and for synaptic plasticity and regeneration of neurons after axotomy (Oliveira et al., 2004). Therefore, MHC class I expression is upregulated in inflammatory processes and during neuronal damage and electrical breakdown. It serves the recognition of antigens by CD8⁺ T cells but is also important for the regeneration and integrity of synaptic connections. In multiple sclerosis, this might lead to a vicious circle resulting in chronic neuronal damage in the CNS, which favours CD8⁺ T cell responses. This, however, is still controversial, with some evidence that immune cell infiltrate in the CNS can be beneficial, supporting the regenerative process by secreting neurotrophic factors (Schwartz and Kipnis, 2001; Foote and Blakemore, 2005). Upregulation of MHC class I expression in multiple sclerosis does not necessarily lead to a CD8⁺ T cell response, as the CNS immune privilege cannot be attributed simply to a lack of MHC class I molecules on the surface of brain cells. Even neurons that overexpress MHC class I are resistant to CD8⁺ T cell-mediated lysis following viral infection, despite increased infiltration of the brain and enhanced viral clearance from MHC-overexpressing infected cells (Rall et al., 1995).

Indeed, CNS damage would occur only if CD8⁺ T cells are activated, clonally expanded and able to invade into the CNS.

Activation, invasion and clonal expansion of CD8⁺ T cells in multiple sclerosis brains

Although most immune cells are effectively separated from the CNS by the blood-brain barrier, activated T cells are allowed to patrol the brain parenchyma (Yeager et al., 2000). It is likely that many of the T lymphocytes in multiple sclerosis lesions are first activated in the periphery and then migrate into the CNS, where they become reactivated when they encounter their target antigens, expand clonally and cause cytotoxic damage (Kivisakk et al., 2004) (Fig. 1). Indeed, in an animal model, activated CD8⁺ T cells travelled freely into the brain, and specifically induced inflammation and tissue damage in the CNS without CD4⁺ T cell help (Cabarrocas et al., 2003). Where and when CD8⁺ T cells initially recognize MHC class I-self-antigen complexes in multiple sclerosis is poorly understood. As target cells (i.e. oligodendrocytes and neurons) lack the necessary co-stimulatory molecules, they are not able to prime CD8⁺ T cells and to generate CD8⁺ T cell memory (Odeberg et al., 2005). The first priming of CD8⁺ T cells can only be initiated by professional APCs, once they have migrated to peripheral lymph nodes (Seder and Ahmed, 2003). Therefore, anti-myelin and/or anti-neuronal cytotoxic CD8⁺ T cells must first be activated by professional APCs that express surface MHC class I-peptide complexes and co-stimulatory molecules.

Fig. 1

Activation, invasion and clonal expansion of CD8⁺ T cells in multiple sclerosis/EAE. Dendritic cells get activated by inflammatory processes (i.e. infections) and their MHC class II molecules present epitopes from autoantigens to CD4⁺ T cells. The primed CD4⁺ T cells license DC to activate CD8⁺ T cells by cross-presenting autoantigens, and these then differentiate into regulatory and cytotoxic T cells. Regulatory CD8⁺ T cells can suppress the immune response against self-antigens by producing IL-10 or by interacting with Qa-1/HLA-E on activated CD4⁺ T cells. When this mechanism is overrun in multiple sclerosis, activated autoreactive cytotoxic CD8⁺ T cells invade the CNS by means of PGSL-1, α₄β₁ integrin (VLA-4) and LFA-1 expression. There they encounter microglia/macrophages (MΦ) which express co-stimulatory molecules (CD80/86, CD40) and MHC class I-peptide complexes (pMHC I). By interacting with these molecules (CD28, CD154, TCR, respectively), CD8⁺ T cells reactivate and expand clonally inside the CNS, where they meet their target antigens presented by MHC class I-expressing oligodendrocytes and neurons. This leads ultimately to tissue damage and neurological deficits.

Classically, only antigens that are derived from proteolytic breakdown of endogenously synthesized proteins in dendritic cells (DCs) have access to the MHC class I antigen presentation pathway (Fig. 2). This view was problematic, indicating, for example, that pathogens that do not infect DCs are not able to elicit an immune response. In multiple sclerosis, this restriction would imply that CD8⁺ T cell-mediated autoimmunity to CNS antigens could develop only if CNS proteins were expressed within professional APCs. Yet, APCs do not seem to express CNS candidate antigens. Instead, DCs may acquire exogenous CNS antigens and present peptides derived from them with MHC class I molecules. Many experiments have shown that exogenous antigens can be transferred directly from antigen-expressing cells to APCs. This phenomenon is called cross-presentation and was long disregarded (Heath and Carbone, 2001; Ackerman and Cresswell, 2004). Indeed, cross-presentation of CNS antigens by brain-derived APCs is responsible for initial priming and retention of fully differentiated CD8⁺ T cells inside the CNS (Calzascia et al., 2003).

Fig. 2

Antigen presentation pathways for the MHC class I and class II molecules. MHC class I molecules present peptides via direct and cross-presentation. Directly presented peptides are derived from endogenously synthesized proteins of either self or pathogen origin. The proteasome breaks down these proteins into peptides and they are then transported by the transporter associated with antigen presentation (TAP) complex into the endoplasmic reticulum for loading onto newly synthesized MHC class I molecules. These pMHC I complexes are then transported via the Golgi to the cell surface to be presented to CD8⁺ T cells. This presentation pathway is used by every cell in the body. By contrast, DCs have the ability to cross-present phagocytosed antigens to CD8⁺ T cells. After endocytosis of exogenous antigens by DCs, they divert these antigens into the MHC class I-processing machinery. In most cases, these antigens will also be processed into the MHC class II presentation pathway for recognition by CD4⁺ T cells. MHC class II molecules present proteins that enter the cell through endocytosis. MHC class II molecules are prevented from binding to endogenous antigens in the endoplasmic reticulum by association with the invariant-chain (Ii). MHC class II complexes–invariant chain (MHC II–Ii) move through the Golgi where li is degraded to class II-associated invariant-chain peptide (CLIP). CLIP is then removed from the MHC class II–CLIP (MHC–CLIP) complexes and exchanged for antigenic peptide (pMHC II), which are transported to the cell surface and recognized by CD4⁺ T cells.

Related to cross-presentation is the question how the CNS antigens traffic from the brain and spinal cord to the lymph nodes (LNs), the site of T cell activation. There are two possible options: either antigens are already taken up in the CNS by APCs, that then migrate to the draining LNs, or they are transported passively in the lymph and are then endocytosed in the LNs by APCs. Despite the lack of lymphatic vessels in the CNS parenchyma, CNS fluids and intracerebrally injected antigens clearly drain along perivascular pathways in the meninges and brain parenchyma to cervical lymph nodes (Weller, 1998). Furthermore, the perivascular spaces surrounding cerebral and meningeal blood vessels contain a population of myeloid cells, which are thought to play an important role as APCs (Hickey and Kimura, 1988; Weller, 1998). They are necessary for initiation of EAE (Hickey and Kimura, 1988; Greter et al., 2005) and they are the only APCs in the CNS capable of efficiently stimulating naïve myelin-specific cells (McMahon et al., 2005). DCs which, in most tissues, endocytose, process and transport potential antigens to lymphoid organs have long gone unrecognized in the CNS. However, it has been shown that DCs are present in normal human meninges, choroid plexus and cerebrospinal fluid (CSF) and can be recruited and/or develop into DCs in injured or inflamed brain tissue (Pashenkov et al., 2001; Pashenkov and Link, 2002). Furthermore, DCs are found in the murine brain and can initiate a T cell-mediated immune response after migrating into cervical LN in an animal model (Serafini et al., 2000; Karman et al., 2004). Recently, it has been demonstrated that CNS-antigen-specific CD8⁺ T cells are primed in the cervical and lumbal LNs, where they acquire an activated phenotype which allows them to migrate specifically into the CNS (Calzascia et al., 2005).

Potential cross-presentation of autoantigens in peripheral LN does not automatically initiate an autoreactive CD8⁺ T cell response, but usually results in CD8⁺ T cell tolerance induction via deletion (Kurts et al., 1997). However, antigen-specific CD4⁺ T cell help can prevent deletion and allows expansion of self-reactive CD8⁺ T cell clones, favouring autoimmunity (Kurts et al., 1997). This has been shown for soluble and cell-associated self-antigens, especially, when APCs are activated by danger signals, e.g. infections, to trigger T_H1 responses (Waldner et al., 2004). Cross-priming is increased in the presence of autoreactive T_H cells and with increasing amounts of antigen in the target tissue. Immune complex-forming autoantibodies, high T cell precursor frequencies and the induction of apoptosis in target tissue further enhance the amount of cross-presented self-antigens (Heath and Carbone, 2001; Kita et al., 2002). Most of these conditions prevail in multiple sclerosis; autoreactive T_H cells have been reported (Oksenberg et al., 1993), autoantibodies have been detected by several groups (Berger et al., 2003; Lily et al., 2004) and apoptosis of neurons and oligodendrocytes is one feature of multiple sclerosis (Barnett and Prineas, 2004).

Molecular mimicry is a possible alternative route of an initial expansion of naïve autoreactive T cells. It proposes that microbial peptides with sufficient sequence or structural similarity to self-peptides can activate autoreactive T cells (Fujinami and Oldstone, 1985; Wucherpfennig and Strominger, 1995). The majority of investigations were done in animal models that examined the T cell receptor (TCR) cross-reactivity on CD4⁺ T cells (Lang et al., 2002). The role of CD8⁺ T cells in molecular mimicry in multiple sclerosis has not been investigated. However, the role of CD8⁺ T cells has been investigated in a mouse model of inflammatory bowel disease using a CD8⁺ T cell clone recognizing both mycobacterial and murine heat shock protein 60 (hsp60). Adoptive transfer of this T cell clone into TCRβ gene-deleted mice showed a massive infiltration into the small intestine and liver, resembling inflammatory bowel disease, whereas a non-cross-reactive T cell clone that only recognizes mycobacterial hsp60 does not cause disease (Steinhoff et al., 1999). Recently, the specificity of oligoclonal IgG antibodies in the CSF of multiple sclerosis patients has been mapped to two EBV proteins expressed in latently infected cells (Cepok et al., 2005). Interestingly, CD8⁺ T cell responses against EBV proteins were higher in multiple sclerosis patients compared with healthy individuals, whereas CD4⁺ T cells responses showed no difference (Cepok et al., 2005). This finding might indicate a prevalence of EBV-specific, cross-reactive CD8⁺ T cell clones in multiple sclerosis patients. However, one has to be very cautious with these kinds of findings, as it is a common phenomenon that CD8⁺ T cells which react against herpes viruses enrich in different chronic inflammatory lesions, including multiple sclerosis (Scotet et al., 1999). This might also be true for antibody productions.

After peripheral priming, CD8⁺ memory T cells can invade the CNS by several routes: from blood to CSF across the choroid plexus, from blood to the subarachnoid space through meningeal vessels or from blood to parenchymal perivascular spaces (Ransohoff et al., 2003). The first step in this process is the adhesion of lymphocytes to blood vessels (Weninger et al., 2002). This phenomenon has been compared in meningeal vessels of mice injected with blood lymphocytes from patients in acute multiple sclerosis relapses or from controls (Battistini et al., 2003). Interestingly, only the CD8⁺ T cells from these patients showed increased adhesiveness in this in vivo model, and it depended on the expression of P-selectin glycoprotein ligand-1 (PSGL-1) by these memory-effector T cells. By contrast, the CD4⁺ T cells instead upregulated vascular cell adhesion molecule-1 (VCAM-1) during acute attacks, but it did not increase their adherence to the murine brain venules. Although this model is xenogeneic, it does show that CD8⁺ T cells from relapsing multiple sclerosis can be selectively recruited into the brain. Notably, the receptor for PSGL-1 is P-selectin which is constitutively expressed in microvessels of the choroid plexus and may mediate T cell trafficking into the brain and CSF (Ransohoff et al., 2003). The preferential recruitment and accumulation of CD8⁺ T cells inside the CNS has also been reported in two animal models (Carson et al., 1999; Brabb et al., 2000). Even after using specific MHC class II-restricted antigens for T cell priming, a greater percentage of CD8⁺ T cells in comparison with CD4⁺ T cells entered the CNS (Carson et al., 1999). This might be attributed to the fact that the memory T cell population (CD44^high) in the periphery contains a higher percentage of CD8⁺ T cells which preferentially cross the blood–brain barrier (Hickey, 1999; Brabb et al., 2000). Selectins have been involved in the primary step of leucocyte migration and may play a particular role in early adhesion processes. In contrast, integrins are mostly involved in the secondary step of the adhesion process (Weninger et al., 2002). Expression of the integrin lymphocyte function associated antigen (LFA)-1 on CD4⁺ and CD8⁺ T cells has since proved to be related to the increase in T2 lesion load in MRI of multiple sclerosis patients and may, therefore, be involved in migration of pathogenic lymphocytes into the CNS (Eikelenboom et al., 2005). Furthermore, CD8⁺ T cells activated in the cervical or lumbar LNs were imprinted to specifically home into the CNS, most probably by expressing characteristic adhesion molecules, one of which is the α₄β₁ integrin (very late antigen-4, VLA-4) (Calzascia et al., 2005).

Having entered the CNS activated CD8⁺ T cells can recognize their antigen on MHC class I expressing cells; if so, they rapidly divide and expand clonally to elicit an orchestrated attack on epitope-bearing cells (Ramakrishna et al., 2004). This subsequent local reactivation of CNS antigen-specific memory T cells, and the consequent inflammation in EAE, apparently depend on the expression of the co-stimulatory molecules CD80 (B7-1) and CD86 (B7-2), as implied by the reduced inflammation and preferential meningeal localization of T cells in mice deficient for these molecules (Chang et al., 2003). Transgenic mice expressing the co-stimulatory ligand CD86 on microglia in the CNS spontaneously develop autoimmune demyelinating disease. Interestingly, the CNS infiltrate showed a predominance of CD8⁺ memory-effector T cells, indicating that myelin-specific autoreactive CD8⁺ T cells are prevalent in wild-type mice, which engage in productive antigen-recognition and are activated in situ by CD86-expressing microglial cells (Zehntner et al., 2003). Co-stimulatory signals from CD40 on CNS-resident APCs acting through CD154 (CD40L) on encephalitogenic T cells are also required for local reactivation and development and progression of EAE (Howard et al., 1999; Becher et al., 2001). Indeed, CD4⁺ and CD8⁺ T cells in the blood of patients with SPMS show constitutive expression of CD154, which is not seen in controls (Jensen et al., 2001). Any T cells that find their cognate antigen in the CNS accumulate and expand, whereas others that cannot recognize their target structures disappear (Flugel et al., 2001). Therefore, at least some of the T cells found in the CNS in multiple sclerosis are probably recognizing specific antigens and are thus implicated in pathogenesis.

A further key step is the production of pro-inflammatory cytokines and chemokines. Indeed, IFN-γ expressing CD8⁺ and CD4⁺ T cells were found to be elevated in peripheral blood and CSF of multiple sclerosis patients compared with normal controls (Wallstrom et al., 2000). Furthermore, in vitro activated CD8⁺ T cells from patients with SPMS showed significantly increased lymphotoxin secretion (Buckle et al., 2003), a product of activated CD8⁺ T cells that is found in active multiple sclerosis plaques and is toxic to oligodendrocytes in vitro (Selmaj et al., 1991a, b). In addition, the CCR5 and CXCR3 chemokine receptor expression on CD8⁺ T cells is best correlated with long-term lesion development in MRI analysis (Eikelenboom et al., 2002) and myelin-specific CD8⁺ T cells can be an important source for pro-inflammatory chemokines that could promote inflammatory responses in multiple sclerosis (Biddison et al., 1997).

A variety of histological evidence implicates CD8⁺ T cells in pathogenesis: although their frequencies are approximately half those of CD4⁺ T cells in normal blood, they exceed them in chronically inflamed multiple sclerosis plaques (Hauser et al., 1986) by 3–10-fold in regions of demyelination and axonal damage (Booss et al., 1983; Babbe et al., 2000). Neurons in the CNS have been regarded as resistant to cytotoxic T cell attack (Joly et al., 1991; Rall et al., 1995). However, several studies have shown that this can be overcome by treating them with IFN-γ, which induces MHC class I expression (Neumann et al., 1995; Rensing-Ehl et al., 1996; Neumann et al., 1997). In vitro, CD8⁺ T cells can form direct and stable adhesions with neurons when they recognize MHC class I-peptide complexes, leading to a selective transsection of neurites (Medana et al., 2001). CD8⁺ T cells were also observed in close proximity to oligodendrocytes and demyelinated axons in brain tissue from multiple sclerosis patients, towards which cytolytic granules were polarized (Neumann et al., 2002). In addition, axonal damage within multiple sclerosis lesions correlates with the number of CD8⁺ T cells (Bitsch et al., 2000; Kuhlmann et al., 2002) and macrophages/microglial cells (Ferguson et al., 1997; Trapp et al., 1998; Kornek et al., 2000), but not with other subsets of immune cells, such as T_H cells. Axonal damage is most prominent in acute multiple sclerosis lesions, when it correlates with the number of infiltrating CD8⁺ T cells, both of which decline later (Kuhlmann et al., 2002). These findings suggest a prominent role of this lymphocyte subset in the pathology of acute multiple sclerosis plaques.

Several studies report oligoclonal expansions in CD8⁺ T cells rather than CD4⁺ T cells in multiple sclerosis. Identified by their clonally specific TCRs or their V-gene usage in the TCR-β-chain (Vβ), these were confined to CD8⁺ T cells in two studies, and were striking in multiple sclerosis brains, blood (Babbe et al., 2000; Skulina et al., 2004) and CSF (Jacobsen et al., 2002). The same clones persisted for 2–5 years, again suggesting that they had expanded in response to cells expressing their target epitope(s) and to be of memory phenotype (Babbe et al., 2000; Skulina et al., 2004). A recent report on blood T cells in multiple sclerosis also showed a significantly skewed T cell repertoire; it was more pronounced in CD8⁺ T cells, which also showed increased expression of IFN-γ, IL-2 and TNF-α mRNAs (Laplaud et al., 2004).

As CD8⁺ T cells are the predominant cells to be clonally expanded in multiple sclerosis lesions, they most probably recognize local target structures and cause damage to CNS tissue. However, activated T cells, especially in the CD8⁺ T cell subset, often contain clonal expansions even in healthy individuals, although usually in older subjects (Hingorani et al., 1993; Posnett et al., 1994; Pannetier et al., 1995; Schwab et al., 1997). Therefore, a challenging question is which, if any, MHC class I-restricted antigens are recognized and attacked in the CNS.

CNS antigens and epitopes recognized by CD8⁺ T cells

Few studies have focused on the potential target epitopes and cell types recognized by infiltrating CD8⁺ T cells. Since multiple sclerosis affects central and not peripheral myelin, it seems logical to concentrate on CNS-specific antigens or exons, particularly those not expressed in the thymus, such as proteolipid protein (PLP) and myelin oligodendrocyte glycoprotein (MOG) (Klein et al., 2000; Bruno et al., 2002). Most of the epitopes of the few current candidates have been identified via prediction algorithms, and their actual recognition in vivo in multiple sclerosis lesions awaits confirmation. Indeed, most of the CD8⁺ T cell responses studied have been in the context of HLA-A2, partly because this allele is so common (Table 2). Although it appears to be protective, many of the HLA-A2-restricted CD8⁺ T cell responses so far studied appear to be cytotoxic. Responses to predicted epitopes in MBP, PLP and myelin-associated glycoprotein (MAG) have been screened in multiple sclerosis patients, but no healthy controls were tested initially. Cytotoxic T cell lines and clones were isolated against peptides MBP_110–118, PLP_80–88, MAG_287–295, MAG_509–517 and MAG_556–564, and they secreted IFN-γ and TNF-α after stimulation by phytohaemagglutinin, a lectin with unspecific mitogenic properties (Tsuchida et al., 1994). Notably, the MBP_110–118/HLA-A2-restricted epitope is clearly processed naturally as the corresponding T cell lines were able to kill HLA-A2 positive freshly isolated oligodendrocytes. Although multiple sclerosis patients and controls initially showed no difference in cytotoxic activity (Jurewicz et al., 1998), these CD8⁺ T cells were more frequently expanded in multiple sclerosis patients against the HLA-A2-restricted MBP_87–95 and MBP_111–119 epitopes. Again, clones from multiple sclerosis patients lysed MBP-transfected HLA-A2⁺ target cells and secreted IFN-γ and TNF-α (Zang et al., 2004).

Table 2

MHC class I immunodominant epitopes in humans and mice

Protein	Human	Mouse
MBP	HLA-A2: 87–95, 110–118	H-2Kk: 79–87
PLP	HLA-A2: 80–88 HLA-A3: 45–53	None known
MOG	None known	H-2Db: 37–46
MAG	HLA-A2: 287–295, 509–517, 556–564	None known

Other CD8⁺ T cell lines derived from multiple sclerosis patients recognized PLP_80–88/HLA-A2 or PLP_45–53/HLA-A3 and secreted chemokines, such as MIP-1α, MIP-1β, IL-16, IP-10 and matrix metalloproteinases (Biddison et al., 1997; Honma et al., 1997). However, the HLA-A2-restricted PLP_80–88 epitope triggered killing regardless of whether the donor was a multiple sclerosis patient or a healthy control (Dressel et al., 1997). A more comprehensive approach used peptide mixtures from several CNS proteins, namely MBP, PLP, MAG, MOG, S100β glycoprotein, oligodendrocyte-myelin glycoprotein, myelin-associated oligodendrocytic basic protein, αβ-crystallin and 2′-3′-cyclic nucleotide 3′-phosphodiesterase. Interestingly, there was no difference in the number of positive responses between multiple sclerosis and normal donors in the CD4⁺ T cell compartment, but the CD8⁺ T cells showed a higher response rate in the multiple sclerosis patients (Crawford et al., 2004), although the epitopes in these mixtures have not been defined yet.

The few identified immunodominant CD8⁺ T cell-restricted epitopes are also included in CD4⁺ T cell immunodominant epitopes. Human MBP_83–99 is presented by HLA-DRB1*1501, HLA-DRB5*0101, HLA-DRB1*0401 and others (Valli et al., 1993; Vogt et al., 1994; Wucherpfennig et al., 1994), whereas MBP_111–129 is presented by HLA-DRB1*0401 (Muraro et al., 1997). Furthermore, the PLP_41–60 peptide is immunodominant for HLA-DR2, HLA-DR3, HLA-DR4, HLA-DQ6 and HLA-DQ8 (Pelfrey et al., 1993; Mangalam et al., 2004). Only few MAG peptides have been examined for their immunodominance with an emphasis on the C-terminal end, i.e. MAG_596–612, MAG_609–626 (Andersson et al., 2002). The overlap of MHC class I and II immunodominant epitopes in multiple sclerosis might reflect preferential processing of exogenous proteins via the MHC class II and the cross-presenting MHC class I pathway.

So far, these studies show that CD8⁺ T cells can, in principle, respond to myelin proteins in multiple sclerosis patients, sometimes more than their T_H cells. More compelling evidence to incriminate them in pathogenesis might include differences in in vivo persistence, in activation state and/or cytokine profiles. What can we learn from the parallel studies in EAE models, e.g. about possible targets for clinical trials?

Contributions of CD8⁺ T cells in animal models

For decades, EAE has been used as a model of immunopathogenesis in multiple sclerosis. It depends on immunizing with myelin antigens in complete Freund's adjuvant; since it is not an ideal way of inducing CD8⁺ responses, there has been a heavy bias towards studying CD4⁺ T cells until recently. Clearly, myelin proteins such as MBP, PLP or MOG can be potent activators of encephalitogenic CD4⁺ T_H1 cells, and these have been isolated and propagated as oligoclonal lines (Steinman, 1999). Since these cloned T_H1 cells from animals with EAE could initiate disease after transfer into healthy animals, most of the research focused on them and ignored CD8⁺ T cells (Ben-Nun et al., 1981; Zamvil et al., 1985a, b). The widely accepted monopoly of CD4⁺ T cells in pathogenesis was further supported by evidence that CD8⁺ T cells in EAE were predominantly suppressive, with minimal effector activity (Jiang et al., 1992; Koh et al., 1992). Indeed, they clearly participate in vivo in the resistance to rechallenge that emerges as B10PL mice recover from a first episode of MBP-induced EAE. In vivo depletion with a monoclonal anti-CD8 antibody reversed this protection, and the mice developed clinical EAE when reimmunized with MBP. Therefore, CD8⁺ T cells obviously require priming during the first immunization to regulate CD4⁺ T cells triggered by the second MBP challenge in this particular in vivo model (Jiang et al., 1992). Moreover, deletion of the CD8 gene (CD8^−/−) on a similar H2^u background converts the course of the EAE from monophasic to chronic relapsing and remitting in the offspring of the CD8^−/− and PL/J mating, though with lower maximum severity (Koh et al., 1992). Thus, in this model, the CD8⁺ cells are mainly regulatory with few effector functions.

A more complex picture emerges from a recent study on MOG-induced EAE in CD4 gene-deleted (CD4^−/−) and CD8^−/− DBA/1 mice (Abdul-Majid et al., 2003). Either deletion substantially reduced EAE-susceptibility. Interestingly, however, this protection appeared to be greater after deletion of CD8. Furthermore, depletion of CD4⁺ cells from CD8^−/− mice protected of them from MOG-induced EAE, whereas 20% of CD4^−/− mice depleted of CD8⁺ cells developed EAE. In addition, the depletion of CD4⁺ but not CD8⁺ T cells protected DBA/1 wild-type mice from EAE, although demyelination was reduced on removal of CD8⁺ T cells. Evidently, both T cell subsets are important for the full disease severity in this particular model.

However, the above mentioned studies using CD4 and CD8 gene-deleted mice are very problematic and may be misleading. It now transpires that in CD4^−/− mice, the CD8 population makes significant responses to several MHC class II-restricted epitopes in addition to MHC class I-restricted epitopes (Tyznik et al., 2004). Many of the CD8⁺ T cells have apparently been selected by recognition of MHC class II in the thymus and may therefore substitute for the missing CD4⁺ T cells; hence, the results require cautious interpretation. One might expect to see the converse in CD8^−/− mice, but that has not yet been formally proven. Furthermore, expression of the CD4 molecule on CD8⁺ T lymphocytes after activation has been reported recently (Kitchen et al., 2004, 2005). The CD4 molecule modulates CD8⁺ T cell function and is critical in vivo for optimal responses against viral and cellular antigens. Therefore, a CD8⁺ T cell response is also compromised in CD4^−/− mice (Kitchen et al., 2005).

Although now doubted, these early studies nevertheless led to the neglect of CD8⁺ T cells in EAE at that time. More convincing evidence in their favour has recently emerged from two animal models in which EAE can be induced without any CD4⁺ T cell help by adoptively transferring CD8⁺ T cells (Huseby et al., 2001; Sun et al., 2001). Goverman's group (Huseby et al., 1999, 2001) reported that C3H mice immunized with MBP show a clonal expansion of CD8⁺ T cells specific for the MBP_79–87 peptide presented by the MHC class I molecule H-2K^k. The transferred cells induce severe disease in naïve recipients; moreover, it resembles multiple sclerosis more closely than does conventional EAE, showing ataxia, spasticity and a higher mortality. On histological analysis, the lesions were located in the brain more than in the spinal cord and showed severe demyelination and perivascular cell death. Co-injection of anti-IFN-γ antibodies markedly reduced disease severity, whereas blocking TNF with a TNF receptor-Fc fusion protein had no such effect (Huseby et al., 2001). The second study used a MOG_35–55 immunization protocol in C57BL/6 mice (Sun et al., 2001). After adoptive transfer of CD8⁺ T cells reactive to MOG_35–55/H-2D^b, disease was severe in wild-type recipients, but absent if they were β-2-microglobulin (β2m) gene-deleted (i.e. MHC class I-deficient). Again, the disease appeared more multiple sclerosis-like than in conventional EAE, with a relapsing and remitting course, and massive infiltration of CD8⁺ T cells and macrophages/microglial cells and demyelination, mainly in the brain rather than in the spinal cord. Interestingly, this MOG_35–55 sequence includes two closely overlapping MHC class II-(IA^b)-restricted MOG_40–48 (Mendel et al., 1996) and MHC class I-(H-2D^b)-restricted MOG_37–46 epitopes (Sun et al., 2003; Ford and Evavold, 2005). Although MOG_40–54 binds to H-2D^b and retains the full ability to induce EAE by activating CD8⁺ T cells in C57BL/6 mice (Sun et al., 2003), it still contains the full CD4⁺ T cell epitope (MOG_40–48), suggesting additional contributions of T_H cells to the resulting disease (Mendel et al., 1996). In further studies, the minimal epitope (MOG_37–46) that binds only to H-2D^b nevertheless induced EAE in C57BL/6 mice (Ford and Evavold, 2005). Labelling peptide:H-2D^b-tetramers identified MOG-specific CD8⁺ T cells early in the disease process, and detected them in the CNS prior to the onset of neurological signs. Furthermore, they secreted IFN-γ but no detectable TGF-β or IL-10, thus showing an effector phenotype, and they did not play a regulatory role in this MOG-induced model (Ford and Evavold, 2005). Nevertheless, regulatory CD8⁺ T cells have been observed in other EAE models.

Suppressive role of CD8⁺ T cells in multiple sclerosis/EAE

As discussed previously, it emerged in 1992 that regulatory or suppressive CD8⁺ T cells in EAE protect against subsequent relapses (Jiang et al., 1992; Koh et al., 1992). Important confirmatory evidence came from the work on CD8⁺ T cells isolated from EAE-recovered mice. These specifically inhibited some MBP-activated CD4⁺ T cell clones in vitro and their depletion was followed by recurrence of EAE. Moreover, they prevented disease in the recipients after transfer to MBP-immunized mice. Therefore, it seems likely that these CD8⁺ suppressor T cells have the ability to restrain the pathogenic MBP-reactive TCR repertoire that is preferentially activated by MBP in vivo (Jiang et al., 2003). Their suppressive actions proved to be restricted to the non-classical MHC class Ib molecule Qa-1, which is the murine homologue of the human HLA-E. It is expressed by some pathogenic CD4⁺ T cells; these proved to be an effective ‘vaccine’ for inducing regulatory CD8⁺ T cells in vivo that protect against EAE (Jiang et al., 1995, 2001, 2003). It is thought that these CD8⁺ regulatory T cells recognize a peptide from the autoreactive TCR expressed in CD4⁺ T cells, which is presented in the context of Qa-1 (Kumar, 2004). Autoreactive CD4⁺ T cells which express the Vβ8.2 TCR undergo apoptotic deletion after expansion of CD8⁺ regulatory T cells (Madakamutil et al., 2003). In further support, recently created Qa-1-deficient mice developed more severe EAE in response to an MHC class II-restricted PLP peptide, and failed to develop the normal resistance to relapses (Hu et al., 2004). Interestingly, in this latter study, the Qa-1-mediated protection was observed only during the secondary, but not the primary response.

Recently, β2m gene-deleted mice (i.e. MHC class I-deficient) were investigated for their disease course after EAE induction with MOG_35–55 or rat MBP (Linker et al., 2005). With both regimens disease course was more severe in β2m gene-deleted mice with axonal damage in immunized mice. These findings suggest a regulatory role for CD8⁺ T cells in this particular model. However, several issues have not been addressed in this study. First, these mice are also devoid of Qa-1 and CD1 expression, as MHC class Ib molecules and CD1 are stabilized by β2m. Secondly, it is now well established that MHC class I molecules are important for axonal regeneration and integrity (Oliveira et al., 2004) and thirdly, β2m-deficient mice suffer from an iron overload syndrome, which might make them more prone to CNS damage (Rothenberg and Voland, 1996). Therefore, it is likely that this report underlines a role for regulatory CD8⁺ T cells and the need for MHC class I expression for neuronal integrity.

Furthermore, a different study has identified a distinct population of antigen-specific CD8⁺CD28⁻ cells with the ability to suppress immune responses by directly interacting with APCs and rendering them tolerogenic (Najafian et al., 2003). With their reduced expression of co-stimulatory molecules, these tolerogenic APC induced antigen-specific unresponsiveness in CD4⁺ T cells, although their exact actions in vivo are not clear. Using the MOG_35–55-induced EAE model in C57BL/6 mice, a recent study showed that depleting CD8⁺ T cells (with monoclonal antibodies) enhanced EAE induction when given before immunization (Montero et al., 2004). CD8⁺ regulatory T cells can be induced by CD40L-activated plasmacytoid DCs (Gilliet and Liu, 2002) or by TGF-β-treated APCs (Faunce et al., 2004). Another novel mechanism is that certain T cells with the same specificity (for MBP_79–87) (apparently with a high affinity) induce unresponsiveness by removing these particular MHC class I–MBP complexes from the APC surface, making them unrecognizable by other specific T cells (Perchellet et al., 2004). However, it is still controversial whether antigen or autoreactive TCR peptide/Qa-1 complex specific recognition is predominant and a general mechanism which is also seen in other circumstances other than EAE. Furthermore, a wider mode of action is reported for CD8⁺ regulatory T cells by other studies with a production of the immunosuppressive cytokine IL-10 (Gilliet and Liu, 2002) or IFN-γ and IL-6 (Filaci et al., 2001). In contrast to CD4⁺ regulatory T cells which are now relatively well characterized in their expression of CD25 and Foxp3 (Sakaguchi, 2005) and which have been reported to show a decrease in their function from peripheral blood of patients with multiple sclerosis as compared with healthy donors (Viglietta et al., 2004), no distinct markers have been found for CD8⁺ regulatory T cells other than a lack of CD28 expression (Sun et al., 1988).

Finally, several studies in multiple sclerosis patients report defects in the suppressive CD8⁺ T cells that normally decrease IFN-γ production or proliferation by activated lymphocytes. Suppressor activity by CD8⁺ T cell lines against anti-CD3- or concavanlin A-activated heterologous lymphocytes was significantly lower in multiple sclerosis patients than in normal controls, whereas suppression by CD4⁺ T cell lines showed no difference (Antel et al., 1986a, b). Having shown that CD8⁺ T suppressor cells can normally be induced in vitro with IL-2, Balashov et al. (1995) found that their inhibitory properties were defective in chronic progressive multiple sclerosis patients compared with either healthy controls or RRMS patients. A standard treatment for multiple sclerosis is glatiramer acetate (GA, Copaxone) which reduces the relapse rate in RRMS patients by ∼30% (Hohlfeld and Wekerle, 2004). In untreated multiple sclerosis patients, in vitro responses of CD4⁺ T cells to GA are comparable with those in healthy controls, whereas those of CD8⁺ T cells are significantly reduced. However, they reportedly return to normal after GA treatment, in parallel with changes in the cytokine profiles of GA-specific CD4⁺ and CD8⁺ T cells, suggesting that their suppressive actions contribute to the therapeutic effects of GA (Karandikar et al., 2002).

Conclusion

To date, research in multiple sclerosis has focused on the pathogenetic role almost exclusively of CD4⁺ T cells. However, there is an increasing appreciation for the role of CD8⁺ T cells in multiple sclerosis, which might reflect the awareness for the complexity of this disease and/or the realization that multiple sclerosis consists of several different diseases with a similar clinical appearance. The arguments in favour of CD4⁺ T cells can now be used equally to support a role for CD8⁺ T cells (Table 3). Genetically, multiple sclerosis is associated not only with HLA class II, but also with HLA class I alleles. Both CD4⁺ and CD8⁺ T cells infiltrate and clonally expand in the CNS, where CD8⁺ T cells are overrepresented. Although the adoptive transfer of CD4⁺ or CD8⁺ T cells can induce EAE, the lesion distribution reproduces that in multiple sclerosis better in the CD8⁺ than in the CD4⁺ T cell models. This applies a fortiori to the involvement of the cytokines TNF-α and IFN-γ. Blockade of TNF-α usually ameliorates CD4⁺-mediated EAE (Steinman, 1999) but has little effect on the CD8⁺-induced models (Huseby et al., 2001). In multiple sclerosis, similar treatment with TNF receptor-Fc fusion protein actually worsened the disease and was discontinued (The Lenercept Multiple Sclerosis Study Group and The University of British Columbia Multiple Sclerosis/MRI Analysis Group, 1999). In addition, recombinant IFN-γ (rIFN-γ) is an efficient treatment for many CD4⁺-induced EAE models, which are often exacerbated by treatment with anti-IFN-γ antibodies (Steinman, 1999), whereas these are beneficial in the CD8⁺-induced EAE model (Huseby et al., 2001). The conclusions from the earlier CD4⁺-induced EAE models prompted a clinical trial of rIFN-γ therapy, but it had to be discontinued owing to the worsening of multiple sclerosis in some patients (Panitch et al., 1987). Thus CD8⁺-induced models reproduce the pathogenic responses and mechanisms in multiple sclerosis in this perspective better than the conventional CD4⁺-induced EAE models.

Table 3

Arguments for a possible rôle of CD8⁺ T cells in multiple sclerosis

Multiple sclerosis is positively associated with the MHC class I allele HLA-A*0301

All neuroectodermal cell types in active multiple sclerosis lesions express MHC class I and are therefore potential targets for cytotoxic CD8+ T cells

Migration of CD8+ T cells into the CNS is favoured over CD4+ T cells

Predominant clonal expansion of CD8+ T cells in the CNS, CSF and blood of multiple sclerosis patients

Overrepresentation of CD8+ T cells in multiple sclerosis plaques compared with blood levels, suggesting a specific enrichment of CD8+ T cells in multiple sclerosis brains

Autoreactive myelin-specific CD8+ T cells can be isolated from multiple sclerosis patients and healthy individuals

CD8+ T cells can initiate and contribute to disease pathology in some EAE models

CD8+ T cell-mediated animal models reflect cytokine effects (i.e. IFN-γ, TNF-α) as they are seen in multiple sclerosis patients

Obviously, depletion of CD4⁺ T cells with monoclonal antibodies is also beneficial in CD4⁺-mediated EAE models (Steinman, 1999); in human multiple sclerosis, there were promising early results in open studies (Lindsey et al., 1994a, b). However, there was no therapeutic effect in a carefully controlled phase II trial with the chimeric anti-CD4 mAb cM-T412 (van Oosten et al., 1997); despite significant and long-lasting depletion of CD4⁺ T cells, there was no difference in the number of active lesions on monthly gadolinium-enhanced MRI over 9 months, the primary measure of efficacy, although the authors conclude that a more aggressive depletion of CD4⁺ T cells might have led to a more substantial reduction in MRI activity (van Oosten et al., 1997). In contrast, depletion of both CD8⁺ and CD4⁺ T cells with an anti-CD52 monoclonal antibody (Campath-1H) led to major reductions in relapses and new lesions, although with little improvement in long-term neurological deficits (Coles et al., 1999; Paolillo et al., 1999). A CD8⁺-depleting antibody might be an interesting new option in multiple sclerosis patients, although it might pose the risk for opportunistic infections, which have recently been reported for treatments with the anti-α₄-integrin (VLA-4 and others) antibody natalizumab, which blocks the invasion of CD4⁺ and CD8⁺ T cells into the CNS and other organs (Singer, 2005). Another option might be an individualized treatment targeting the pathogenic T cell clones, which have expanded in each multiple sclerosis patient. However, it will pose a major challenge to set up an efficient screening method to quickly identify the epitopes which are recognized by individual multiple sclerosis patients' T cells.

Other trials have tried this approach by using altered peptide ligand (APL) analogous of the immunodominant HLA-DR2-restricted MBP_83–99 epitope; in vitro, they specifically inhibit T_H responses because of subtle changes in their TCR-contact residues. They also seemed very promising in vivo in mice, when given at the time of immunization with the (unchanged) encephalitogenic peptide, and they even reversed established paralytic disease (Karin et al., 1994; Gaur et al., 1997). However, in open or placebo-controlled studies (Bielekova et al., 2000; Kappos et al., 2000), there were no significant differences in clinical and MRI parameters versus either untreated patients (Kappos et al., 2000) or the patients' own previous baseline (Bielekova et al., 2000). Some patients even developed exacerbations, which were possibly linked to the APL treatment (Bielekova et al., 2000). Despite these results, APL remain an interesting therapeutic concept, which might give better results in future trials if the targeted epitopes are widened to several CD4- and CD8-restricted peptides.

Most studies in EAE focus on the acute phase, which is dominated by activation of macrophages and microglial cells by CD4⁺ rather than CD8⁺ T cells, and is largely independent of CD8⁺ T cells (Steinman, 1999). Thus, although CD4⁺-mediated EAE has taught us valuable lessons about T cell priming and infiltration into the CNS, CD8⁺-induced EAE may actually be an interesting model for the effector phase in human multiple sclerosis. Indeed, CD8⁺ T cells have long been known to be important in other autoimmune diseases, including type I diabetes mellitus (Hanninen et al., 1992; DiLorenzo and Serreze, 2005), rheumatoid arthritis (Kang et al., 2002), autoimmune thyroiditis (Sugihara et al., 1995; Stassi and De Maria, 2002), vitiligo (Ogg et al., 1998) and inflammatory bowel disease (Steinhoff et al., 1999; Bouma and Strober, 2003). Even if CD8⁺ T cells are the main effectors in multiple sclerosis, CD4⁺ T cells will still be very important for disease induction and maintenance; in viral infections, we know that they are required early during primary antigen exposure to promote both the differentiation of CD8⁺ effector cells into long-lived memory cells and their subsequent maintenance (Shedlock and Shen, 2003; Sun and Bevan, 2003). By analogy, multiple sclerosis might still be initiated by molecular mimicry of T_H epitopes from myelin proteins by microbial sequences (Oldstone, 1987; Wucherpfennig and Strominger, 1995; Lang et al., 2002), since that can also lead to ‘licensing’ of APC to activate autoantigen-reactive CD8⁺ T cells (Smith et al., 2004).

As in the CD4⁺ population, some CD8⁺ T cells can be protective (Jiang and Chess, 2004). If multiple sclerosis/EAE is initiated by cross-reactive microbial epitope(s), presented by MHC class II molecules on APCs, the responding T_H may start to express Qa-1/HLA-E-–self-peptide complexes (Fig. 1). In turn, they could activate regulatory CD8⁺ T cells that suppress any CD4⁺ T cells expressing the same Qa-1/HLA-E-self-peptide complex. Since expression of Qa-1 is restricted to specifically activated CD4⁺ T cells, the peptides it presents will probably include those induced by antigen activation or a peptide of the TCR itself (Kumar, 2004). Although known to prevail in some monophasic CD4⁺-induced EAE models, this mechanism may be defective in multiple sclerosis—leading to the generation of autoreactive CD8⁺ T cells (Balashov et al., 1995; Jiang and Chess, 2004). However, so far HLA-E has not yet been studied on human CD4⁺ T cells, and its action could be similar in multiple sclerosis. In order to therapeutically expand regulatory T cells, studies aiming at understanding the underlying mechanisms of regulatory CD8⁺ T cell induction in multiple sclerosis are needed, since this particular issue might be very different from the phenomena reported in EAE.

Further studies are clearly needed to address the role of CD8⁺ T cells in multiple sclerosis. They should aim at investigating the specificity of CD8⁺ T cells within brain lesions of multiple sclerosis patients and their pathogenic relevance. It is of importance to determine whether these T cells have a regulatory or a disease mediating phenotype. Clonally expanded CD8⁺ T cells in the blood, CSF and brain might be the key to a deeper understanding of their pathogenic relevance in multiple sclerosis. The autoantigens of these clonally expanded T cells need to be identified to be targeted specifically. However, as multiple sclerosis in its clinical course and its pathogenetic presentation is a very heterogeneous disease, target antigens are likely to be evenly diverse. Multiple sclerosis is now histologically divided into four subtypes with predominant lesions consistent of macrophage-mediated demyelination (type I), antibody-mediated demyelination (type II), distal oligodendrogliopathy (type III) and demyelination secondary to oligodendrocyte damage (type IV) (Lucchinetti et al., 2000), one of which might reflect a predominant CD4⁺ T cell-mediated disease whereas a different pattern might be caused by a predominant CD8⁺ T cell attack. Although many approaches have been taken, none has, so far, led to an identification of the antigen repertoire recognized in multiple sclerosis (Hohlfeld and Wekerle, 2004). This is, however, not surprising in the context of the heterogeneity. As many multiple sclerosis patients have silent lesions already before their clinical symptoms, a phenomenon called epitope spreading (Vanderlugt and Miller, 2002) (i.e. a spreading of autoreactive responses to different antigens within the same molecule or to antigens within other molecules) is likely to occur very early in the disease course, making it even more difficult to recognize single target antigens in multiple sclerosis. Therefore, there is a pressing need for more suitable models for studying the chronic phase, the better to mimic ongoing multiple sclerosis.

In conclusion, it is most likely that CD4⁺ T cells and CD8⁺ T cells are equally important in multiple sclerosis. However, both cell types have only been addressed in a reductionist approach by studying one or the other. Integrating the roles of both T cell subtypes will be an important new goal in understanding the pathogenesis of multiple sclerosis and in devising more suitable therapies.

We thank Nick Willcox for a critical reading of the manuscript and valuable suggestions. The authors' work is supported by the Medical Research Council (UK). M.A.F. is supported by the Deutsche Forschungsgemeinschaft (DFG FR1720/1-1).