|Home | About | Journals | Submit | Contact Us | Français|
Immune-based self-recognition and failure to modulate this response are believed to contribute to the debilitating autoimmune pathology observed in multiple sclerosis (MS). Studies from its murine model, experimental autoimmune encephalomyelitis (EAE), have shown that neuroantigen-specific CD4+ T cells are capable of inducing disease, while their immune sibling, the CD8+ T cells, have largely been ignored. To understand their role in autoimmune demyelination, we first confirmed that, similar to our observations in human MS, there is robust induction of neuroantigen-reactive CD8+ T cells in several models, including MOG35–55/CFA-induced EAE. However, MOG35–55-specific CD8+ T-cells, when purified, were unable to adoptively transfer disease into naïve mice (in contrast to CD4+ T-cells). In fact, we observed that the transfer of these neuroantigen-specific CD8+ T cells was able to suppress the induction of EAE and to inhibit ongoing EAE. These regulatory CD8+ T cells produced IFN-γ and perforin and were able to kill MOG loaded CD4+ T-cells as well as CD4-depleted APC, suggesting a cytotoxic/suppressor mechanism. Inhibition of EAE was associated with both the modulation of APC function as well as decreased MOG-specific CD4+ T cell responses. Our studies reveal a novel and unexpected immune regulatory function for neuroantigen-specific CD8+ T cells and have interesting biologic and therapeutic implications.
Multiple sclerosis (MS) is an immune-mediated demyelinating disease of the central nervous system (CNS), characterized by an inflammatory infiltrate that includes both CD4+ and CD8+ T cells, among other mononuclear cells . Several studies from others and us provide evidence for the involvement of CD8+ T cells in the immune pathogenesis and/or regulation of MS [2–5]. In fact, CD8+ T cells comprise the majority of lymphocytes found in MS lesions with oligoclonal expansion, indicating their active participation in the lesion [4,6–12]. However, there is a relative paucity of knowledge regarding their functional role in MS, likely due to the focus on CD4+ T cells as the putative pathogenic and regulatory cells in MS and its animal model, experimental autoimmune encephalomyelitis.
EAE is a well-established animal model of MS that can be induced in several strains of mice by immunization with myelin-related antigens or by adoptive transfer of neuroantigen-specific CD4+ T cells. Prior EAE studies using CD8(−/−) mice, β2-microglobulin(−/−) mice or antibody depletion of CD8+ T cells in vivo are suggestive of both a pathogenic [13–17] as well as a regulatory role [16,18–20] for these CD8+ T cells. However, the antigenic specificity of immune regulatory CD8+ T cells remains somewhat of a mystery. In recent studies, CNS-specific CD8+ T cells were capable of inducing EAE in certain models, suggesting a pathogenic role for these CNS-targeted cells. For example, myelin basic protein (MBP)-specific CD8+ T cells in C3H mice  or myelin oligodendrocyte glycoprotein (MOG)35–55-specific CD8+ T cells in C57BL/6 (B6) mice could induce, augment or transfer disease in certain settings [14,17,21].
In this study, we attempt to further elucidate the role of myelin-specific CD8+ T cells in the context of wildtype EAE. We demonstrate the presence of neuroantigen-specific CD8+ T cell responses in several models of EAE. Importantly, using adoptive transfer approaches in wildtype immune-replete mice, we show that these cells do not display pathogenecity. Rather, we provide the first evidence that neuroantigen-reactive CD8+ T cells inhibit EAE, through suppression of CD4+ T cell and/or antigen-presenting cell function, mediated in part, by direct killing in an antigen-specific manner. These studies unveil an important and novel aspect of immune regulation during ongoing autoimmunity.
C57BL/6 (B6) female mice were purchased from Taconic (Hudson, NY) and the UT Southwestern Mouse Breeding Core Facility (Dallas, TX). SJL/J female mice were purchased from National Cancer Institute (Bethesda, MD). B10.PL mice were purchased from Taconic. B6.129 CD8−/− mice were purchased from Jackson Laboratory (Bar Harbor, ME). All mice were housed and bred in the UT Southwestern Medical Center Animal Resource Center and used according to approved IACUC protocols.
Six to 8 week-old C57BL/6 mice were immunized subcutaneously at two injection sites with 200 μg MOG35–55 (MEVGWYRSPFSRVVHLYRNGK, UT Southwestern Protein Chemistry Technology Center) or whole bovine myelin basic protein (MBP, Sigma) emulsified in CFA supplemented with 4 mg/ml Mycobacterium tuberculosis (MTB, H37Ra, Difco). SJL/J mice were immunized with proteolipid protein peptide 139–151 (PLP139–151, HSLGKWLGHPDKF) emulsified in CFA while B10.PL mice were immunized with MBP Ac1–11 (ASQKRPSQRSK) emulsified in CFA. In some experiments, EAE was induced in B6 mice using 200 μg PLP178–191 in CFA. Ovalbumin peptide 323–339 (OVA323–339, ISQAVHAAHAEINEAGR) was used as a peptide control. On days 0 and 2 post-immunization, 250 ng of pertussis toxin (PTX, List Biological Laboratories) was administered intraperitoneally in 100 μl of phosphate buffered saline (PBS). Disease severity was monitored daily and EAE clinical disease scored according to the following scale: 0, no paralysis; 1, loss of tail tonicity; 2, hind limb weakness; 3, partial hind limb paralysis; 4, complete hind limb paralysis; 5, hind limb paralysis and forelimb weakness/moribund state; 6, death. In all experiments with multiple groups, all groups were represented across multiple cages and all scoring was performed in a blinded manner. In mice that showed signs of EAE (clinical score ≥1), incidence, mortality, disease onset, time to peak, peak score, cumulative score (sum of all scores from disease onset to day 20), outcome (final score at day 20), and grade of remission (difference between peak score and outcome) data were recorded.
B6 mice were immunized with MOG35–55/CFA. Lymph node cells and/or splenocytes were harvested at either 10 or 20–25 days post-immunization. These cells were then incubated for 72 h at 37 °C in Dulbecco's Modified Eagle's Medium supplemented with 10% fetal calf serum (FCS), L-glutamine, penicillin, streptomycin, HEPES buffer, non-essential amino acids, sodium pyruvate and β-mercaptoethanol, stimulated with MOG35–55 (20 μg/ml) and mouse rIL-2 (10 pg/ml). Live cells were then separated using a Ficoll gradient. CD4+ and/or CD8+ cells (purity >90%) were positively selected using magnetic beads (Miltenyi Biotech, Germany). A total of 5–15 × 106 CD4+, CD8+ or bulk splenocytes were injected intravenously into naïve, wildtype B6 female mice and monitored for disease according to the previously outlined scale. To test for the ability of CD8+ T cells to transfer EAE, culture conditions were modified across several experiments by inclusion of murine IL-12, IL-23 or varying cell and stimulating peptide concentrations. In some experiments, 250 ng pertussis toxin was also administered intraperitoneally on days 0 and 2.
Female B6 mice were immunized with MOG35–55/CFA or OVA323–339/CFA. A total of 20–25 days post-immunization, splenocytes were stimulated with either MOG35–55 or OVA323–339 (20 μg/ml) respectively and murine rIL-2 (10 pg/ml) for 72 h at 37 °C at a concentration of 7.5 × 106 cells/ml. Live cells were then isolated on a Ficoll gradient. CD8+ cells in PBS were then injected into naïve, wildtype B6 via tail vein. The next day, these mice were immunized with MOG35–55/CFA for active EAE induction and monitored for disease.
Draining lymph node cells and/or splenocytes from immunized mice were harvested. CD8+ cells were magnetically separated using a negative selection protocol (Miltenyi Biotech, Germany) to recover “untouched” CD8+ cells (>90%). Cultures were set up in triplicates with either 4 × 105 bulk or 5 × 104 CD8+ cells per well. When necessary, irradiated splenocytes from naïve or immunized mice were used as APC at a ratio of 1:5 (CD8+ T cells:APC). After 72 h in culture, cells were pulsed with 0.5 μCi/well of 3H-thymidine for 18–20 h. Cells were then harvested on glass fibers mats and counted using a Betaplate counter (Wallac, Gaithersburg, MD).
Proliferation assays were performed using a carboxyfluorescein succinimidyl ester (CFSE)-dilution assay as described previously . Splenocytes were harvested and used either in bulk proliferation assays or as a source for CD8+ cells, which were isolated using a negative selection magnetic bead protocol. Bulk splenocytes or “untouched” CD8+ cells were suspended at 1 × 106 cells/ml in PBS and incubated at 37 °C for 7 min with 0.25 μM CFSE, followed by addition of serum and two PBS washes. Subsequently, CD8+ cells were suspended at 0.5–1 × 106/ml of media. Irradiated, CD8-depleted splenocytes (3500 rads) from mice immunized with OVA323–339/CFA or from naïve mice were used as APC at a ratio of 1:5. On day 5, cells were washed with staining buffer and labeled with phycoerythrin (PE)-conjugated-anti-CD8 and allophycocyanin (APC)-conjugated-anti-CD4 antibodies (Caltag/Invitrogen, Carlsbad, CA). After a short incubation at 4 °C, cells were washed and fixed in 1% paraformaldehyde (PFA, Electron Microscopy Sciences, Hattfield, PA). Flow cytometric data were acquired on a BD FACSort using BD CellQuest Software or BD LSR II flow cytometer using FACSDiva software. For analysis, FlowJo (Treestar, Ashland, OR) software was used to gate on lymphocytes and further on the CD4+ CD8− or CD8+ CD4− T cell subsets. The mean background proliferation was calculated based on the proliferating fractions (PF) in media alone. Proliferation was considered significant if the delta proliferation fraction (ΔPF = % sample proliferation – %mean background proliferation) was >1 and the stimulation index(SI = % sample proliferation ÷ % mean background proliferation) was >2
Splenocytes from naïve female B6 mice were harvested for use as target cells. These cells were stained with 0.25 μM CFSE (CFSE high) or 0.05 μM CFSE (CFSE low) and stimulated overnight with Con A (10 μg/ml), with concomitant loading with either MOG35–55 or OVA323–339 (20 μg/ml). In some experiments, an additional dye (CMTPX) was used in order to distinguish four target populations. A portion of CFSE high and CFSE low cells were further stained with CMTPX (Invitrogen, Eugene, OR) according to manufacturer's protocol and loaded with alternative antigen. Subsequently, populations bearing distinct antigens were mixed in equal proportions and 20 × 106 cells were injected via tail vein into naïve, PBS/CFA-immunized or MOG/CFA-immunized mice at 25 days post-immunization. A total of 48 h after injection, mice were sacrificed, splenocytes harvested, fixed in 1% PFA and collected using a BD LSR II. Data were analyzed using FlowJo software. Killing was measured as a difference in ratio between the MOG/CFA immunized mice and either naïve or PBS/CFA-immunized mice as controls.
For specific target cell populations, CD4+ cells were purified via positive selection using magnetic beads. CD4+ and CD4-depleted subsets were then stained separately according to the procedure described above. Analysis showed CD4+ populations to be >90% pure and CD4-depleted populations to contain <10% CD4+ cells.
Total RNA was extracted from purified CD8+ cells derived from in vitro-activation cultures using RNeasy Mini Kit (Qiagen, Valencia, CA) and reversed transcribed to cDNA using Ready-To-Go T-primed First Strain Kit (Amerisham, Piscataway, NJ) per manufacturer's protocol. The following primer pairs were used to evaluate relative expression level of each mRNA: β-actin: F:GTG GGC CGC TCT AGG CAC CAA; R:CTC TTT GAT GTC ACG CAC GAT TTC, IL-10: F:CAG AGC CAC ATG CTC CTA; R:GGA GTC GGT TAG CAG TAT G, TNF-α: F:CAT CTT CTC AAA ATT CGA GTG ACA A; R:TGG GAG TAG ACA AGG TAC AAC CC, IL-4: F:ACA GGA GAA GGG ACC CCA T; R:GAA GCC CTA CAG ACG AGC TCA, Perforin: F:AAC TCC CTA ATG AGA GAC GCC; R:CCA CAC GCC AGT CGT TAT TGA. All primers were purchased from Invitrogen (Carlsbad, CA). Quantitative PCR assays were performed in 25 μl reaction volumes using Brilliant SYBR Green QPCR Master Mix (Stratagene, La Jolla, CA) on an MX3000p thermocycler (Stratagene, La Jolla, CA) as described previously [2,23]. The following thermocycler conditions were used: 10 min at 95 °C, followed by 40 cycles of 45 s at 95 °C, 1 min at 72 °C followed by an extension phase of 3 min at 95 °C and 5 min at 55 °C, and completed with a product dissociation cycle. The SYBRGreen fluorescence was read during the extension phase of the reaction. The specificity of the product was confirmed by dissociation curve analysis as well as visualization of product size on 3% agarose gels. The data were normalized to β-actin (assigned a value of 1) using the standard ΔCT method and expressed as the mean relative expression (+SEM) of the indicated molecule.
In all experiments with multiple experimental groups, differences in clinical scores, peak scores, disease onset and remission indices were evaluated using two-tailed Student's t-test. Statistical significance of proliferation studies were evaluated by using twoway ANOVA analysis using Graphpad Prism Software. Differences were considered significant if p ≤ 0.05.
In previously published studies, it has been shown that immunization with neuronal antigens can result in the induction in CD8+ T cell responses in some models of EAE [19,21]. As a first step toward studying their role during autoimmune demyelination, we investigated whether the presence of such responses was unique to specific models of EAE. Using established protocols, we immunized C57BL/6 (B6) mice with MOG35–55/CFA, SJL/J mice with PLP139–151/CFA and B10.PL mice with MBP Ac1–11/CFA . Lymph node and spleen cells were harvested at various time points post-immunization (day 10; days 20–25; days 30–35). As seen in Fig. 1A (example of chronic EAE, typical of B6 mice), these time points correspond to the pre-disease, acute and chronic disease phases of MOG35–55-induced EAE. Bulk cells were stained with CFSE and cultured in the presence of no antigen, immunizing antigen, irrelevant antigen (such as OVA323–339) or concanavalin A (Con A). Following 5 days of culture, CD4+ and CD8+ proliferation was quantified by CFSE dilution, showing not only the expected CD4+ T cell response, but also robust dose-dependent CD8+ T cell proliferation (Fig. 1B). Parallel 96-well cultures were performed for traditional tritiated thymidine-based proliferation assays and confirmed robust antigen-specific proliferation in bulk lymph node cells (not shown), as well as splenocyte cultures (Fig. 1C). In flow cytometric assays, CD4+ T cell responses were detected at pre, acute and chronic phases of disease, while CD8+ responses showed slightly delayed kinetics, being consistently detected by day 20 post-immunization, suggesting the possibility that they may not play a central role in initiating disease. Of note, CD8+ T cell responses to the immunizing antigen were not unique to the B6/MOG35–55 EAE model on the H-2KbDb background. As shown in Fig. 1D, CD8+ responses could be detected following immunization with various neuroantigenic peptides in different strains of mice. In each case, the expected CD4+ T cell responses could also be detected (data not shown).
In order to confirm that this CD8+ T cell proliferation was indeed antigen-specific and not a bystander phenomenon in a bulk culture, we isolated pure populations of CD8+ T cells from the spleens of immunized mice using magnetic bead sorting (>90% purity), stained them with CFSE and cultured them in the presence (or absence) of various stimuli. To ascertain absence of cytokine secretion by APC, we used irradiated, CD8-depleted splenocytes from OVA323–339-immunized mice as APC in these cultures. Flow cytometric evaluation on day 5 confirmed the antigen specificity of the CD8 T cell response (Fig. 2A). Parallel thymidine-based assays using purified CD8+ T cells also showed dose-response pattern to the appropriate stimulating antigen (Fig. 2B) and a lack of response to irrelevant antigens (not shown), confirming antigenic specificity. Such antigen-specific CD8+ T cell responses were detected in several different models of EAE, as indicated in Fig. 2C.
Since previous publications have suggested that MOG35–55 immunization may induce a pathogenic population of T cells [13,17], we decided to focus the remainder of our studies on this specific model. A potential pathogenic role for these cells has been shown through adoptive transfer of in vitro stimulated cells into either wildtype B6 mice or T cell deficient scid mice [13,17]. To investigate the function of MOG35–55-specific CD8+ T cells in wildtype disease, we asked whether a magnetically sorted population of in vitro stimulated, highly pure CD8+ T cells could transfer EAE to naïve mice. Using in vitro protocols similar to published methods, we stimulated splenocytes from MOG35–55-immunized mice for 72 h. Following this activation, we transferred either bulk splenocytes or magnetically sorted CD4+ or CD8+ T cells from these cultures into naïve wildtype B6 mice via tail vein injections. We observed that either bulk splenocytes or sorted CD4+ T cells could robustly transfer EAE to naïve mice. In contrast, MOG-stimulated CD8+ T cells were found to be consistently incapable of adoptively transferring disease (Fig. 3A). Analysis of average peak score and disease incidence, suggested that CD4+ T cells were the major contributors to disease induction. Thus, over multiple experiments, none of the mice receiving MOG-stimulated CD8+ T cells showed signs of EAE (0/32), whereas MOG-stimulated CD4+ T cells were able to transfer robust disease, as published previously (p < 0.05).
Although we clearly observed failure of CD8-mediated EAE induction, we could not completely exclude the possibility that peptide manufacturing, culture media variations or cytokine supplementation may be affecting the results, hence we repeatedly attempted disease transfer by using various protocols that differed in number of cells transferred, pertussis toxin administration, method of T cell subset isolation (negative vs. positive selection) and the addition of various cytokines to the in vitro culture condition (IL-2, IL-12, or IL-23). In each of these attempts, highly purified MOG35–55-specific CD8+ T cells were not able to confer disease when adoptively transferred to naïve mice. Thus, we concluded that MOG-stimulated CD8+ T cells were not pathogenic by themselves in wildtype mice.
We next tested whether these cells may contribute to the pathogenesis of the disease in the context of CD4+ stimulation/help. It has been proposed that MOG-reactive CD8+ pathogenecity may be augmented by the addition of MOG-reactive CD4+ T cells in order to induce disease . Thus, we first stimulated splenocytes from MOG-immunized mice in the manner described above and separately stimulated splenocytes from OVA323–339-immunized mice with OVA323–339 as controls. Of note, the magnitude of MOG- and OVA-specific CD8+ T cell responses was comparable (data not shown). CD8+ T cells were isolated from these cultures and injected into naïve, wildtype mice via tail vein (designated as day −1). On day 0, mice were immunized with MOG35–55/CFA in order to induce active EAE. These experiments demonstrated that MOG-specific CD8+ T cells did not result in worsening of disease. Importantly and in contrast to their purported pathogenic role, we observed that MOG-specific CD8+ T cells, in fact, significantly and consistently abrogated active EAE, as shown in Fig. 3B (representative experiment) and Table 1 (cumulative data). Amelioration of EAE paralysis was found to affect both the acute and chronic phases of the disease as evidenced by the decrease in peak score and increase grade of remission of MOG-induced paralysis. In vitro-stimulated (pre-activated and expanded) CD8+ T cells (Fig. 3B; Table 1) had a stronger protective effect than those that were obtained ex vivo (data not shown).
Although we observed a clear decrease in disease severity upon pre-immunization adoptive transfer, there was a possibility that these CNS-reactive CD8+ T cells may still exhibit a pathogenic role in the context of ongoing EAE. Thus, active EAE was induced by MOG35–55 immunization, followed by transfer of MOG- or OVA-activated CD8+ T cells at the peak of the EAE course. Again, instead of exacerbating the disease, injection of these MOG-reactive CD8+ T cells resulted in a rapid and sustained decrease in disease severity, as shown in Figure 3C (representative experiment) and Table 2 (cumulative data).
Following the unexpected discovery that neuroantigen-specific CD8+ T cells were immune regulatory in function, we next sought to characterize the mechanism whereby these cells exert their function. Using QPCR, we quantified the relative level of effector cytokines on an aliquot of these in vitro stimulated MOG-reactive CD8+ T cells, which we had observed to be protective during an adoptive transfer. These cells showed significant expression of IFN-γ, IL-4, TNF-α, and perforin, suggesting both cytotoxic functionality as well as the ability for cytokine-mediated immune modulation (Fig. 4A). Supernatant cytokine levels were also measured and confirmed the robust production of IFN-γ and small but significant amount of IL-4 (Fig. 4B), although IL-10 levels were not significant above background (data not shown).
Next, we directly tested whether these cells were able to kill antigen-loaded immune target cells. For this, we used a previously described in vivo killing assay , taking advantage of the ability to detect pre-marked cell populations using intracellular dyes. Target cells where prepared by labeling naïve B6 splenocytes with high or low levels of CFSE. In some experiments, aliquots of each population were then stained with a different dye, CMTPX, allowing us to look at four target populations in a single mouse. These trackable populations were then loaded with different antigens (MOG35–55 or OVA323–339) by incubating overnight in the presence of Con A (to generate antigen-loaded “Con A blasts”). Next, these cells were washed, mixed together at a 1:1 ratio and injected intravenously into PBS/CFA- or MOG/CFA-immunized mice. Either wildtype or CD8(−/−) mice were used to ascertain the role of CD8+ T cells. Two days post-injection, splenocytes were harvested and CFSE-high and CFSE-low populations were quantified by flow cytometry. As shown in Fig. 4C, when OVA- and MOG-loaded cells were injected into wildtype B6 mice immunized with PBS/CFA (top left panel), the ratio of the two populations remained constant (~50%). In contrast, the number of MOG-loaded cells was significantly diminished in the MOG/CFA-immunized mouse (top right panel), indicating antigen-specific killing. Moreover, this killing was almost completely abrogated in MOG-immunized CD8(−/−) mice (bottom panels), confirming that it is predominantly mediated by MOG-specific CD8+ T cells and eliminating the possibility that preferential trafficking of MOG-loaded cells would account for their disappearance.
These results were confirmed in multiple replicates, by using different permutations of CFSE- and CMTPX-stained cells for loading control vs. MOG antigen. Thus, killing was observed only on MOG-loaded target cells, regardless of whether these represented the CFSE high or low peaks or the CMTPX (+) or (−) peaks (data not shown). Importantly, the target cells in these assays were Con A blasts which were loaded with no antigen, OVA or MOG peptide. Thus, the observed killing is not non-specific killing of activated cells; rather, a MOG-specific killing of MOG-loaded target cells by MOG-specific CD8+ T cells.
To further delineate the preferred immune target population for MOG-specific killing by CD8+ T cells, we separated our antigen-loaded target cells into a CD4+ fraction or a CD4-depleted fraction. In repeat and reciprocal experiments, we were able to observe killing of both MOG-loaded CD4+ cells, as well as MOG-loaded CD4-depleted APC populations (Figs. 4D,E, cumulative data shown in Fig. 4F).
Our observations suggest that the immune regulatory effect of MOG-reactive CD8+ T cells is mediated, at least in part, by affecting APC as well as by killing potentially pathogenic CD4+ T cells. To address these hypotheses, we next performed two sets of related experiments. In the first set, we asked whether MOG-specific CD4+ T cell responses were affected in mice as a result of adoptive transfer of MOG-reactive CD8+ T cells. As seen in Fig. 5A, mice that were protected from EAE by adoptively transferred MOG35–55-stimulated CD8+ T cells had significantly diminished MOG-specific CD4+ Tcell responses, compared to mice that received either no CD8+ Tcells or OVA323–339-reactive CD8+ Tcells.
In the next set of experiments, we evaluated the function of APC obtained from mice that were protected by the transfer of MOG-reactive CD8+ T cells. First, T cell-depleted splenocytes, from protected and control mice, were exposed to lipopolysaccharide (LPS) and evaluated for their cytokine secretion by ELISA. As shown in Fig. 5B, APC derived from protected mice showed significantly higher levels of IL-10 than controls and tended to show lower levels of IL-12 production, suggesting an anti-inflammatory phenotype. Differences in TNF-α secretion did not reach statistical significance. Next, APC from protected or control mice were used in proliferation assays using MOG-specific CD4+ T cells isolated from MOG-immunized mice as responders. Cultures were stimulated with no antigen, MOG35–55 or Con A. As shown in Fig. 5C, APC from MOG-CD8-protected mice showed a significant decrease in their ability to stimulate an antigen-specific T cell response as well as their ability to assist in Con A-mediated stimulation. Thus, immune regulatory MOG-reactive CD8+ T cells resulted in modulation of APC function and reduction in pathogenic CD4+ T cell responses.
The role of CD8+ T cells in the process of autoimmune pathology has been both understudied and controversial. While it is known that CD8+ T cells represent the predominant T cell in an MS lesion and are oligoclonally expanded at the site of pathology, the antigenic specificity of these cells and their role is not known. There have been multiple reports showing an immune regulatory role for CD8+ T cells as early as the 1970s [26,27]. After a period of disfavor, the place of the regulatory CD8+ T cell as an important contributor to peripheral tolerance is undergoing a revival [5,20,25,28–32]. While not as well characterized as the CD4+CD25+FOXP3+ Tregs [33–36], it is becoming clear that CD8+ T cells play a role in the peripheral regulation of autoimmune pathology [5,28,37–42]. However, in most cases, the antigenic specificity of these regulatory T cells is unclear. Qa-1-restricted cytotoxic/suppressor CD8+ T cells have been shown to have specificity for TCR-derived peptides presented by pathogenic CD4+ T cells [5,23,30,38–40,42]. We have similarly shown copolymer therapy-induced, antigen-specific HLA-E-restricted CD8+ T cells in patients with MS [5,23,30]. In contrast, the role of CNS-specific CD8+ T cells is still incompletely understood.
It makes intuitive sense that a CNS-targeted, MHC Class I-restricted CD8+ T cell response would likely have a pathogenic role in disease and, perhaps, the oligoclonally-expanded CD8+ T cells in MS lesions represent such a response. In fact, there are reports demonstrating MHC-restricted in vitro cytotoxic killing of oligodendrocytes by CNS-specific CD8+ T cells from MS patients [43,44] While these studies show the potential of a pathogenic role, the in vivo role of these cells remains poorly understood. It is entirely possible that specificity for certain antigens in the correct MHC context may allow for CD8+ T cell-mediated CNS pathogenesis. Previous studies in certain models of EAE also suggest a pathogenic role of CNS-specific CD8+ T cells, such as MBP-specific cells in a C3H mouse [14,15] and MOG35–55-specific CD8+ cells in the B6 mouse [13,17]. In addition, recent mouse models based on transgenically expressed CNS-sequestered antigens combined with TCR-transgenic CD8+ T cells or HLA-transgenic mice also suggest that these cells may have pro-inflammatory potential [45–47]. However, most of these studies involved either induced homeostatic expansion of CD8+ T cells in T cell-deficient mice or the use of transgenic manipulation.
In the current study, we have directly addressed the role of MOG35–55-induced CD8+ T cells in a classic model of EAE using wildtype B6 mice. Interestingly, while attempting to replicate their purported pathogenic role, we generated the first evidence for a novel and unexpected regulatory role for neuroantigen-specific CD8+ T cells in EAE. In multiple attempts that included varying in vitro and in vivo conditions, we were unable to transfer disease with MOG-activated CD8+ T cells. In fact, these cells attenuated the induction of EAE and were able to reduce the severity of ongoing EAE. Thus, it seems that while one can generate models wherein CD8+ T cells can target myelin-expressing cells and cause CNS inflammation, innately these cells play a predominate regulatory role in MOG35–55/CFA-induced EAE. This corroborates with the fact that MOG35–55-induced EAE is more severe in CD8−/− B6 mice , especially at suboptimal doses of inducing antigen.
In the usual setting of MOG35–55-induced EAE, the generation of a pathogenic CD4+ T cell response preceded the detection of MOG35–55-specific CD8+ T cell responses, as the CD8+ T cell responses were reproducibly detected after days 20–25 post-immunization. Thus, the dynamics of pathogenic vs. regulatory mechanisms clearly give the upper hand to the pathogenesis. Moreover, the numeric balance between the two populations also appears to be important, as we did not see significant enhanced adoptive EAE from preparations that contained only CD4+ T cells (as compared to bulk transfers). However, when an adequate number of pre-primed CD8+ T cells were transferred, they could effectively inhibit EAE.
The mechanisms of this inhibition appear to depend, at least in part, on the ability of the CD8+ Tcells to modulate APC function and CD4+ Tcell responses. These cells produce cytokines that may effect such immune modulation. For example, even IFN-γ may not be a purely pathogenic cytokine and could mediate down modulation of EAE [37,48–54]. Moreover, MOG35–55-specific CD8+ T cells also have direct cytotoxicity towards MOG35–55-loaded target cells, but not towards a non-specifically activated T cell. In that regard, these regulatory CD8+ Tcells are distinct from previously described Qa-1-restricted cytotoxic/suppressor T cells, which are capable of killing activated CD4+ T cells by recognizing heat shock proteins or other self antigens loaded in the context of Qa-1. In those examples, once activated, the suppressor cells appear to suppress in a non-antigenspecific manner, i.e. with disregard to the priming antigen. In contrast, MOG-reactive CD8+ T cells maintain their antigenic specificity, appearing to require MOG presentation. We have observed in recent experiments that MOG35–55-induced CD8+ Tcells were unable to inhibit PLP178–191-induced EAE in B6 mice and were also unable to inhibit MOG35–55-induced EAE in the absence of MHC Class I (unpublished observations). This is compatible with a model wherein these CD8+ T cells could directly kill proinflammatory populations of APC or CD4+ T cells, provided these can cross-present MOG in the context of their class I MHC. It is still unclear whether such presentation involves Class Ia or Ib molecules and what effector molecules are utilized for suppression. These issues will be the focus of future experimentation. While we have clearly demonstrated the ability of these cells to regulate immune responses in the periphery, it also remains to be seen whether these immunologic effects may be exerted directly at the site of pathology through CNS trafficking.
Thus far, our studies have not revealed a specific immunophenotypic feature of these CD8+ T cells that might allow us to sort them prior to antigenic stimulation. Thus, we do not know whether these cells might belong to the varied ex vivo CD8+ Treg population described before, such as CD8+ CD25+ , CD8+ CD28− [20,56] CD8+CD122+ , CD8+CD75s+  or CD8aa . Our experiments have relied on their antigenic specificity as a mode of stimulating and expanding these cells in vitro. It is likely that CD8+ Tregs, as a whole, may represent distinct populations, depending on the context in which they are derived, similar to the varying phenotypes of CD4+ Tregs in distinct disease settings (CD4+CD25+FOXP3+ vs. Tr1 vs. Th3 vs. Th2).
Finally, there may be important therapeutic as well as biological implications for these observations. While using autoreactive T cells as a model of therapy may seem unorthodox, there is a clear drive to generate autoantigen-reactive regulatory T cells for adoptive immunotherapy, especially with regard to CD4+CD25+FOXP3+ Tregs. Other forms of autoreactive CD4+ Tregs (Tr1, Th3) have also shown promise in animal models and are being considered as potential strategies. The role of human neuroantigen-specific CD8+ T cells in the context of MS, is of yet, unknown. We have shown that there is a high prevalence of CNS-specific CD8+ T cell responses in both healthy individuals and MS patients. If these cells, in fact, have immune regulatory functions, one may be able to harness this for therapeutic benefit in a very targeted manner.
In terms of the biological implications of such a response, it is tempting to speculate that this may represent an underappreciated arm of immune tolerance. Hence, in the setting of an ongoing robust CD4+ T cell response (such as that induced by antigen/CFA immunization), one could envision the generation of a CD8+ T cell response specific to the same inducing antigen, with the inherent function of controlling the expansion of the initial response. It remains to be seen whether such a cross-talk is inherent during non-pathogenic immune responses and what types of presentation requirements are needed for its generation.
To summarize, our studies demonstrate a novel role for neuro-antigen-specific CD8+ T cells, revealing a potential pathway of intrinsic immune regulation during ongoing inflammation. This aspect of immune modulation may have implications in human MS as well as other immune-mediated disorders and these studies should hopefully pave the way for intensified investigations in these settings.
This report provides the first evidence that CNS-specific autoreactive CD8+ T cells may play an immune regulatory role during autoimmune demyelination. Thus, purified MOG35–55-specific CD8+ T cells derived from wildtype B6 mice were incapable of transferring EAE to naïve mice and, in fact, resulted in the amelioration of disease when transferred either prior to EAE induction or during ongoing disease. These cells appear to operate through a cytotoxic/suppressor mechanism, likely resulting in both the direct cytotoxic killing of pro-inflammatory cells as well as modulation of APC.
The authors would like to thank Maycie M. Garibay and Dr. Chris L. Ayers for discussions and manuscript review. This work was supported, in part, by grant awards from the National Institutes of Health (R01AI053439, R01AI065463, K24AI079272) and National Multiple Sclerosis Society (Harry Weaver Neuroscience Scholar Award JF2118-A-2).