Characterization of the Pathogenic Autoreactive T Cells in Cyclosporine-Induced Syngeneic Graft-Versus-Host Disease¹

Cyclosporine (CsA)³ is a widely used immunosuppressive drug that selectively suppresses T lymphocyte-dependent immune responses (1). Although CsA has potent immunosuppressive activity, this drug paradoxically disrupts the mechanisms governing self tolerance, in part by inhibiting the thymic-dependent clonal deletion of autoreactive T cells (2, 3). In fact, administration of CsA after autologous or syngeneic bone marrow transplantation (BMT) in humans and in rodent models elicits a T lymphocyte-dependent autoimmune syndrome with pathology similar to graft-vs-host disease (GVHD) occurring after allogeneic BMT (4, 5).

The onset of the CsA-induced autoaggression syndrome, termed syngeneic GVHD, is associated with the development of a highly restricted repertoire of CD8⁺ autoreactive T cells that promiscuously recognize MHC class II molecules (6, 7, 8, 9). MHC class II recognition occurs even though the autoreactive T cells do not have the appropriate cell surface restriction element (CD4) (6, 9). In the Lewis rat model of syngeneic GVHD, the predominant autoreactive population expresses the Vβ8.3/8.5 TCR V region gene segment (8, 9). Adoptive transfer studies confirm that the CD8⁺ Vβ8.5⁺ T cells that develop under the influence of CsA are responsible for initiation of this autoaggression syndrome (8).

Recent studies in humans and in rodent models reveal that the pathogenic effector T cells recognize a peptide from the MHC class II invariant chain (CLIP) presented in the context of MHC class II (9, 10). Furthermore, there appears to be a functional interaction between the TCR and the flanking regions of CLIP that extend beyond the peptide-binding domain of the MHC class II molecule, which may partially explain the promiscuous specificity of the autoreactive T cells (9). Although T cell clones can be detected that require either the N-terminal or C-terminal flanking regions in vitro, it appears that only those cells responsive to CLIP with the N-terminal flanking region are pathogenic in vivo (9). These cells, when assessed in a local graft-vs-host reaction (GVHR) assay, induce histologic changes consistent with a GVHR (i.e., apoptosis/dyskeratosis). Comparatively, clones responsive to CLIP with the C-terminal flanking region are not pathogenic in vivo. The underlying mechanisms accounting for the pathogenicity, the apparent restriction of the TCR repertoire, and the target Ag in syngeneic GVHD remain unclear.

The present studies characterize the pathogenic clones that mediate syngeneic GVHD. The results demonstrate that the pathogenic T cells produce IFN-γ, whereas the nonpathogenic clones produce IL-4 and IL-10. Differential cytokine production may partially explain the pathogenic potential of the autoreactive clones responsive to CLIP with the N-terminal flanking region. Recognition of the target Ag by the pathogenic T cell clones requires an autoreactive TCR that is highly conserved with respect to both Vβ and Vα gene utilization. In contrast, there is significant diversity in the autoreactive TCR D and J regions that define the CDR3 domain. Moreover, peptide constructs containing the N-terminal flanking region of CLIP can be recognized by the autoreactive T cell clones. These findings, taken together, suggest that the apparent functional interaction between this flanking region and the TCR is critical in explaining the promiscuous nature of these pathogenic T cells.

The isolation and initial characterization of CD8⁺ (confirmed flow cytometrically) autoreactive T cell clones from Lewis rats with syngeneic GVHD are described in detail elsewhere (8, 9). Briefly, lymphocytes from spleens, lymph nodes, and peripheral blood from animals with biopsy-confirmed syngeneic GVHD were cultured at limiting dilution utilizing irradiated (3000 rad) syngeneic spleen cells as APCs in complete tissue culture medium containing IL-2 (10 U/ml). The clones were expanded by restimulation (every 7 to 10 days) with irradiated syngeneic spleen cells (2 × 10⁴ cells/macrotiter well). The in vitro specificity of the effector T cells was defined using the JAM test to measure their ability to kill specifically loaded target cells (see below) (11). Pathogenicity of the autoreactive T cell clones was confirmed in vivo using a local GVHR footpad assay, as described elsewhere (9).

Killing activity was assessed using a [³H]thymidine-based assay (JAM), as described by Matzinger, that measures DNA fragmentation and cell death (11). The target cells (PHA blast cells; 5–10 × 10⁶) were pulsed with 2.5 μCi/ml [³H]thymidine for 18 h and washed three times before assay. Graded numbers (1 × 10⁴ to 1 × 10⁵) of the effector T cell clones and the target cells (5 × 10³) were coincubated for 4 h before harvest. Variability between replicate cultures averaged less than 5%.

The sequences of parent CLIP (aa 82–103) and the peptides utilized in the present studies are given in Table I and include the truncated variants of CLIP containing just the MHC class II-binding domain (aa 90–100) or this domain with either the N-terminal or C-terminal flanking region (aa 86–100, 90–104), CLIP with inverted flanking regions, an MHC class II-binding allospecific peptide from the immunogenic region of MHC RT1.A, and chimeric constructs of this allopeptide with the N-terminal and C-terminal flanking regions of CLIP (12, 13). The peptides, chemically synthesized and purified by high pressure liquid chromatography, were obtained from Quality Controlled Biochemicals (Hopkinton, MA). The peptides (>92% purity) were diluted to 10 μM in RPMI 1640 prior to loading MHC class II-positive lymphoblasts, as previously described (9). Previous dose-response studies revealed that maximal enhancement of killing was achieved by pretreating the target cells with 1 μM peptide (9).

RNA was extracted and purified with Trizol reagent (Life Technologies, Gaithersburg, MD), according to the protocol provided by the manufacturer. In brief, the cloned T cells were harvested and washed twice in PBS. Cell lysate was prepared from 5 × 10⁴ cells in 500 μl Trizol reagent with adequate mixing. After adding 100 μl chloroform, the solution was well mixed and centrifuged. The supernatant was collected and extracted once with chloroform. RNA was precipitated with 2-propanol and rinsed with 70% ethanol. Purified RNA was dissolved in 20 μl diethyl-pyrocarbonate-treated distilled water.

cDNA was prepared with Ready-To-Go reverse-transcription (RT) kit (Pharmacia Biotech, Piscataway, NJ), according to the protocol provided by the manufacturer. RT reaction was performed in a total volume of 33 μl and primed with random hexamer. For each sample, two RT reactions were conducted, the cDNA products were pooled, and 2 μl cDNA was used for PCR. Oligonucleotide primers for routine PCR analysis of Vβ and Vα gene usage have been previously described (8, 14, 15, 16). Primers used for PCR analysis of cytokines and PCR amplification before TCR sequencing are summarized in Table II (17). For cytokine analysis, actin was also amplified as a semiquantitative indicator of cDNA used in each reaction. The PCR was conducted in a total volume of 25 μl containing 20 mM Tris-HCl (pH 8.4), 50 mM KCl, 1.5 mM MgCl₂, 200 μM dNTP, 100 nM each of the primers, and 5 U Taq DNA polymerase (Life Technologies). The thermal cycler (Perkin-Elmer, Branchburg, NJ) was programmed as 95°C × 1 min, 60°C × 1 min, 72°C × 1 min, 30 cycles. The PCR products were extracted with chloroform, and electrophoresis was performed on a 1.5% agarose gel. Gels were stained with ethidium bromide to visualize the bands.

For each T cell clone, three PCR reactions were conducted and the PCR products were pooled for cloning. PCR products were cloned into the pT7 blue plasmid vector by using the pT7 Blue Perfectly Blunt Cloning Kit (Novagen, Madison, WI). DNA sequence was determined by dideoxy chain termination method and the Sequenase Version 2.0 DNA Sequencing Kit (Amersham Life Science, Cleveland, OH). The sequencing reaction was primed by the −40 Primer supplied by the kit. The TCR genes utilized by the clones were identified by comparison with previously reported TCR gene sequences (14, 16, 18).

Tissues were analyzed for expression of MHC class II and CLIP by immunohistochemistry, as previously described (19). Briefly, frozen sections were incubated with mouse anti-rat MHC class II mAb (Serotec; Harlan Bioproducts for Science, Indianapolis, IN) or with affinity-purified (CLIP-Sepharose columns) rabbit anti-rat CLIP (developed by Quality Controlled Biochemicals, Hopkinton, MA). The specificity of these Abs has been previously reported (9). The activity of the anti-CLIP Ab could be inhibited specifically with 10 μM of peptide. Controls consisted of sections incubated with normal mouse serum or prebled rabbit IgG. After 60 min of incubation, the sections were washed in Tris-buffered saline (pH 8.2) with 1% milk (TBS-milk) and incubated (30 min) with biotinylated F(ab′)₂ goat anti-mouse IgG or F(ab′)₂ goat anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA), respectively. Subsequently, the sections were washed with TBS-milk, incubated for 60 min with avidin-alkaline phosphatase, and developed with Fast Red. The sections were counterstained with hematoxylin.

Previous studies in our laboratory indicated that the autoreactive T cells that mediate syngeneic GVHD recognize CLIP in the context of MHC class II (9). Discrete subsets of these T cells exist that require not only the MHC class II-binding domain of CLIP, but also flanking regions that extend outside of the MHC peptide pocket. Interestingly, only the T cell clones that require the N-terminal flanking region of CLIP, which represent the majority (>75%) of CLIP-reactive T cell clones, were pathogenic in vivo (9).

Semiquantitative PCR was used to examine the cytokine production of the pathogenic T cell clones. The RNA was isolated from the clones 9 days after their last stimulation with irradiated APCs. To verify that equal amounts of RNA were used in each RT-PCR reaction, actin was also amplified along with the target cytokines (IL-2, IL-4, IL-10, IFN-γ). Fig. 1 shows the cytokine profiles of the T cell clones. All five pathogenic clones produced readily detectable mRNAs for IL-2 and IFN-γ. Neither mRNAs for IL-4 nor IL-10 were detected from the pathogenic clones. In contrast, the three nonpathogenic clones produced large amounts of IL-4 and IL-10 cytokine mRNA transcripts with no detectable mRNA transcripts for IL-2 or IFN-γ. As an internal control, cytokine mRNA transcripts could not be detected from APCs cultured alone for 9 days (data not shown). Ten additional clones of similar specificity (8 N-terminal restricted; 2 C-terminal restricted) revealed an identical cytokine profile with IFN-γ and IL-2 produced by the N-terminal restricted clones, and with IL-4 and IL-10 produced by the C-terminal restricted clones. Moreover, stimulation of the pathogenic clones with a truncated variant of CLIP containing the N-terminal flanking region resulted in a significant increase in IFN-γ mRNA production after 3 h in culture, as shown in Fig. 2. Stimulation with the variant containing the C-terminal flanking region or just the MHC class II-binding domain of CLIP did not induce up-regulation of IFN-γ mRNA.

The capacity to make IFN-γ implies that the pathogenic T cells are able to induce up-regulation of MHC class II-CLIP, the target Ag. Since the tongue is one of the primary target tissues for syngeneic GVHD in the rat, the expression of MHC class II and CLIP was assessed in animals with active disease. Histopathologic examination revealed that there is a marked up-regulation of MHC class II molecules in the tongue during active disease, as shown in Fig. 3,A. Staining is most intense at the epidermal-dermal junction of the mucosal epithelium. Similarly, pronounced expression of CLIP was also detected in this region of the epithelium (Fig. 3,C). Inflammatory mononuclear cells infiltrating the tissue also stained with the anti-MHC class II and anti-CLIP Abs, and probably include dendritic cells, monocytes, and activated T cells that express MHC class II (6, 7, 9, 20, 21, 22). Controls (normal mouse serum or rabbit-prebled IgG) showed no significant staining (Fig. 3, B and D). Examination of tissue from normal animals revealed no significant staining with either the anti-MHC class II or anti-CLIP Abs (data not shown), and is consistent with the results from previous studies indicating that MHC class II gene products are not expressed in normal, nonstimulated epithelial cells (20, 21, 22).

In the current study, pathogenic T cell clones were typed for TCR usage by PCR. As shown in Table III, all of the pathogenic clones (clones 1, 2, 3, 6, and 12) were Vβ8.5, while four of these clones expressed Vα11. The remaining clone was Vα10⁺. The three nonpathogenic clones (clones 18, 71, and 77) were all Vα10⁺, but used Vβ2, 5, and 8.1, respectively. Seven additional N-terminal restricted clones expressed Vβ8.5 and Vα11. Since the TCR restriction of the pathogenic clones appeared highly conserved, their TCRs were sequenced to further characterize their clonotypic receptors.

The representative sequence of the N-D-N domain and Jβ gene utilization of the TCR β-chain from each T cell clone are given in Fig. 4. Among the five pathogenic clones, clones 1 and 2 were identical, using the same N-D-N region and Jβ2.6 segments. No similarities in the N-D-N or Jβ regions were found among the other three Vβ chains of the pathogenic clones (clones 3, 6, and 12). To explore the possibility that Jβ2.6 is preferentially used by the Vβ8.5⁺/Vα11⁺ T cell clones, seven additional Vβ8.5⁺/Vα11⁺ T cell clones with the N-terminal restricted specificity were sequenced. Jβ2.6 was used by five of the seven clones (clones A, B, C, D, and E), while the other two clones expressed different J region genes. Also, there was no homology in the N-D-N for these seven clones. Vβ8.5 identity for all clones was confirmed during sequencing. Sequences of the α-chain from a select number of Vβ8.5⁺Vα11⁺ pathogenic T cell clones were also assessed. As shown in Fig. 5, each clone used a different Jα segment. No significant homology was found in the N-D-N region among the four clones, although valine and glycine were found at high frequency.

Previous studies in our laboratory suggested that there may be a functional interaction between the N-terminal flanking region of CLIP and the V region of the TCR β-chain at or near the SEB binding site (9). This interaction could explain the bias in the TCR repertoire to cells expressing a V region gene that allows responsiveness to SEB (8, 9, 23). A series of studies was conducted to evaluate the importance of this N-terminal flanking region. Ten clones were randomly selected from the total pools of lytic clones with either N-terminal or C-terminal sp. act. Fig. 6,A illustrates the lytic profile for 10 representative pathogenic clones. Target cells loaded with the MHC class II-binding domain of CLIP were not effectively killed by the pathogenic clones at limiting E:T cell ratios (5:1). Killing was maximal when target cells were loaded with CLIP containing the N-terminal flanking region. In contrast, killing was negligible when CLIP containing the C-terminal flanking region was loaded. A reciprocal pattern was observed with 10 C-terminal restricted, nonpathogenic clones (Fig. 6 B). To assess the relative spatial importance of the N-terminal and C-terminal flanking regions, target cells were loaded with a construct of CLIP that had the N- and C-terminal flanking regions inverted. Neither the pathogenic T cell clone nor the nonpathogenic clones killed the target cells loaded with the dyslexic peptide. The effect of adding these flanking regions to an MHC class II-binding allopeptide was also assessed. The pathogenic T cell clones did not kill the cells loaded with the allopeptide. In contrast, cells loaded with the chimeric allopeptide containing the N-terminal flanking region were recognized by the pathogenic T cell clones. The level of killing was comparable with cells loaded with CLIP containing the N-terminal flanking region. On the other hand, addition of the C-terminal flanking region to the allopeptide did not have a similar effect for the C-terminal restricted nonpathogenic clones.

Syngeneic GVHD is mediated by a highly conserved repertoire of CD8⁺Vβ8.5⁺ T cells that promiscuously recognize MHC class II (6, 7, 8, 9, 24). Recognition of MHC class II molecules by the autoreactive T cells is facilitated by the presentation of a peptide from the invariant chain, CLIP (9, 10). Essential for the promiscuous recognition of MHC class II independent of the CD4 accessory molecule is the interaction between the N-terminal flanking region of CLIP and the Vβ segment of the TCR (9, 24). Clonal analysis, however, reveals two discrete subsets of T cells that require either the N-terminal or the C-terminal flanking region (9). Although both subsets of cells can functionally recognize target cells in vitro, surprisingly, only the cells that require the N-terminal flanking region of CLIP are pathogenic in vivo. These cells induced histologic changes consistent with a GVHR (with marked epithelial cell destruction and dyskeratosis) (25). In contrast, clones requiring the C-terminal flanking region of CLIP were nonpathogenic, as assessed in this in vivo model. Analysis of cytokine production revealed that the pathogenic clones produced IFN-γ and IL-2, type 1 cytokines. Comparatively, the nonpathogenic clones produced type 2 cytokines (IL-4, IL-10).

The production of type 1 cytokines may underlie, in part, the pathogenicity of the N-terminal restricted clones. Release of IFN-γ in the target tissue would lead to the up-regulation of the MHC class II-CLIP complex, the target Ag of the pathogenic clones. In fact, the present studies demonstrate that MHC class II and CLIP are up-regulated in the target tissue during active syngeneic GVHD. A similar mechanism occurs during acute GVHD following allogeneic BMT, leading to the up-regulation of MHC class II in the target tissue (20, 21, 22). Interestingly, recent evidence indicates that there is discoordinate surface expression of IFN-γ-induced MHC class II proteins in nonprofessional APCs (26). Following IFN-γ stimulation, the production of DM, the class II molecule that facilitates displacement of CLIP with nominal peptide (27), is delayed relative to expression of DR and the coordinated expression of the invariant chain (26, 27). Moreover, there is an absence of DM and DR colocalization, also leading to a failure or delay in the exchange of CLIP for nominal peptides. In effect, cell surface MHC class II expression on nonprofessional APCs after exposure to IFN-γ would be largely associated with CLIP, which is critical for stabilization of the MHC class II molecule (12, 13, 25, 26). Thus, IFN-γ-induced up-regulation of the MHC class II-CLIP complex enhances the potential for target tissue destruction by the autoreactive T cells that mediate syngeneic GVHD.

The role of the C-terminal restricted, nonpathogenic clones in syngeneic GVHD remains unclear. The nonpathogenic clones appear to express an array of Vβ molecules, while there is a marked restriction to Vβ8.5 in the pathogenic T cell subset (8, 9). These clones produce IL-4 and IL-10, which do not actively promote up-regulation of MHC class II in the target tissue. In this regard, allospecific CD8⁺ cells that secrete type 2 cytokines do not mediate acute GVHD (28). On the other hand, the C-terminal restricted subset of cells may promote the development of the chronic phase of syngeneic GVHD, characterized by fibrosis and sclerosis (20). Consistent with this hypothesis is the observation that cells secreting type 2 cytokines play a critical role in the pathogenesis of chronic GVHD after allogeneic BMT (29, 30). Alternatively, the C-terminal restricted, nonpathogenic clones may represent a subset of regulatory T cells that may modify the response of the pathogenic T cells. Current studies are attempting to define the role of this subset in vivo.

The pathogenic T cells that mediate syngeneic GVHD not only have a unique specificity for the MHC class II-CLIP complex, but also are restricted in their use of TCR V region genes (8, 15). Molecular analysis of the TCR of the pathogenic T cell clones reveals a remarkable conservation with respect to Vα and Vβ gene utilization. Virtually all of the pathogenic clones express the Vβ8.5 and Vα11 TCR molecules. Interestingly, these same V region components are utilized by the CD4⁺ autoreactive T cell subset in syngeneic GVHD (8, 24, 31). This autoreactive T cell subset, which also has specificity for CLIP, is not pathogenic per se, but can amplify the activity of the CD8⁺ pathogenic T cells (30). On the other hand, sequence analysis of the CD8⁺ clones reveals marked heterogeneity in the CDR3 domains of both Vα and Vβ chains that, in part, define the peptide specificity of the TCR. It appears that multiple CDR3 constructs can be used to recognize CLIP since there is degenerate recognition, a “looseness of fit” between the TCR and the MHC class II- peptide complex (32). In fact, recent evidence indicates that specific T cell clones can recognize a multitude of peptides presented by MHC molecules (33, 34). On the other hand, loading the truncated variant containing just the MHC class II-binding domain of CLIP (35) was insufficient to allow killing of the target cells at low E:T cell ratios. Killing of target cells loaded with this truncated variant of CLIP could only be demonstrated at much higher E:T cell ratios with a limited efficiency (50–60% maximum killing), indicating specificity for the MHC class II-binding domain of CLIP, but with weak affinity (9). These data, however, suggest that effective recognition requires an additional interaction, outside of the MHC-binding domain.

The N-terminal flanking region of CLIP, which extends beyond the terminus of the peptide-binding groove of MHC class II molecules, plays a major role in recognition (35, 36, 37). Previous studies suggest that the N-terminal flanking region interacts with MHC class II molecules at or near the SEB binding site and can stabilize peptide binding (38, 39). On the other hand, the hypothesis that effective recognition of the MHC class II-CLIP complex by the pathogenic T cells requires an interaction of this flanking region with the Vβ chain of the TCR outside of the CDR3 domain, is based on the findings that SEB pretreatment of the autoreactive T cells inhibits their lytic function (9), and that the N-terminal flanking region of CLIP prevents the SEB-dependent ligation of the TCR V region segment with MHC class II molecules (40). The results from the current studies indicate that this interaction is quite important, and defines the specificity of the pathogenic T cells. Loading a chimeric construct of an MHC class II-binding allopeptide with the N-terminal flanking region of CLIP (consisting of the KPVSP amino acid sequence) promoted effective recognition of the target cells. Target cells loaded with the unmodified allopeptide could not be killed even at high E:T ratios (9), suggesting that this interaction may override inadequate complementarity between the CDR3 domain of the TCR and the peptide. Recognition of the peptide bound in the groove of MHC class II by the autoreactive T cells appears, at best, to be secondary to that of the interaction between the N-terminal flanking region of CLIP and the TCR V region. Furthermore, the results from the present studies suggest that the spatial orientation of this flanking region is also of critical importance. Inversion of the flanking regions on CLIP does not allow for recognition and killing of the target cells. These data are consistent with recent studies indicating that T cell recognition of Ag requires the proper orientation of the TCR relative to the MHC class II-peptide complex. (41) The CDR1 domains of the TCR α- and β-chains must be appropriately aligned with the α- and β-chains of MHC class II molecules to allow recognition of the peptide. Therefore, it appears that the N-terminal flanking region must also be in alignment with a specific V region component of the TCR. This interaction of the flanking region on CLIP, when presented in the context of MHC class II, appears to be near a contact site between the Vβ segment of the TCR, the MHC class II determinant, and SEB. Such a mechanism may explain the V region restriction of the autoreactive T cell repertoire in syngeneic GVHD, in particular to a Vβ gene that allows responsiveness to SEB (8, 9, 23, 42). The conserved Vα segment may also play a significant role since recent studies indicate that the TCR α-chain can influence the response to superantigens (43). Nevertheless, the interaction between this flanking region of CLIP and the clonotypic Ag receptor would strengthen the avidity of the TCR-CLIP-MHC complex and may bypass the requirement for the appropriate cell surface accessory molecule (i.e., CD4).

Comparatively, the nonpathogenic clones are restricted by the C-terminal flanking region of CLIP. Although the analysis was limited, these clones express an array of Vβ molecules, but appear to share a common Vα element. Whether a similar interaction occurs between the C-terminal flanking region of CLIP and the V region elements of the TCR is currently under investigation.

In summary, the results of the present studies reveal that syngeneic GVHD is mediated by a highly restricted subset of T cells that produce type 1 cytokines. Moreover, the interaction between the N-terminal flanking region of CLIP and the autoreactive TCR is critically important, allowing for effective target cell recognition and tissue destruction. Interestingly, analysis of an autoaggression syndrome in humans occurring after clinical autologous BMT and CsA treatment reveals virtually identical autoimmune mechanisms (10). The analogous promiscuous recognition of MHC class II molecules and dependence on CLIP of the autoreactive T cells demonstrate the fundamental importance of this mechanism.

We acknowledge Ms. Kathy Suttka for her excellent secretarial assistance, and Dr. Teresa Pfaff-Ameese for her assistance with the immunopathology. We also thank Dr. Ephraim Fuchs for his critical review of this manuscript.