Regulation of Encephalitogenic T Cells with Recombinant TCR Ligands1

We have previously described recombinant MHC class II β1 and α1 domains loaded with free antigenic peptides with potent inhibitory activity on encephalitogenic T cells. We have now produced single-chain constructs in which the peptide Ag is genetically encoded within the same exon as the linked β1 and α1 domains, overcoming the problem of displacement of peptide Ag from the peptide binding cleft. We here describe clinical effects of recombinant TCR ligands (RTLs) comprised of the rat RT1.B β1α1 domains covalently linked to the 72–89 peptide of guinea pig myelin basic protein (RTL-201), to the corresponding 72–89 peptide from rat myelin basic protein (RTL-200), or to cardiac myosin peptide CM-2 (RTL-203). Only RTL-201 possessed the ability to prevent and treat active or passive experimental autoimmune encephalomyelitis. Amelioration of experimental autoimmune encephalomyelitis was associated with a selective inhibition of proliferation response and cytokine production by Ag-stimulated lymph node T cells and a drastic reduction in the number of encephalitogenic and recruited inflammatory cells infiltrating the CNS. The exquisitely selective inhibition could be observed between molecules that differ by a single methyl group (the single amino acid residue difference between RTL-200 (threonine) and RTL-201 (serine) at position 80 of the myelin basic protein peptide). These novel RTLs provide a platform for developing potent and selective human diagnostic and therapeutic agents for treatment of autoimmune disease.

T he pathogenesis of a variety of human diseases including multiple sclerosis (MS), 3 rheumatoid arthritis, diabetes, autoimmune uveitis, transplant rejection, and graft-vs-host disease appear to involve Ag-specific CD4 ϩ T cells (1)(2)(3)(4). It is thought that pathogenic T cells home to the target tissue where autoantigen is present and, after local activation, selectively produce Th1 lymphokines (5). This cascade of events leads to the recruitment and activation of lymphocytes and monocytes that ultimately destroy the target tissue (6).
Activation of CD4 ϩ T cells in vivo is a multistep process initiated by coligation of the TCR and CD4 by the MHC class II/peptide complex present on APC (signal 1), as well as costimulation through additional T cell surface molecules such as CD28 (signal 2). Ligation of the TCR in the absence of costimulatory signals disrupts normal T cell activation, inducing a range of responses from anergy to apoptosis (4 -6). Thus, a direct approach toward Ag-driven immunosuppression is to present the complete TCR ligand, Ag in the context of MHC, in the absence of costimulatory signals that are normally provided by specialized APCs. Toward the long-term goal of targeted Ag-driven immunosuppression of pathogenic T cells, we have developed a family of novel recombinant TCR ligands (RTLs) encoding portions of the ␤1 and ␣1 domains of MHC class II molecules. 4 In a previous study, we demonstrated that RTLs loaded with soluble peptide Ag had potent inhibitory activity on encephalitogenic T cells (7). Recently, we described the biochemical characterization of single-chain constructs in which the peptide Ag was genetically encoded within the same exon, a design that favored specific loading and high occupancy of peptide Ag in the MHC binding cleft, and the ability to alter the encoded Ag of interest using standard molecular biology techniques (8). We developed our initial RTL constructs for testing in experimental autoimmune encephalomyelitis (EAE), a paralytic, inflammatory, and sometimes demyelinating disease mediated by CD4 ϩ T cells specific for CNS myelin components, including myelin basic protein (MBP). EAE shares a number of immunological similarities with the human demyelinating disease MS (9) and has been a useful model for preclinical testing of therapies for the human illness (10 -16). In Lewis (LEW) rats, the dominant encephalitogenic determinant resides in the 72-89 peptide of guinea pig MBP (Gp-MBP-72-89) (17), and active immunization with this peptide in CFA can induce a severe paralytic episode starting on day 10 -11 and lasting 5-7 days with associated formation of inflammatory perivascular lesions within the CNS (18). Interestingly, the corresponding 72-89 peptide of rat MBP (Rt-MBP-72-89) that differs by a single conservative residue (T instead of S at position 80) has greatly reduced encephalitogenic and tolerogenic activity in LEW rats. We produced RT1.B-derived RTLs covalently linked to Gp-MBP-72-89, Rt-MBP-72-89, or CM-2 peptides and investigated the regulatory effects of these novel constructs on actively or passively induced EAE. We here demonstrate that RTL-201 (␤1␣1-Gp-MBP-72-89), but not RTL-200 (␤1␣1-Rt-MBP-72-89) nor RTL-203 (␤1␣1-CM-2), could both suppress and treat clinical signs of EAE through a mechanism that inhibited T cell activation and resulted in a striking reduction of CNS-infiltrating cells. These results illustrate the ability of RTL constructs (with covalently coupled Ag) to inhibit clinical and histological EAE.

RTLs
Methods for cloning, expression, purification, and biochemical analysis of these molecules have been described previously (7,8). In brief, genes encoding single-chain RTLs were constructed by splicing the sequence encoding the amino terminus of the rat RT1.B ␣1 domain to the sequence encoding the carboxyl terminus of the ␤1 domain. The 5Ј end of the gene contained an insertion sequence that encoded a covalently coupled antigenic peptide and a thrombin cleavage site embedded within a flexible linker (19). Within this linker, a unique SpeI restriction endonuclease site allowed production of RTLs with different covalently coupled peptides by simply cutting the construct and directionally cloning in the DNA fragment of interest. The organization and logic of the RTL design has been previously described (8). There is only a single amino acid difference between RTL-200 (encoding rat-MBP-72-89) and RTL-201 (encoding Gp-MBP-72-89). This difference is at position 80 (MBP numbering) within the antigenic peptide derived from MBP (PQKSQR(S/T)QDENPVVHF). Residue 12 is a threonine in RTL-200 and a serine in RTL-201. RTL-203 contains a covalently coupled cardiac myosin-derived peptide, CM-2 (KLELQSALEEAEASLEH), that has been described previously (8).

Induction of EAE
Active EAE was induced in rats by s.c. injection of 25 g guinea pig MBP (Gp-MBP) or 300 g Gp-MBP-69 -89 peptide in CFA supplemented with 100 g Mycobacterium tuberculosis strain H37Ra (Difco, Detroit, MI). The clinical disease course induced by the two emulsions was essentially identical, with the same day of onset, duration, maximum severity, and cumulative disease index. For passive transfer of EAE, short-term T lymphocyte lines were selected with Gp-MBP-69 -89 peptide from lymph node (LN) cells of naive rats or from rats immunized 10 days earlier with Gp-MBP-69 -89/CFA. After 72 h stimulation, 10 ϫ 10 6 blasting T cells were transferred i.v. into the tail vein on day 0. Details of this procedure have been described previously (23). The rats were assessed daily for changes in clinical signs according to the following clinical rating scale: 0, no signs; 1, limp tail; 2, hind leg weakness, ataxia; 3, paraplegia; and 4, paraplegia with forelimb weakness, moribund condition. A cumulative disease score was obtained by summing the daily disability scores over the course of EAE for each affected rat, and a mean cumulative disease index (CDI) was calculated for each experimental group.

Analysis of CNS mononuclear cell infiltrates from control and protected animals
Spinal cord mononuclear cells were isolated by a discontinuous Percoll gradient technique and counted as previously described (17). The cells were stained with fluorochrome (FITC or PE)-conjugated Abs specific for rat CD4, CD8, CD11b, CD45ra, TCR BV8S2, and CD134 (PharMingen, San Diego, CA) for 15 min at room temperature and analyzed by flow cytometry. The number of positive-staining cells per spinal cord was calculated by multiplying the percent staining by the total number of cells per spinal cord.

Ag-specific proliferation assays
T cell recovered from the LN of control and RTL-treated animals were suspended at 2 ϫ 10 4 cells in 200 l/well and cocultured with 1 ϫ 10 6 irradiated APC plus Gp-MBP-69 -89 Ag as described previously (23). The cultures were incubated for 3 days, the last 18 h in the presence of [ 3 H]thymidine (0.5 Ci/10 l/well). The cells were harvested onto glass fiber filters, and [ 3 H]thymidine uptake was assessed by liquid scintillation. Mean cpm Ϯ SD were calculated from triplicate wells, and differences between groups determined by Student's t test.

5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) staining of passively transferred cells
For tracking passively transferred cells, short-term T cell lines were stained with the fluorescent tracking dye CFSE (C-1157; Molecular Probes, Eugene, OR) at 0.5 M concentration for 15 min at 37°C in RPMI 1640, washed two times with 10-fold excess RPMI 1640, and then injected into the animals at 1 ϫ 10 7 cells/300 l RPMI 1640. On days 2, 4, and 7, animals were sacrificed, organs harvested as described previously (8), and cells from each organ analyzed by FACS.

Results
Three

Suppression and treatment of EAE using recombinant TCR ligands
RTL constructs were evaluated for their ability to suppress the induction of, as well as to treat existing signs of, actively induced EAE in LEW rats. Intravenous injection of 300 g of RTL-201 in saline on days 3, 7, 9, 11, and 14 after injection of Gp-MBP or Gp-MBP-69 -89 peptide/CFA suppressed the induction of clinical ( Fig. 1 and Table I) and histological (not shown) signs of EAE. All of the control animals that were untreated, that received 300 g of RTL-200, or that received 300 g of RTL-203 developed paralytic EAE ( Table I). All of the control animals that received 20 g Gp-MBP-69 -89 peptide alone (the dose of free peptide contained in the 300 g dose assuming complete cleavage of the peptide from RTL-201 in vivo) developed paralytic EAE (Table I) Table II). The mild course of EAE (mean cumulative index (MCI) ϭ 3.6 Ϯ 0.7) in the RTL-201-treated group was significantly less than the severe course of EAE in the control group (MCI ϭ 15.8 Ϯ 0.52), the RTL-200-treated group (MCI ϭ 13.6 Ϯ 0.77), or the RTL-203-treated group (MCI ϭ 14.4 Ϯ 0.6), although the duration of disease (7 Ϯ 1 days) was the same in both groups (Table II).
Consistent with the complete lack of inflammatory lesions in spinal cord histological sections (not shown), suppression of active EAE with RTL-201 essentially eliminated the infiltration of activated inflammatory cells into the CNS (Table III) (Table III).

Treatment of passively transferred EAE using recombinant TCR ligands
The  (Table IV) and histological (not shown) signs of EAE. In contrast, 19 of 20 of the control animals that were untreated, 5 of 6 animals that received 300 g of RTL-200 (encoding Rt-MBP-72-89), and 3 of 3 animals that received 300 g of RTL-203 (encoding Rt-CM-2) developed paralytic EAE (Table IV). RTL-200 produced a marked but statistically insignificant decrease in the CDI of EAE (Table IV). The specificity of the effect of RTL-201 on passively transferred disease was further evaluated after i.v. transfer of 15 ϫ 10 6 activated Gp-MBP-87-99-specific T cells to naive recipient rats. Intravenous injection of 300 g RTL-201 (encoding Gp-MBP-72-89) in saline on days 1, 3, and 5 after passive transfer of encephalitogenic Gp-MBP-87-99-specific T cells had no effect on the induction of clinical (Table IV) or histological (not shown) signs of EAE. Similarly, all five untreated rats developed paralytic EAE (Table IV).

Specific inhibition of T cell proliferation responses
Consistent with inhibition of EAE in vivo, treatment of animals with RTL-201 also specifically inhibited proliferation responses of T cells cultured ex vivo. Draining LN (DLN) cells were recovered from both treated and control animals at the peak of actively induced EAE and stimulated in vitro with Gp-MBP-72-89 (50 g/ml) or whole Gp-MBP (10 g/ml). The proliferative response was measured 72 h later using a standard [ 3 H]thymidine incorporation assay. T cells from RTL-201-treated animals showed an ϳ3-fold decrease in their proliferative response to Gp-MBP-72-89 or whole Gp-MBP compared with T cells from untreated animals (Fig. 3). T cells from RTL-200and RTL-203-treated animals showed a mild (about 25%) decrease in proliferative response to Gp-MBP-72-89 or whole Gp-MBP (Fig. 3), consistent with the effect of these molecules on suppression of EAE in vivo ( Fig. 1 and Table I).

CFSE staining of passively transferred cells
Passive transfer experiments were repeated after labeling activated Gp-MBP-72-89-specific T cells with the fluorescent dye CFSE. Intravenous injection of 300 g RTL-201 (encoding Gp-MBP-72-89) in saline inhibited infiltration of donor-derived CD4 ϩ T cells (CFSE ϩ ) into the CNS (Fig. 4, upper right quadrants; Table V). In passively transferred EAE, onset of disease typically occurred between days 3 and 5. Two days after passive transfer (before disease induction), there was a Ͼ10-fold decrease in the number of labeled CD4 ϩ T cells in the CNS of RTL-201 (encoding Gp-MBP-72-89)treated animals vs RTL-203-treated or untreated controls (40 vs 450 -500 cells/spinal cord) (Table V), and at onset of disease (Day 4) there was a 300-fold difference (200 vs 45-60,000 cells/spinal cord) ( Table V).

Discussion
The results presented above demonstrate that RTLs in which the Gp-MBP-72-89 peptide Ag was genetically encoded within the same exon as the linked ␤1 and ␣1 domains derived from MHC class II RT1.B possessed a potent and exquisitely selective ability to suppress and treat active and passive EAE mediated by T cells specific for this peptide. We have previously demonstrated that "empty" RTLs loaded with Gp-MBP-72-89 peptide suppressed and treated actively induced EAE (7). The major limitation when using these complexes was controlling the specific loading of the empty RTLs and maintenance of this complex in vivo. To overcome this limitation, we have added a sequence encoding an amino-terminal linker and the antigenic peptide of interest. These relatively small (ϳ200 aa residues) RTLs can be produced in Escherichia coli in large quantities and refolded from inclusion bodies, with a final yield of purified protein between 15 and 30 mg/L of bacterial culture. The design of the constructs allows for substitution of sequences encoding different antigenic peptides using restriction enzyme digestion and ligation of the constructs. Structural characterization using circular dichroism demonstrated that these molecules retained the anti-parallel ␤-sheet platform and antiparallel ␣-helices observed in the native class II heterodimer, and the molecules exhibited a cooperative two-state thermal folding-unfolding transition. The RTLs with the covalently linked Ag-peptide showed increased stability to thermal unfolding relative to "empty" RTLs (8).
The design of this second generation of constructs containing covalently linked Ag favors equimolar loading of peptide Ag within the binding cleft of the MHC class II ␤1␣1-derived RTL moiety, yet allows ready exchange of the encoded Ag of interest using straightforward molecular biology techniques. Of crucial importance to the  RTL concept, covalently bound Ag minimizes potential extraneous biological effects of peptide that might disassociate from our previously described "empty" RTLs (7). In experiments described here, the ability of RTL-201 to protect and treat EAE could not be accounted for by antigenic peptide released from the construct, because an equivalent molar concentration of free Gp-MBP-72-89 peptide had no inhibitory effect on EAE (Table I), nor did the construct alone or in association with a different noncovalently bound peptide (7). Thus, inhibition of EAE required the combination of both MHC and peptide components of the RTL. The Gp-MBP-72-89 epitope represents the dominant encephalitogenic determinant in the LEW rat, whereas the corresponding 72-89 peptide of rat MBP differs by a single conservative residue (T instead of S at position 80). However, this seemingly minor difference in sequence has profound immunological effects in LEW rats, with the Rt-MBP-72-89 peptide possessing about 10-fold less encephalitogenic activity on a molar basis (28,29). Most T cells induced after immunization with Gp-MBP-72-89 are RT1.B restricted, but are only weakly stimulated with Rt-MBP-72-89 (30). Conversely, most T cells induced after immunization with Rt-MBP-72-89 are also RT1.B restricted, but can still be stimulated efficiently with Gp-MBP-72-89 (28). These findings suggest that TCR interactions are stronger with the Gp than with the rat epitope. This could occur if the hydroxyl side chain at position 80 of serine provided a dominant interaction with hydrogen accepting TCR residues of Gp-MBP-72-89-specific T cells; in Rt-MBP-72-89, the presence of the extra methyl group in threonine might limit the mobility and hence the binding of the hydroxyl group, thus impeding this potentially important interaction. Studies are now underway in our laboratory to directly examine the differences in Rt-and Gp-MBP-72-89 peptide binding to the RT1.B-derived RTLs.   The clinical effects of RTL-200 and RTL-201 on EAE contrasted strikingly. RTL-201 (encoding Gp-MBP-72-89) possessed potent suppressive and therapeutic activity for actively induced disease, substantially reducing the proliferative response of DLN T cells (Fig. 3). Even more striking, RTL-201 completely suppressed passive EAE induced after transfer of Gp-MBP-72-89-specific T cells (Table IV), but had no effect on passive EAE induced with a different I-E (RT1.D)-restricted T cell line specific for a distinct encephalitogenic determinant, MBP-87-99 (Table IV). Finally, treatment with RTL-201 prevented infiltration of both CSFE-labeled donor T cells as well as host recruited inflammatory cells into the CNS (Tables III and IV). In contrast, RTL-200 (encoding Rt-MBP-72-89) and a second control, RTL-203 (encoding rat CM-2), showed only a very mild suppression of actively and passively induced disease ( Fig. 1; Table I and IV), a slight reduction of infiltrating cells into the CNS (Table III), and only a mild decrease in the proliferative response of DLN T cells (Fig.  3). The remarkable decrease in the number of inflammatory cells that penetrated the blood-brain barrier after RTL-201 treatment may provide important insights into the mechanism of action of these molecules in vivo. We are currently exploring the possibility that RTL treatment alters expression of cell-surface adhesion markers that are up-regulated after Ag challenge and that are required for crossing the blood-brain barrier.
Our results demonstrating selective inhibition of EAE with the Gpbut not Rt-MBP-72-89 RTL are in full agreement with a previous study (31) that showed that orally administered Gp-MBP-68 -88 conferred resistance to an encephalitogenic challenge with Gp-MBP-68 -88 or Rt-MBP-68 -88. However, Rt-MBP-68 -88 did not confer oral tolerance to either molecule. These observations suggest that some MHC/peptide interactions with TCR can both stimulate and tolerize Ag-specific T cells, whereas other interactions may stimulate but not tolerize. Perhaps the simplest explanation for this disparity in tolerizing activity might be a difference in ligand binding avidity to the TCR, with tolerance induction occurring only in high-avidity T cells. In accordance with the differential avidity model of self-tolerance (32), we would speculate that T cells with a high avidity for RT1.B/Rt-MBP-72-89 have been deleted in LEW rats by negative selection in the thymus, leaving TCRs with moderate to low avidity for native Rt-MBP-72-89 but high avidity for Gp-MBP-72-89. However, an important implication is that the moderate-to low-avidity TCR repertoire that remains after thymic selection to self-Ags may not be amenable to tolerization with self-sequences but would require altered peptide ligands with higher avidity for the TCR. Although differences in avidity for the Gp-and Rt-MBP-72-89 epitopes have not been assessed, the development of specifically labeled RTLs may now make such studies feasible. Our data demonstrated the clinical utility of RTLs with covalently coupled Ag in regulating pathogenic T cells in EAE. These studies provide a template for engineering human homologues that may be useful in treatment of autoimmune diseases such as MS that likely involves inflammatory T cells directed at CNS proteins.