Autoreactive T cells can be protected from tolerance induction through competition by flanking determinants for access to class II MHC

Abstract

It is not clear why the N-terminal autoantigenic determinant of myelin basic protein (MBP), Ac1–9, is dominant in the B1O.PL (H-2^u) mouse, given its weak I-A^u-MHC binding affinity. Similarly, how do high-affinity T cells specific for this determinant avoid negative selection? Because the MBP:1–9 sequence is embryonically expressed uniquely in the context of Golli-MBP, determinants were sought within the contiguous N-terminal “Golli” region that could out-compete MBP:1–9 for MHC binding, and thereby prevent negative selection of the public response to Ac1–9, shown here to be comprised of a Vβ8.2Jβ2.7 and a Vβ8.2Jβ2.4 expansion. Specifically, we demonstrate that Ac1–9 itself can be an effective inducer of central tolerance induction; however, in the context of Golli-MBP, Ac1–9 is flanked by determinants which prevent its display to autoreactive T cells. Our data support competitive capture as a means of protecting high-affinity, autoreactive T cells from central tolerance induction.

The mechanism which allows T cells specific for self determinants to circumvent immunological tolerance remains a fundamental question. Although it is possible for an autoantigen to be geographically sequestered in an immunologically privileged site (e.g., the eye), studies have demonstrated thymic expression of encephalitogenic proteins (1, 2). T cells specific for the N terminus of myelin basic protein (MBP) have been well described in the H-2^u mouse model of experimental autoimmune encephalomyelitis (EAE). Theories on how Ac1–9-specific T cells escape negative selection have been described (3). These previous studies attribute the lack of central tolerance induction of the Ac1–9-specific repertoire to Ac1–9's relatively poor binding affinity for the I-A^u MHC class II molecule. This mechanism, however, does not entirely explain how T cells specific for the Ac1–9-determinant escape negative selection, given that this determinant is still dominant within the context of MBP. Of particular interest is the “public” Vβ8.2Jβ2.7 response used by all B10.PL mice when primed with Ac1–9 (4). Such high-affinity, self-directed T cells are usually purged during development (5). Thus, the Ac1–9-specific T cell repertoire of the B10.PL mouse seems to be one of the few exceptions to the usual finding that “public” high-affinity T cells directed against dominant self-determinants are preferentially tolerized.

In this context, it is of interest that analysis of the genetic region upstream to the “classical” MBP exons revealed a larger unit containing several novel exons, termed the “Golli” region (genes of the oligodendrocyte lineage) (6). The 5′ Golli exons are transcribed from their own upstream promoter in frame with the classical MBP exons and truncated products from this gene complex are found expressed at high levels in the fetal thymus; meanwhile, in its classical form, MBP has never been detected in the developing thymus.

Based on this knowledge, we considered the proposition that a strong MHC binding flanking determinant might out-compete MBP:1–9 for presentation in the thymus by a mechanism involving competition for MHC loading. Here we show that an overlapping determinant just upstream from MBP:1–9 exists within the Golli-MBP junctional region. This determinant preferentially binds to I-A^u, preempting the classical 1–9 determinant from being displayed to Ac1–9-specific T cells. In addition to this determinant, a second, flanking C-terminal determinant in the 7–20 region also competes for MHC binding. We hypothesize that the presence of an intact, high-affinity, Ac1–9-specific repertoire in the adult is a result of competitive occupancy of I-A^u by one or both of these flanking determinants.

Materials and Methods

Generation of Golli Gene Knockout (KO) Mice.

The Golli-KO mice are described in detail elsewhere (A.T.C., R. R. Voskuhl, P. M. Pribyl, K. Kampf, and C.W.C., unpublished data). Briefly, exon 2 of the MBP gene, containing the translation initiation site for all Golli transcripts, was deleted to generate Golli-deficient mice. A 7.5-kb exon 2-containing fragment of 129SVEv mouse DNA was cloned into pSport (GIBCO/BRL). The clone was linearized with StuI and exon 2 was removed by ExoIII exonuclease digestion. The ends were repaired with T4 DNA polymerase and exon 2 was replaced by the neopoly(A) gene (Stratagene) oriented in the reverse direction. Mouse embryonic stem (ES) cells were transfected with the targeting construct. Targeted ES cell clones were then injected into C57BL/6 blastocysts and chimeric founders were backcrossed to C57BL/6. Mice were then backcrossed for an additional five generations onto the B10.PL background. After three backcrosses animals expressing I-A^b were discarded. Mice homozygous for H-2^u were then backcrossed for an additional two generations on to the B10.PL background.

Proteins.

Recombinant murine Golli-MBP proteins were produced as described in detail elsewhere (2). Murine MBP was purified from mouse brains (Rockland, Gilbertsville, PA) as described (7).

Peptides.

Peptides were synthesized on an Advanced ChemTech synthesizer and purified on a reverse phase column by high performance liquid chromatography. Purity was then determined by mass spectrometry and capillary electrophoresis. Amino acid sequences of peptides used in this study can be found in Tables 1 and 2.

Table 1.

LDVM 1–9 binds to I–A^u 30-fold better than Ac 1–9

Peptide	Sequence	IC50, μM
LDVM 1–9	LDVMASQKRPSQR	0.25
DVM 1–9	DVMASQKRPSQR	3.0
VM 1–9	VMASQKRPSQR	5.3
Ac1–9	Ac-ASQKRPSQR	7.4

Peptides were assayed for their relative binding affinity for purified I–A^u molecules by using an inhibition test, measuring the concentration that leads to 50% inhibition of binding of a labeled standard peptide. 

Table 2.

LDVM 1–9 cannot induce EAE, whereas other peptides lacking the acetyl group can do so

Experiment no.	Mouse strain	Antigen	Dose per mouse, μg	Sequence	Incidence (%)	Mean day of onset	Mean severity
1	B10.PL	Ac1–9	100	Ac-ASQKRPSQR	2/2 (100)	12.0	3.0
	B10.PL	LDVM 1–9	100	LDVMASQKRPSQR	0/3	—	—
2	F1	Ac1–9	20	Ac-ASQKRPSQR	3/4 (75)	23.7	2.7
	F1	G (1–9) Y4	200	GASQYRPSQR	4/4 (100)	17.3	3.3
	F1	LDVM 1–9	200	LDVMASQKRPSQR	0/4	—	—
3	F1	Ac1–11	50	Ac-ASQKRPSQRSK	3/3 (100)	13.0	4.3
	F1	(1–9) Y4	100	ASQYRPSQR	4/4 (100)	14.0	2.5
	F1	LDVM (1–9) Y4	100	LDVMASQYRPSQR	2/4 (50)	13.0	1.0
4	F1	Ac1–9	100	Ac-ASQKRPSQR	3/3 (100)	9.0	4.3
	F1	BG21	200	Whole protein	2/6 (33)	32.0	3.0
	F1	G 1–9	100	GASQKRPSQR	4/4 (100)	17.5	4.0
	F1	G (1–9) Y4	100	GASQYRPSQR	3/3 (100)	13.0	3.0
	F1	M 1–9	100	MASQKRPSQR	1/3 (33)	21.0	5.0
	F1	M(1–9) Y4	100	MASQYRPSQR	2/3 (67)	12.0	4.0
5	F1	mMBP	100	Whole protein	3/3 (100)	10.3	3.3
	F1	Golli only	400	Whole protein	0/6	—	—
6	F1	Ac1–9	50	Ac-ASQKRPSQR	4/4 (100)	10.8	3.8
	F1	LDVMR(2–5)	100	LDVMRSQKR	1/3 (33)	21.0	3.0
	F1	1–9	100	ASQKRPSQR	3/4 (75)	27.7	2.3

The F₁ strain used was (B10.PL × SJL)F₁. The clinical severity of EAE was scored daily as follows: 1, toss of tail tonus; 2, hind limb weakness or forelimb involvement alone; 3, total hind limb paralysis; 4, hind and forelimb paralysis; and 5, moribund/death. Mean severity was calculated by using scores from only those mice exhibiting clinical signs of EAE. 

Mice.

SJL/J and B10.PL/J mice were purchased from The Jackson Laboratory. These strains as well as (SJL × B10.PL) F₁ mice were bred in our mouse colony. Female mice were used at 8–20 weeks of age. Mice were age-matched for each experiment.

Class II Peptide Binding Assays.

MHC class II peptide-binding assays were conducted as described in detail elsewhere (8).

Lymph Node Proliferation Assay.

T cell proliferation assays were conducted as described in detail elsewhere (9).

Hybridomas.

The Ac1–9-specific hybridomas 172.10 and BP1A1.2 were maintained in this laboratory. LDVM 1–9-specific T cell hybridomas were created by fusing LDVM1–9 primed lymph node cells with BW5147 (α−/β−). Briefly, four B10.PL mice were immunized with LDVM1–9 and draining lymph nodes were removed on day 10. T cells were expanded in the presence of peptide (10 μg/ml) and then fused to BW5147 (α−/β−). Peptide-responsive hybridomas were identified by IL-2 production using a bioassay with IL-2-requiring HT-2 cells.

Induction of EAE.

For induction of EAE, mice were immunized s.c. with 100 μl of a complete Freund's adjuvant emulsion containing 200 μg of Mycobacterium tuberculosis H37Ra (Difco) and the noted amount of protein or peptide divided among three sites as described (7). One and three days later, mice were injected i.p. with 100–200 ng of purified pertussis toxin (List Biological, Campbell, CA) in 0.5 ml BSS.

Fetal Thymic Organ Culture.

Thymus lobes dissected from 14- to 15-day-old fetal Vβ8.2Jβ2.4 T cell receptor (TCR) transgenic animals (10) were placed on the surface of filters (0.45 mm pore size, Millipore) supported on blocks of surgical Gelfoam (Upjohn). Originally Ac1–9 was added for 8 h to day-8 cultures at 0.1, 1, and 10 μg/ml, because this T cell was only of medium affinity for Ag, and Ac1–9 itself is not a strong binder. All concentrations were found to effectively eliminate Ac1–9-specific CD4⁺ CD8⁺ double positive cells. For all subsequent experiments antigen was added to cultures at 1 μg/ml. For experiments in which cells were stained with propidium iodine (PI) and annexin V, cultures were harvested 6 h after antigen administration. Thymocyte suspensions were prepared by gently forcing the intact lobe through an 18 gauge needle.

Immunoscope Analysis.

Immunoscope analysis was performed as has been described in detail elsewhere (5, 9).

Results and Discussion

Seeking a Strong Competitive Binding Determinant for the Class II Molecule, I-A^u.

MHC binding studies were conducted to identify candidate peptides within Golli capable of being preferentially captured by class II MHC molecules. These studies were performed on overlapping 15-mer peptides, offset by five residues, covering the entire “Golli only” sequence, as well as the Golli-MBP junctional region, 13-mer peptide, LDVM1–9 (LDVMASQKRPSQR). Several peptides from the Golli region that had the ability to bind significantly to purified I-A^u molecules were identified (E.M., A.S., and E.E.S., unpublished data). Interestingly, among these, the LDVM1–9 peptide was the best binder to I-A^u with a 30-fold better binding affinity than Ac1–9 (Table 1). One of several plausible explanations for the higher binding affinity of LDVM1–9 relative to 1–9 is the existence of a second determinant within the 13-mer with hydrophobic residues at positions −4 (leu) and −1 (met). This view is supported by the fact that substitution of Arg with Met at position 4 in Ac1–9 (ASQMRPSQR) leads to a substantial increase in I-A^u binding (8). To test the importance of the N-terminal residues, the I-A^u binding capacities of different LDVM1–9 truncated analogs were measured. Removal of the N-terminal leucine residue from the 13-mer greatly diminished (>10-fold) its affinity for I-A^u (Table 1). This finding has also been recently reported by Garcia and colleagues (11).

Ac1–9-Specific T Cells Do Not Respond to LDVM1–9 or to Golli-MBP.

To evaluate whether upstream residues encoded from Golli exons could affect the reactivity of Ac1–9-specific T cells, Ac1–9, LDVM1–9, and BG21 [Golli (1–134)–MBP (1–56)] were tested for their ability to induce the Ac1–9-specific hybridoma 172.10. Whereas this clone responded well to Ac1–9 and whole MBP, it was unable to respond to LDVM1–9 or BG21, each containing residues 1–9 (data not shown). Therefore, extending the MBP:1–9 determinant in the N-terminal direction by including the Golli residues, LDVM interfered with activation of this hybridoma.

We also sought the complementary evidence by isolating lymph node cells from LDVM1–9-immunized B10.PL mice and testing their proliferative response to various peptide antigens in vitro. Consistent with the results discussed above for Ac1–9-specific T cells, no cross-reactivity between Ac1–9 and LDVM1–9 could be demonstrated (Fig. 1A). Specifically, only LDVM1–9 and DVM1–9 could stimulate LDVM1–9 primed T cells; no recall responses were seen to M1–9 or Ac1–9, and there was only a small response to VM1–9. In conclusion, these results suggest that when the Ac1–9 determinant is flanked by LDVM, a completely new determinant is created.

Figure 1.

Ac1–9 cannot stimulate LDVM1–9 primed T cells. (A) B10.PL mice were immunized with LDVM1–9 and on day 10, in vitro proliferative responses were measured. Only LDVM1–9 and DVM1–9 gave significant recall responses (S.I. >3). M1–9 and Ac1–9 each failed to recall a response. Each bar represents cells pooled from two mice. Data are shown from one of three experiments with similar results. Only two pools of cells were assayed against M(1–9)Y4. (B) Twenty-two LDVM1–9 hybridomas were screened against N-terminally truncated peptides. In all cases, the best responses were seen to LDVM1–9 (○) itself, with poorer responses to DVM1–9 (◊); in seven hybridomas, responses to BG21 (▵) were obtained at high doses, as shown here. None of the LDVM1–9-induced hybridomas were able to respond to Ac1–9 (□) or MBP:1–9 (+) at any concentration. One representative hybridoma is shown, which fails to respond to VM1–9 (×), M1–9 (*), Ac1–9 (□), and MBP:1–9 (+).

At the clonal level, 22 LDVM1–9-specific T cell hybridomas were tested for cross-recognition of either Ac1–9 or unacetylated 1–9. None were able to do so. The majority (15 of 22) of these hybridomas responded only to LDVM1–9, or to both LDVM1–9 and DVM1–9. Seven of the 22 hybridomas responded to BG21 itself, showing that the LDVM1–9 determinant can be processed from BG21 and presented by splenic antigen presenting cells (Fig. 1B).

The Importance of the N-Terminal Acetyl Group of Ac1–9.

Given that other studies have supported a requirement for the N-terminal acetyl group in the activation of Ac1–9-specific T cells (12), the inability of LDVM1–9, which lacks an acetyl group, to activate Ac1–9-specific T cells might be expected. However, one cannot conclude that all unacetylated peptides will be poor stimulators of Ac1–9-specific T cells. MHC class II molecules can bind 13- to 70-mer peptides; thus, the positively charged N terminus of a determinant will often reside outside of the MHC binding groove. Thus, an alternative view is simply that the acetyl group allows the MBP:1–9 residues to bind within the I-A^u class II groove by shielding the charged N terminus. If this were indeed the key function of the acetyl group of Ac1–9, one might hypothesize that a longer, 13-mer peptide with the N-terminal extension, LDVM, would be an equivalent inducer of the 172.10 hybridoma, which is not the case. An alternative hypothesis is that the LDVM amino acids constitute a portion of a competitive register that preempts 1–9 binding. To test this hypothesis, we used amino acid residues known to bind poorly to the pockets within I-A^u (8, 13) to destroy the MHC binding capacity of the left-side “LDVM” register. The resulting peptide, [SDVGASQKRPSQR = SDVG (1–9)], with MHC-binding residues L and M ablated, was tested and found to be an excellent activator of the Ac1–9-specific hybridoma 172.10 (data not shown). These results are consistent with the hypothesis that LDVM1–9 contains an N-terminal dominant register competitive with MBP:1–9, whose preferential binding to I-A^u prevents the 1–9 register from binding to the MHC class II molecule in an appropriate orientation to permit activation of Ac1–9-specific T cells.

In addition to the LDVM register, overlapping 1–9 on its left side, a right side flanking determinant which binds to I-A^u has also been described (14). To study possible competitive capture by this right side flanking region, we prepared the peptide LDVMASQKRPSQRSKYLATA (=LDVM 1–16). Like LDVM1–9, this 20-mer failed to stimulate the Ac1–9-specific hybridoma, 172.10. To determine whether the right side flanking region also plays a competitive role, mutants of the 20-mer designed to abrogate binding of either or both of these flanking registers were prepared. Again amino acid residues known to bind poorly to the I-A^u binding pockets were used, and the following three peptides were compared with LDVM1–16: SDVGASQKRPSQRSKYLATA, LDVMASQKRPSQRSKDEATA, and SDVGASQKRPSQRSKDEATA. The data in Fig. 2 show that only when both flanking regions are neutralized by mutation does the ability to stimulate Ac1–9-specific T cells become rescued. With only one flank mutated, the remaining wild-type flank was able to competitively exclude 1–9 from stimulating its specific hybridoma. Clearly, MBP:1–9 without its N-terminal acetyl group is a more than adequate inducer of specific T cells when competing flanking determinants are functionally excised from the antigen.

Figure 2.

An MBP Ac1–9-specific hybridoma (172.10) was not stimulated by BG21 or LDVM1–9. To study competition for MHC binding by the left-sided LDVM register and by the right-sided MBP:7–20 register, the LDVM1–9 determinant (“LDVM”) was extended to include seven additional MBP residues forming the 20-mer LDVM1–16 (LDVMASQKRPSQRSKYLATA). The Ac1–9-specific hybridoma 172.10 failed to be stimulated by LDVM1–16 (◊). By using amino acid substitutions known to bind poorly within the pockets of I-A^u, three additional peptides were then synthesized [SDVG1–16 (▵), LDVM1–16(12D,13E) (○), SDVG1–16(12D,13E) (+)] designed to destroy the left-side, right-side, or both flanking registers, respectively. Only when both left- and right-sided registers were destroyed could the peptide stimulate 172.10. A second approach was also used in which the binding of the MBP:1–9-register within the 20-mer LDVM1–16 was increased in an effort to increase the 1–9-register's competitiveness for class II bindng. The increased binding was achieved by substituting 4Y for 4K in the original 1–9 register, to yield LDVMASQYRPSQRSKYLATA (×). This peptide was successful at stimulating the Ac1–9-specific hybridoma 172.10. The Golli-mbp protein BG21 (*) was unable to stimulate the 172.10 hybridoma showing that the 1–9 determinant is not processed from Golli-MBP. Ac1–9 was of course successful at stimulating 172.10 (□).

In summary, all of the evidence presented is consistent with LDVM1–16 bearing three overlapping determinants: an N-terminal register that includes the residues LDVM, a register that comprises the “1–9” determinant of classic MBP, and a carboxy register within the 7–16 region.

LDVM1–9 Cannot Induce EAE, Whereas Other Peptides Lacking the Acetyl Group Are Able to Cause Disease.

Having analyzed the activity of the LDVM1–9 peptide and its truncated analogs at the level of MHC binding and T cell antigenicity, the capacity of the same antigens to induce EAE was studied. In Table 2, the ability of several Ac1–9 variants to induce disease in the B10.PL mouse or the (B10.PL × SJL) F₁ mouse is presented. These data show that the acetyl group is not absolutely required for pathogenicity because several peptides lacking an N-terminal acetyl group are nevertheless able to induce EAE. In fact, disease induction was only slightly affected when the acetyl group of Ac1–9 was replaced with glycine. Unacetylated 1–9 and M1–9 were also able to induce EAE, though not nearly as well as Ac1–9 itself.

Importantly, LDVM1–9 was unable to induce EAE in any of nine mice tested. Histological studies were performed on selected mice and verified the lack of any pathology (data not shown). This finding agrees with the proliferation data discussed earlier, showing that LDVM1–9 was unable to stimulate Ac1–9- specific T cells, despite the presence of the 1–9 sequence.

LDVM(1–9)Y4 Allows the C-Terminal Moiety of LDVM1–9 to Gain Parity and Cause EAE.

An attempt was made to alter the competitive balance between the two registers within LVDM1–9 by increasing the binding affinity of its C-terminal moiety. In the case of the Ac1–9 peptide, substituting the lysine at position 4 with tyrosine or methionine (to yield Ac-ASQYRPSQR or Ac-ASQMRPSQR) greatly increases its binding to the I-A^u MHC class II molecule (8). It could be predicted that the same single amino acid change (LDVMASQYRPSQR) should also increase the binding to I-A^u of the right side register within LVDM1–9, recognized by Ac1–9-specific T cells. If correct, this substitution should permit substantial recognition by Ac1–9-specific T cells. In fact, when a Y4 substitution was introduced into LDVM1–9, the resulting peptide, LDVM(1–9)Y4 (LDVMASQYRPSQR), gained the capacity to stimulate an Ac1–9-specific T cell line (Fig. 3), was able to induce EAE (Table 2), and became an excellent stimulator of an Ac1–9-specific T cell hybridoma (data not shown). LDVM(1–9)Y4 was found to bind to I-A^u with an affinity over 50-fold better than LDVM1–9 (ref. 11 and data not shown). When this same substitution was made within the 20-mer LDVM (1–16), the resulting peptide gained the ability to stimulate the Ac1–9-specific hybridoma, 172.10 (Fig. 2). These results are consistent with the notion that once the 1–9 register overcomes its poor binding to I-A^u, it can achieve a degree of codominance with the flanking N- and C-terminal registers.

Figure 3.

LDVMR2–5 and LDVM(1–9)Y4 can stimulate an I-A^u restricted Ac1–9-specific T cell line. An I-A^u restricted Ac1–9-specific T cell line was cultured with the indicated peptide antigens and proliferation was measured via a ³H-thymidine incorporation assay. Significant responses (S.I. >3) were seen to Ac1–9 as well as to many of its variants, all of which lack N-terminal acetyl groups. LDVM1–9 failed to stimulate this Ac1–9-specific cell line. The background value was 850 cpm.

LDVMR2–5, with Only Two Amino Acids in Direct Sequence Alignment with Ac1–9, Is Able to Induce EAE and Stimulate Ac1–9-Specific T Cells.

We next set out to determine whether the LDVM moiety of the 13-mer competed for and occupied the same site within I-A^u as Ac1–9, or simply hindered the binding of Ac1–9-specific T cells by interfering at a distinct site on I-A^u, flanking the binding groove. It has previously been shown that the arginine at position 5 of Ac1–9 serves as the major TCR contact residue in B10.PL mice (15). In addition, modeling of the Ac1–9 determinant within I-A^u provides strong evidence that the arginine at position 5 makes contact with the MHC molecule and also extends out of the groove making it available for TCR recognition (13). Therefore, LDVMRSQKR (LDVMR2–5) was synthesized to determine its effect on the stimulation of Ac1–9-specific T cells: in this variant, the only alteration from LDVMASQKR (LDVM1–5) is the R for A substitution. Thus, only the two arginines at positions 5 and 9 are in direct sequence alignment with Ac1–9, and it has already been shown that 9R is irrelevant for stimulation of Ac1–9-specific T cells (16). Because 4M is favored over 4K for MHC binding (8), and because arginine is the fifth residue in both the Ac1–9 and LDVMR2–5 sequences, it was predicted that if the LDVM moiety competed for the same site within I-A^u the 4M I-A^u anchor would focus 5R in just the right position for T cell stimulation. Consistent with this prediction, LDVMR2–5 stimulated Ac1–9-specific but not LDVM1–9-specific T cells; it could also induce EAE (Fig. 3; Table 2). These results underscore the importance of the 5R residue in the Ac1–9 molecule for T cell recognition, and argue for direct competition at a single site between the N- and C-terminal registers of LDVM1–9. They also demonstrate that a shared single amino acid is quite adequate to account for T cell recognition.

The LDVM1–9-Specific Repertoire Is Subjected to Strong Tolerogenic Pressures.

Data presented here conclusively demonstrate that the MBP:1–9 determinant fails to be generated from LDVM1–9 and Golli-MBP. However, little evidence has thus far been presented for thymic expression of the left-side determinant within LDVM1–9. Golli-MBP has been shown to be expressed well in the thymus (2) and splenic or thymic antigen presenting cells pulsed with the Golli-MBP protein BG21 can stimulate LDVM1–9-specific T cells (Fig. 1B, and data not shown). If the left side register of LDVM1–9 were expressed thymically from fetal Golli-MBP, one would predict that the endogenous expression of the bridging LDVM1–9 should profoundly affect the development of LDVM1–9-specific T cells. To test this hypothesis, we used a Golli knockout mouse, produced by A. Messing and A.T.C. (see Materials and Methods), that expresses “classical” MBP but lacks the thymic expression of Golli-MBP. The mouse was prepared by deleting exon 2 of the Golli–MBP gene and replacing it with a neopoly A gene oriented in the reverse direction. The resultant animals did not express Golli proteins in any tissue, but expression of classic MBP was unaffected (A.T.C., R. R. Voskuhl, P. M. Pribyl, K. Kampf, and C.W.C., unpublished data). Priming wild-type animals with Golli-MBP resulted in only a small recall response to LDVM1–9, presumably owing to tolerance induction to determinants available from Golli-MBP. In comparison, Golli KO animals primed with Golli-MBP mount a greatly enhanced response to LDVM1–9 (Fig. 4). Accordingly, in the wild-type mouse, the LDVM1–9 determinant is apparently expressed endogenously in a tolerogenic fashion, whereas the 1–9 residues of MBP remain protected from tolerogenic display.

Figure 4.

Golli-KO mice mount an enhanced response to LDVM1–9. Golli-KO (□) and wild-type congenic B10.PL (○) animals were immunized with the Golli-MBP protein J37. Ten days later, draining lymph node cells were harvested and cultured in serum-free medium with the indicated amounts of LDVM1–9. Golli-KO animals mounted robust responses to LDVM1–9, whereas lymphocytes derived from wild-type animals proliferated in response to LDVM1–9 only at higher concentrations (20 μg/ml) of antigen. The average background cpm values were 5,022 and 3,077 for wild-type and KO animals, respectively. No proliferation was observed to in vitro recall with Ac1–9 (data not shown).

Although Possessing a Poor Binding Affinity for I-A^u, Ac1–9 Is an Effective Inducer of Central Tolerance.

Previous studies have described a Vβ8.2Jβ2.7 Ac1–9-directed T cell response present in all B10.PL animals (4). To characterize the Ac1–9-specific immune response by using CDR3-length spectroscopy, B10.PL mice were immunized with Ac1–9, and their lymph nodes were removed 9 days later for analysis. This revealed two public expansions, the previously characterized Vβ8.2Jβ2.7 expansion and a novel Vβ8.2Jβ2.4 expansion, present in every animal. Both expansions contained a third complementarity determining region of 9 aa (Fig. 5). TCR affinity studies revealed that the DAGGGY TCR was of higher affinity than the Vβ8.2Jβ2.7 DASGGN TCR (E. S. Ward, personal communication). Additional studies indicated that although both of these clones are involved in the pathogenesis of passively induced EAE, the Vβ8.2Jβ2.7 DAGGGY clone dominates actively induced disease (P.v.d.E., E.M., D. Huffman, S. S. Wilson, V. Kumar, and E.E.S., unpublished data). The intermediate affinity of the DASGGN TCR makes this clone especially suited for studies involving central tolerance (results obtained with this clone cannot be discounted by the T cell having too low or too high an affinity for its ligand). A T cell clone expressing the novel Vβ8.2Jβ2.4 expansion has been previously isolated and used to prepare a TCR transgenic animal (10). This Vβ8.2Jβ2.4 transgenic animal develops spontaneous EAE and has been backcrossed onto the B10.PL background.

Figure 5.

Immunizing B10.PL mice with Ac1–9 results in the expansion of two public clonotypes. B10.PL mice were immunized with Ac1–9 and draining lymph nodes were removed 10 days later and parallel cell suspensions were cultured with Ac1–9 and medium alone. Immunoscope analysis revealed that cultures incubated with Ac1–9 had significant expansions of a Vβ8.2Jβ2.7 and a Vβ8.2Jβ2.4 clonotype. Sequence analysis of the Vβ8.2Jβ2.7 expansion revealed predominant CDR-3 sequences of GDAGGGY and GDAGGSY. Clonal analysis of the Vβ8.2Jβ2.4 expansion revealed a variety of CDR-3 sequences: GDAGSQN, GDASGGN, GDAGRQN, and GDAGAQN.

Because previous studies have suggested that, owing to its weak affinity for I-A^u, Ac1–9 was incapable of inducing central tolerance (3), we tested this hypothesis directly in fetal thymic organ culture (Fig. 6). Consistent with its ability to activate Ac1–9-specific cells, our preliminary studies show that when fetal thymi isolated from Ac1–9-specific TCR transgenic animals (10) were incubated with Ac1–9 for 8 h, an almost complete loss of CD4, CD8 double positive cells was observed (double positive cells are the thymic population most sensitive to negative selection; ref. 17). In contrast, incubation with LDVM1–9 resulted in no significant reduction in the number of these cells (Fig. 6A). To determine whether the reduction of double positive cells was caused by apoptosis or by some other phenomenon, we stained T cells isolated from parallel cultures with a marker for apoptotic cells, annexin V, and a marker for necrotic cells, PI. Roughly corresponding to the loss of double positive cells in experiment 3 (19% of the total thymocyte population), there was a comparable increase in the number of thymic annexin V-positive cells [(12.98 + 8.55) − (3.92 + 1.87) = ≈16%] (Fig. 6B). A significant increase was observed in the number of PI-negative/annexin V-positive and PI-positive/annexin-V positive fetal thymic T cells after a 6-h incubation with Ac1–9. Together, these data support the conclusion that even peptides possessing weak affinities for MHC class II molecules can be effective inducers of central tolerance provided that they are displayed in the thymus. These results are consistent with those obtained by Allen and colleagues, who demonstrated that certain antagonistic peptides could negatively select CD4⁺ T cells (18). In addition, the ability of Ac1–9 to negatively select Ac1–9-specific T cells has been confirmed by Lafaille and colleagues (personal communication). The differences between our study and that previously published by Wraith and colleagues (3) may be owing to the poorer affinity of the Ac1–9-specific T cells used in their study (19) or due to the fact that they delivered the Ag i.p. and then examined the thymus rather than directly looking at fetal thymic organ culture.

Figure 6.

Ac1–9 can be an effective inducer of negative selection, provided it is expressed in the thymus. Fetal thymi isolated from Ac1–9-specific TCR-tg B10.PL animals were cultured with or without Ac1–9. (A) Culture with Ac1–9 (1 μg/ml) for 8 h resulted in a dramatic reduction in the number of CD4⁺ CD8+ double positive thymocytes (P < 0.005 paired Student's t test). (B) To determine whether this loss was caused by apoptosis, thymic lobes from parallel cultures were stained with annexin V and propidium iodide. A significant increase in the number of annexin V-positive/PI-positive and annexin V-positive/PI-negative cell populations was observed after incubation with Ac1–9 for 6 h (P < 0.005 χ² test).

Our results are directly complementary to those described recently by Garcia and colleagues (11), Additionally, we provide direct evidence, by using LDVM1–9-specific T cells, for the existence of an N-terminal LDVM register in addition to the C-terminal register previously described (14), as well as demonstrating competition at the same site within the I-A^u groove. Our results also make use of the self (murine) sequence of MBP rather than the foreign guinea pig sequence used by Garcia and colleagues.

In summary, these results support a model in which competition for MHC loading is an event with crucial implications for central tolerance induction. The enforced lack of MBP:1–9 display protects even its highest affinity T cell population, the Vβ8.2Jβ2.7 DAGGGY T cell, from tolerance induction, permitting its survival in the periphery. It is precisely such high-affinity T cells that can be extremely dangerous when they become available for induction by Ac1–9 in mature MBP. We believe that this may be a general escape mechanism for driver clones, and unless regulatory mechanisms supervene, autoimmune disease may be difficult to avoid.

Abbreviations

EAE

experimental allergic encephalomyelitis

KO

knockout

MBP

myelin basic protein

TCR

T cell receptor

PI

propidium iodine