Negative selection of thymocytes expressing the D10 TCR

Abstract

We have analyzed the patterns of positive and negative selection of thymocytes expressing the T cell antigen receptor (TCR) from the D10.G4.1 T cell clone. This TCR confers reactivity to several non-self MHC class II alleles with a remarkably broad range of avidities. Therefore, negative selection can be studied when induced by high-, intermediate-, or low-avidity interactions with endogenous peptide–MHC complexes, all within the same TCR transgenic system. These data directly demonstrate that MHC class II–peptide ligands that fail to activate mature T cells can promote negative selection of immature thymocytes. Additionally, we show that negative selection of thymocytes can occur at two distinct “time points” during development depending on the avidity of the TCR for the MHC–peptide complex. Finally, we show that the self-peptide repertoire plays a significant role in selection because alteration of the self-peptide repertoire by disruption of the H2-Ma gene drastically alters selection of D10 TCR-expressing thymocytes.

The MHC is composed of the most polymorphic set of genes ever identified. Much of the allelic variation in MHC class II genes lies within the floor of the peptide-binding groove. The result of this variation is that each MHC allele is able to bind a different set of peptides (1). Theoretically, increasing the variety of peptides that can be displayed at the surface of antigen-presenting cells (APCs) should increase the likelihood that an effective immune response can be mounted. Because most organisms carry more than one allele at each MHC locus, a wide array of potentially antigenic peptides can be displayed and, presumably, contribute to the adaptive immune response. Indeed, an MHC heterozygote advantage has been documented in the extended survival of HIV-1-infected patients (2).

The tremendous variety of MHC alleles coupled with potential variations in self-peptides suggests that the repertoire of MHC–peptide complexes expressed by each individual within a species is unique. To accommodate this variety, the cumulative T cell antigen receptor (TCR) repertoire of a species must be extremely diverse. Within an individual, however, only a relatively small subset of the possible TCR specificities is required for the immune system to be competent. Indeed, TCR specificities that are useful for one individual may be useless or even harmful to another (3). All individuals of a species, however, carry the same or similar gene segments that recombine to generate complete TCR genes. Therefore, it is reasonable to propose that the immature TCR repertoire of each individual is initially similar, but during development in the thymus, this repertoire is molded to fit the needs of the individual (4).

Intrathymic development shapes the TCR repertoire by two processes described as positive and negative selection (5). Positive selection is believed to pick out useful TCR specificities on a matrix of self-peptide–self-MHC ligands presented on epithelial cells within the cortex of the thymus. The TCRs that will undergo positive selection are able to recognize self-MHC alleles with a low affinity (6). The specificity of TCRs that can be positively selected is also influenced by the repertoire of intrathymic self-peptides displayed by the MHC molecules (7–13).

Negative selection is induced by the same matrix of self-peptide–self-MHC ligands. Intrathymic negative selection is thought to remove developing thymocytes expressing TCRs with high avidities for endogenous peptide–MHC complexes, thereby helping to establish self-tolerance. The analysis of various TCR transgenic systems has led, however, to a rather confused understanding of the mechanics of the negative selection process. Differences in results obtained from these systems may be in part attributed to inherent differences in the TCRs used (e.g., avidities for endogenous MHC–peptide complexes), differences in transgene expression, and the means by which negative selection is promoted. Indeed, negative selection in these various systems has been induced by a variety of means including infection (14–16), injection (17–20), superantigens (21–23), or introduction of high levels of antigenic peptides (17, 24–29), but rarely simply by the presence of endogenous peptide–MHC ligands (30, 31). These systems complicate interpretations because the negatively selecting ligands are transiently expressed, expressed at artificially high levels, or presented by physiologically irrelevant cell types. Also, these systems have not examined the role that specific self-peptides play in negative selection. Additionally, much of this work was done with MHC class I-restricted TCRs.

Because negative selection of thymocytes is induced by TCR interactions with self-peptide–self-MHC complexes, examination of this process is best studied with a system in which negative selection is mediated by these endogenous peptide–MHC complexes. Such a system, however, would be most useful if the interaction of the TCR with MHC could be controlled across a wide range of avidities. The mice described here that were made transgenic for the genes expressing the TCR cloned from the D10.G4.1 (D10) T cell clone provide such a system. Of the eight MHC I-A alleles examined, only I-A^k induced positive selection of thymocytes expressing the D10 TCR. Six of these I-A alleles, however, promoted negative selection of D10, even though three of them did not stimulate proliferation of mature D10 T cells. Negative selection was found to occur at two distinct developmental points.

Experimental Procedures

Animals.

Transgenic mice were housed in specific pathogen-free environments at Yale and the Sloan-Kettering Institute. Other mice were purchased from The Jackson Laboratory. H2-Ma deficient mice were a gift from Martin and van Kaer, Vanderbilt University School of Medicine, Nashville, TN (32). D10 transgenic mice are similar to those described by Sant'Angelo et al. (33); however, the mice used for this article have the TCR α- and β-transgenes cointegrated into the same genomic site.

T Cell Proliferation Assays.

D10 T cells were harvested from the lymph nodes and spleens of 6- to 8-week-old I-A^k-expressing D10 transgenic, TCR Cα-negative mice. CD4⁺ T cells were enriched by negative selection by using antibodies against MHC class II (10.2.16) and CD8 (2.43) followed by incubation with anti-mouse IgG, anti-mouse IgM, and anti-rat IgG antibodies that were linked to BioMag magnetic beads (PerSeptive Biosystems, Framingham, MA). Labeled cells were then removed with a magnet. A sample of the cells was stained with anti-CD4 (Sigma) and 3D3, the clonotypic antibody against the D10 TCR, and analyzed by a fluorescence-activated cell sorter (FACS) to check their purity. Typical preparations yielded cell populations that were more than 90% CD4⁺3D3⁺. APCs were prepared by incubation of single-cell suspensions of splenocytes with antibodies against CD4, CD8, and Thy 1, followed by incubation with rabbit complement and mitomycin C for 30 min at 37°C. Mixed lymphocyte experiments were set up in triplicate in 96-well, round-bottom plates with 4 × 10⁴ D10 T cells and titrated numbers of APCs. Assays were pulsed with 1 μCi per well of [³H]thymidine after 48 h and were harvested after an additional 24 h. The experiments shown all were performed at the same time to reduce variation. T cell assays were performed at least two times and similar results were obtained.

FACS Analysis.

Single-cell suspensions were made by dissociation with glass slides. Cells were incubated in round-bottom plates on ice with antibodies for 30 min, washed, and, when necessary, incubated with streptavidin secondary reagents for 30 min on ice. Cells were washed and run a BD FacScan or LSR. Antibodies used were anti-CD4 quantum red, anti-CD8 PE (Invitrogen) and 3D3-biotin.

Results

Naïve CD4 T Cells from D10 TCR Transgenic Mice Respond to Non-Self MHC.

The D10.G4.1 (D10) T cell clone responds to a peptide derived from chicken conalbumin (CA-wt) bound to the MHC class II allele, I-A^k (34, 35). The D10 T cell clone also responds to a wide array of non-self MHC class II alleles including I-A^b,q,v,p&d (36, 37). D10 T cells do not proliferate in response to APCs expressing I-A^f,r,s&u (38, 39). Monoclonal anti-MHC class II antibodies block stimulation of D10 T cells by these various ligands.

We used our DNA constructs for the D10 TCR α- and β-chains (33) to generate transgenic mice with both genes cointegrated into the same genomic site. The responses to some of the alleles mentioned above by T cells isolated from D10 transgenic mice are shown in Fig. 1A. D10 T cells were harvested from H-2^k mice carrying a disruption in their germ-line TCR Cα genes (40). The lack of endogenous TCR α-chains, coupled with strong allelic exclusion by the D10 TCR β-chain transgene results in nearly all of the isolated T cells expressing only the D10 TCR. Strong proliferation of the predominantly naive D10 transgenic T cells occurs when incubated with titrated numbers of T cell-depleted, mitomycin C-treated splenocytes from C57BL/6 (I-A^b) or B10.G (I-A^q) mice. A minor, but notable, response to B10.D2 (I-A^d) splenocytes was also seen. The responses of D10 T cells to I-A^f,r,s&u were negligible.

Figure 1.

The response of naïve D10 T cells to non-self MHC. T cells (5 × 10⁴) purified from D10 TCR transgenic, TCR Cα-deficient mice were incubated for 72 h in the presence of titrated numbers of T-depleted, mitomycin C-treated APCs expressing the indicated MHC I-A alleles. Responses to B10.BR splenocytes that express the restricting MHC allele I-A^k loaded with titrated amounts (10, 0.1, 0.01, and 0.001 μM) of CA-wt peptide are included for comparison. (A) Splenocytes homozygous for MHC class II; (B) splenocytes heterozygous for MHC class II.

The responses of the naïve D10 T cells to non-self MHC molecules were consistent, but not identical with those seen with the D10.G4.1 clone (ref. 36; data not shown). Overall, the same hierarchy of responses to non-self MHC alleles (I-A^b >I-A^q≫I-A^d) was found in both the naïve D10 T cells, the D10 clone, and D10 T cell lines. Also, the lack of or very limited responses to I-A^f,r,s&u is also consistent with the D10.G4.1 clone. For comparison, the response of transgenic D10 T cells to APCs expressing I-A^k plus titrated amounts of conalbumin (CA-wt) peptide are also shown in Fig. 1A. Overall, these data demonstrate that recognition of different non-self MHC alleles by the D10 TCR results in a wide range of proliferative responses.

Much of the data presented in later figures rely on mice that are heterozygous for I-A^k and one of the nonstimulating MHC alleles. It is possible that mixed chain pairing between I-A^k and the nonstimulating MHC allele creates a novel MHC molecule in such mice (41), and that this hybrid MHC allele is recognized by the D10 TCR. Fig. 1B demonstrates that this recognition does not occur. As in Fig. 1A, the response of the D10 T cells to I-A^k plus titrated amounts of CA-wt peptide is shown for comparison and also as a positive control for the experiment. The response of naïve D10 T cells to APCs heterozygous for I-A^k and one of the MHC alleles (I-A^b,q,d) that stimulated D10 was similar, but maximum responses were reduced (data not shown).

Efficient Negative Selection of Thymocytes Expressing the D10 TCR.

Intrathymic negative selection is believed to eliminate TCR specificities that are overtly self-reactive (3, 5). To examine this process, we generated mice that carry the transgenes for the D10 TCR, its restricting MHC molecule, I-A^k, and also an MHC allele known to stimulate D10. The presence of a selecting allele allows for the discrimination between negative selection and lack of positive selection. These mice all have wild-type TCR α and β loci and, thus, can potentially express other TCRs.

Positive selection of thymocytes expressing the D10 TCR is observed in 2-week-old B10.BR mice, which express I-A^k (Fig. 2A). FACS analysis with antibodies against CD4, CD8, and the anti-D10 TCR antibody, 3D3 (34), demonstrates a skewing of developing thymocytes toward the CD4 lineage in these mice. Electronically gating on 3D3^hi cells demonstrates that many of the mature thymocytes are CD4⁺. A significant number of 3D3⁺ double-negative (DN) cells are also present in D10 transgenic mice. Many transgenic mice have an increase in this cell type that may be caused by early TCR α-chain pairing with the β-chain transgene (42). Analysis of lymph node cells from D10 transgenic mice (Fig. 2B) also demonstrates skewing of the T cells toward the CD4 lineage. When only 3D3⁺ cells are examined, the skewing is even more pronounced. Most, but not all, of the CD8⁺ 3D3⁺ T cells are likely to express a second TCR α-chain because their frequency in D10 transgenic mice bred to Cα knockout mice is much lower (data not shown).

Figure 2.

FACS analysis of thymocytes and lymph node cells from mice carrying the D10 TCR transgenes expressing MHC alleles that stimulate D10 T cells. Single-cell suspensions of (A) thymocytes and (B) lymph node cells from 2-week-old mice expressing the indicated MHC I-A alleles were stained with CD4, CD8, and the anti-D10 clonotypic antibody, 3D3. FACS plots show either the total cell population or are electronically gated on 3D3^hi thymocytes or 3D3⁺ T cells. Numbers within the plots are the percentage of cells that fall into each quadrant. Percentages in 3D3-gated plots represent only the positive cells and not the total population. Note that these mice are able to express endogenous TCR α- and β-chains.

Massive intrathymic negative selection is apparent in 2-week-old D10 transgenic mice that are heterozygous for the MHC class II alleles I-A^b and I-A^k. As can be seen in Fig. 2A, almost no CD4⁺CD8⁺ (double-positive, DP) cells develop in these D10/I-A^b/k mice and the few 3D3^hi cells are nearly all CD4⁻CD8⁻ (DN). Equally as striking was the reduction in thymic cellularity. The total cell number dropped from more than 50 × 10⁶ cells in D10/I-A^k mice to 6 × 10⁶ cells in the D10/I-A^b/k mice. Indeed, 2-week-old mice were used because the thymus in older mice was extremely difficult to dissect reliably. Almost no 3D3⁺CD4⁺ lymphocytes could be detected (Fig. 2B). Similar results were obtained from mice expressing I-A^q (I-A^q/k) and, somewhat surprisingly, from mice expressing the weakly reactive I-A^d (I-A^d/k) allele. In both cases, DP cells were nearly absent, as were peripheral 3D3⁺CD4⁺ T cells (Fig. 2B). Thymic cellularity was also greatly diminished in these mice (Table 1).

Table 1.

Average total thymocyte and lymphocyte cell number from two to four animals of each genotype

Strain	Thymocytes	Lymph nodes
2 weeks
B10.BR (k/k)	51  × 106	4.7  × 106
C57Bl/6 (b/k)	6  × 106	5.4  × 106
B10.G (q/k)	9.3  × 106	4.8  × 106
B10.D2 (d/k)	11  × 106	9  × 106
4 weeks
B10.BR (k/k)	197  × 106	46  × 106
B10.M (f/k)	191  × 106	24  × 106
B10.RIII (r/k)	106  × 106	33  × 106
B10.S (s/k)	123  × 106	36  × 106
B10.PL (u/k)	97  × 106	44  × 106
C57Bl/6 (b/b)	9.1  × 106	27  × 106
H-2Ma−/− (b/b)	62  × 106	36  × 106

MHC Alleles That Do Not Stimulate Mature D10 T Cells Can Induce Negative Selection of Thymocytes Expressing the D10 TCR.

We continued our evaluation of negative selection in D10 TCR transgenic mice by examining the impact of MHC alleles that do not appreciably stimulate proliferation of mature D10 T cells as shown in Fig. 1A. For these experiments, mice were killed at 4 weeks of age. Positive selection of D10 thymocytes in 4-week-old I-A^k mice was similar to that seen in the 2-week-old mice used above as determined by FACS analysis (Fig. 3A). One significant difference was in the percentage of 3D3⁺ DN thymocytes, which was reduced as compared with the 2-week-old mice. Also, the total numbers of both thymocytes and lymphocytes was increased in the 4-week-old D10/B10.BR mice (Table 1).

Figure 3.

FACS analysis of thymocytes and lymph node cells from mice carrying the D10 TCR transgenes expressing nonstimulatory MHC alleles. Single-cell suspensions of the thymocytes and lymph node cells from 4-week-old mice expressing the indicated MHC I-A alleles were stained with CD4, CD8, and the anti-D10 clonotypic antibody, 3D3. FACS plots show either the total cell population or are electronically gated on 3D3^hi thymocytes or 3D3⁺ T cells. Numbers within the plots are the percentage of cells that fall into each quadrant. Percentages in 3D3-gated plots represent only the positive cells and not the total population. Note that these mice are able to express endogenous TCR α- and β-chains.

D10 transgenic mice carrying the I-A^f/k MHC alleles fail to develop CD4 single-positive (SP) thymocytes (Fig. 3A). The lack of CD4 SP thymocyte development must be caused by negative selection mediated by the nonstimulating I-A^f allele. Peripheral CD4⁺ T cells expressing the D10 TCR are also essentially absent in these mice (Fig. 3B). In sharp contrast to negative selection mediated by the I-A^b,q,&d alleles, however, the development of DP cells is essentially normal. Additionally, thymic cellularity is not noticeably reduced (Table 1).

Similar results were obtained from D10 transgenic mice expressing the I-A^r and I-A^u alleles (Fig. 3). Again, DP cells develop and are 3D3 positive, but CD4⁺3D3⁺ cells are absent (Fig. 3A). Thymic cellularity in these mice is reduced by about half (Table 1). Closer examination of the proliferation assay in Fig. 1 suggests that D10 T cells may actually weakly respond to these “nonstimulatory” alleles (an increase over stimulation with k/k expressing APCs of ≈1,000–4,500 cpm). However, equivalent proliferation in response to I-A^s was seen; and this allele, as shown in Fig. 3, does not participate in negative selection of D10. Lack of negative selection by I-A^s is clear, because 3D3⁺CD4⁺ thymocytes (Fig. 3A) and lymphocytes (Fig. 3B) accumulate in these I-A^s/k heterozygous mice.

Although the I-A^s allele does not negatively select D10, it is possible that I-A^s positively selects thymocytes expressing this TCR. To evaluate the specificity of positive selection of D10, we further backcrossed the D10 I-A^s/k mice to B10.S mice to produce D10 I-A^s/s homozygous mice. FACS analyses of single-cell suspensions of the thymus and lymphocytes from 4-week-old mice of this genotype are shown in Fig. 4. Few, if any, 3D3⁺CD4⁺ thymocytes or lymphocytes can be detected. Therefore, I-A^s neither negatively nor positively selects thymocytes expressing the D10 TCR. Thus, the coexpression of I-A^s and I-A^k has little, if any, effect on positive selection.

Figure 4.

Thymocytes expressing the D10 TCR are neither positively nor negatively selected in mice homozygous for I-A^s. Staining was performed as in Fig. 3 with antibodies against CD4, CD8, and the D10 TCR (3D3 clonotypic antibody). The few thymocytes that progress to the CD4 SP stage are not positive for 3D3 and, therefore, these cells must be expressing endogenous TCR α-chains.

Alteration of the Intrathymic Self-Peptide Repertoire Alters Negative Selection.

Although self-peptides seem to be critical for intrathymic positive selection (refs. 7–9, 13, 24, and 43; most recently reviewed in refs. 44 and 45), the role of self-peptides in negative selection is not as clear. We took advantage of the alloreactivity of D10 to the I-A^b allele and the availability of H2-Mα-deficient mice (32) to address this question. The predominant MHC–peptide complexes expressed in H-2Mα-deficient mice are I-A^b loaded with a set of peptides known as CLIP (class II invariant chain peptides), which are derived from the MHC class II invariant chain (32, 46, 47). The recognition of I-A^b by the D10 TCR is clearly influenced by the peptide bound within the groove, because APCs from H-2Mα-deficient mice fail to activate D10 T cells strongly as compared with APCs from wild-type C57BL/6 mice (Fig. 5A). The weak response of D10 T cells to high numbers of H-2Mα-deficient APCs is highly reproducible. This residual response suggests that either D10 T cells recognize the I-A^b–CLIP complex with a low avidity or that D10 T cells are responding to the presence of one or more endogenous peptides presented independent of the function of H-2M.

Figure 5.

APCs expressing predominantly I-A^b–CLIP complexes do not efficiently stimulate D10 T cells and alter intrathymic negative selection. (A) Stimulation of 4 × 10⁴ naïve T cells from D10 TCR transgenic TCR Cα^−/− mice with titrated numbers of T cell-depleted, mitomycin C-treated splenocytes from H2-Ma-negative mice or wild-type C57BL/6 mice. The small response of the D10 T cells when stimulated by high numbers of H-2Ma-negative APCs is reproducible. FACS analysis of (B) thymocytes and (C) lymph node cells from D10⁺ I-A^b/b H-2Ma^−/− and D10 TCR I-A^b/b transgenic mice. Cells were stained with antibodies against CD4, CD8, and the anti-D10 TCR clonotypic antibody, 3D3. Numbers are the percentage of cells within each quadrant.

To investigate directly the role of self-peptides in negative selection, we introduced the D10 TCR transgenes into H-2M-deficient mice by breeding. These D10⁺ I-A^b/b H-2Ma^−/− mice were compared with similarly aged D10⁺ I-A^b/b mice. The thymuses taken from the D10⁺ I-A^b/b H-2Ma^−/− mice had approximately six times as many cells as the D10⁺ I-A^b/b mice (Table 1), which immediately suggested that negative selection of the D10-expressing thymocytes was impacted by the altered self-peptide repertoire expressed in these mice. FACS analyses confirmed the change in negative selection because a near-normal percentage of 3D3^lowCD4⁺CD8⁺ thymocytes were seen in D10⁺I-A^b/b H-2Ma^−/− mice as compared with D10⁺I-A^b/b mice, in which the DP population was reduced to nearly zero (Fig. 5B). The SP thymocytes and the CD4 and CD8 T cells found among the lymph node cells in the D10⁺ I-A^b/b mice were nearly all negative for the clonotypic 3D3 anti-D10 TCR antibody (Fig. 5 B and C). These T cells presumably have escaped negative selection by eliminating expression of the D10 TCR either by deletion of the transgene (48) or some type of receptor editing (28). These non-D10 T cells are more prominent in these 6-week-old mice as compared with the 2-week-old D10 I-A^b/k mice shown in Fig. 2, presumably because of T cell expansion, as is also seen in older D10 I-A^b/k mice. What is clear, however, is that early negative selection of D10 TCR-expressing thymocytes is directly impacted by the repertoire of expressed self-peptides.

Thymic cellularity (62 × 10⁶) was reduced 3-fold in D10⁺ I-A^b/b H-2Ma^−/− mice as compared with 4-week-old I-A^k D10 TCR transgenic mice (197 × 10⁶) and 2-fold as compared with D10 transgenic mice expressing the nonselecting I-A^s/s allele (123 × 10⁶) as shown in Table 1. These data suggest that the lack of 3D3⁺CD4⁺ T cells in these mice may be caused by negative selection during the DP-to-SP transition because of interactions with residual, non-CLIP self-peptides that are expressed in H-2Mα-deficient mice. However, because the selecting I-A^k allele is not present, we cannot eliminate the possibility that the lack of SP cells is caused by a failure of positive selection.

Discussion

We have examined the specificity of positive and negative selection of thymocytes expressing the TCR from the D10.G4.1 T cell clone (34). The D10 TCR confers reactivity to a wide range of non-self MHC class II molecules (36, 38, 39); therefore, it provides a unique system to study intrathymic negative selection mediated by endogenous MHC–peptide complexes. Negative selection of thymocytes expressing transgenic TCRs has been observed in several different systems (refs. 49–53, for example) and recently reviewed in refs. 54 and 55. In most of these systems, negative selection was reported to occur before the CD4⁺CD8⁺ stage. It has been suggested that such early negative selection may be caused by higher than normal levels of TCR expression on immature thymocytes (56). Recently, elegant work by Davis and colleagues (26) demonstrated that negative selection of antigen-specific T cells occurs throughout thymic development. By using MHC tetramers to track moth cytochrome c (MCC peptide)-specific T cells, this group was able to detect negative selection even in DN thymocytes. Negative selection in this system, however, relied on the de novo expression of a known agonist peptide.

Consistent with our previous work (27), our current data support a model in which negative selection primarily depends on the avidity of the TCR for the available self-peptide–self-MHC complexes rather than depending on the maturation state of the developing thymocyte. The MHC class II alleles I-A^b, I-A^q, and I-A^d vary widely in their stimulation of naïve D10 T cells with I-A^b being very strong, I-A^q being somewhat less strong, and I-A^d being much weaker. All three of these alleles induced very early negative selection of developing D10 thymocytes. Negative selection of D10 thymocytes also occurred in mice expressing the apparently nonstimulatory MHC alleles I-A^f, I-A^r, and I-A^u. In sharp contrast to I-A^b,q&d, negative selection induced by I-A^f,r&u occurred after the CD4⁺CD8⁺ DP stage of development. The failure to progress from the DP to SP stage was not caused by a lack of positive selection, because in all cases the mice were also expressing the selecting I-A^k allele.

Our data also provide a direct demonstration of the exquisite sensitivity of developing thymocytes as compared with their mature T cell counterparts. The I-A^f,r&u MHC class II alleles do not appreciably stimulate mature D10 T cells, but these alleles are capable of mediating efficient negative selection of D10 TCR-expressing thymocytes. Our data are consistent with findings from other groups (14, 57–59), but in experiments done by those groups, activation of immature thymocytes was done by infection with viral variants (14) or in vitro incubation of thymocytes with the addition of exogenous peptide (57, 58) or superantigen (59). The data presented here represent an in vivo demonstration of the enhanced sensitivity of developing thymocytes to endogenous MHC–peptide complexes. Differences in the presented peptide antigens might account for some of the differences in activation of D10 T cells and thymocytes. However, in vitro assays with thymic APCs harvested from mice with “nonstimulatory” MHC alleles did not stimulate mature D10 T cells (data not shown).

We also suggest that high levels of ligand (self-peptide–self-MHC complexes) expression, presumably on cortical epithelial cells, may mediate early (DN to DP) negative selection. The later (DP to SP) negative selection is likely to be a result of thymocyte interactions with dendritic cells. This conclusion is supported by studies that used H2-Mα-deficient mice that are transgenic for the D10 TCR. These mice predominantly express MHC class II I-A^b loaded with CLIP peptides (32). In these mice, DP thymocytes develop, but the transition to SP is blocked. The substantial reduction in thymic cellularity in these mice, as compared with the thymus from mice with positively or nonselecting alleles, suggests that this block is caused by negative selection, which, however, cannot be concluded directly because the selecting I-A^k allele is not present in the D10⁺I-A^b/b H-2Ma^−/− mice. Regardless of the interpretation of the block in the transition from DP to SP, these data clearly demonstrate that negative selection, similar to positive selection, can be influenced by specific intrathymic self-peptides, contrary to the idea that the specificity of the T cells that are negatively selected is primarily directed toward MHC contact points.

Several important conclusions can be drawn from these data. First, negative selection can occur at least two different times during development, either before or after the DP stage. Negative selection before the DP stage implies that cells other than medullary dendritic cells, perhaps cortical epithelial cells, can mediate deletion. Second, the alteration of negative selection seen in the H-2Ma^−/− mice demonstrates that negative selection depends, at least in part, on specific recognition of self-peptides. Our data also demonstrate that negative selection eliminates TCR specificities that are not overtly self-reactive. Finally, our data provide direct in vivo support of other work suggesting that immature thymocytes are more sensitive to activation than their mature counterparts (57–59).

Abbreviations

TCR

T cell antigen receptor

APCs

antigen-presenting cells

FACS

fluorescence-activated cell sorter

CLIP

class II invariant chain peptides

SP

single-positive

DP

double-positive

DN

double-negative