Insertion of the dibasic motif in the flanking region of a cryptic self-determinant leads to activation of the epitope-specific T cells

Efficient induction of self tolerance is critical for avoiding autoimmunity. The T cells specific for the well-processed and -presented (dominant) determinants of a native self protein are generally tolerized in the thymus, whereas those potentially directed against the inefficiently processed and presented (cryptic) self epitopes escape tolerance induction. We examined whether the crypticity of certain determinants of mouse lysozyme-M (ML-M) could be attributed to the nonavailability of a proteolytic site, and whether it could be reversed to immunodominance by engraftment of a novel cleavage site in the flanking region of the epitope. Using site-directed mutagenesis, we created the dibasic motif (RR or RK; R = arginine, K = lysine), a target of intracellular proteases, in the region adjoining one of the three cryptic epitopes (46-61, 66-79, or 105-119) of ML-M. Interestingly, the mutated lysozyme proteins, but not unmutated ML-M, were immunogenic in mice. The T cell response to the altered lysozyme was attributable to the efficient processing and presentation of the previously cryptic epitope, and this response was both epitope and MHC haplotype specific. In addition, the anti-self T cell response was associated with the generation of autoantibodies against self lysozyme. However, the results using one of three mutated lysozymes suggested that the naturally processed, dibasic motif-marked epitope may not always correspond precisely to the cryptic determinant within a synthetic peptide. This is the first report describing the circumvention of self tolerance owing to the targeted reversal of crypticity to dominance in vivo of a specific epitope within a native self Ag.     

The self-directed T cell repertoire against mouse lysozyme reflects the influence of the hierarchy of its own determinants and can be engaged by a foreign lysozyme.

The T cell repertoire is shaped by the processes of positive and negative selection. We have previously shown that mice are tolerant to a native self-Ag, mouse lysozyme (ML), but they respond vigorously when challenged with different ML peptides ("cryptic" self-determinants). In this study, we have addressed the issue of the physiological significance of both the hierarchy (dominance/crypticity) of self-determinants within ML and the anti-cryptic, self (ML)-directed T cell repertoire. Our results demonstrate that there are several ML peptides that bind well to MHC but are totally nonimmunogenic when tested for proliferative T cell response and cytokine secretion: a subset of these peptides presumably represent the originally dominant self-determinants of ML, which have rendered the T cells tolerant during thymic selection. Other ML peptides, which bind well to MHC and are immunogenic, correspond to the cryptic determinants of ML: T cells against cryptic ML determinants escape tolerance induction. Thus, the mature T cell repertoire against ML bears the direct imprint of the hierarchy of self (ML)-determinants. Interestingly, hen egg white lysozyme could prime T cells in vivo that were cross-reactive with certain cryptic ML determinants, and vice versa, without requiring any coimmunization with the foreign lysozyme and ML peptide(s). Moreover, repeated, deliberate priming and expansion of T cells by hen egg white lysozyme immunization concomitantly enhanced T cell response to such cross-reactive ML determinants. This reciprocal self-foreign determinant cross-reactivity may play a previously unrecognized, but crucial, role in the expansion and diversification of self-reactive clones in the autoimmune response.     

T cells that recognize peptide sequences of self MHC class II molecules exist in syngeneic mice.

T cell reactivity toward self MHC class II molecules has been recognized in syngeneic MLR in a number of studies, where the T cells are believed to recognize the combination of self/nonself peptide and self MHC molecule. We investigated the stimulation of T cell proliferation by synthetic peptides of sequences corresponding to the first polymorphic amino terminal domain of alpha- and beta-chains of self I-A molecules. Both unprimed and primed T cells responded to a number of peptides of alpha 1 and beta 1 domains of self I-Ad molecules. The response was dependent on the presentation of I-Ad peptides by syngeneic APC and was blocked by anti-class II MHC mAb. Upon further investigation it was observed that I-Ad peptides could inhibit the stimulation of Ag-specific MHC class II-restricted T cell hybridoma due to self presentation of peptides rather than to direct binding of free peptides to the TCR, further supporting their affinity/interaction with intact self MHC class II molecules. The peptide I-A beta d 62-78 showed high affinity toward intact self MHC II molecule as determined by the inhibition of Ag-specific T cell stimulation and yet was nonstimulatory for syngeneic T cells, therefore representing an MHC determinant that may have induced self tolerance. Thus we have shown that strong T cell proliferative responses can be generated in normal mice against the peptides derived from self MHC class II molecules and these cells are part of the normal T cell repertoire. Therefore complete tolerance toward potentially powerful immunodominant but cryptic determinants of self Ag may not be necessary to prevent autoimmune diseases.     

Characterization of CD8+ cytotoxic T lymphocyte/tumor cell interactions reflecting recognition of an endogenously expressed murine wild-type p53 determinant

p53 mutations are frequently found in human cancers and are often associated with the overexpression of wild-type (WT) protein or peptide sequences, supporting the notion that WT p53 epitopes may serve as potential targets for tumor immunotherapy. We have developed a cytotoxic T lymphocyte (CTL)/p53 tumor-associated antigen (TAA) model, based on immune recognition of a WT p53 determinant. WT p53-peptide-specific, major histocompatibility complex (MHC) classI-restricted CTL were produced from immunocompetent C57BL/6 (H-2b) mice after immunization with a previously defined WT p53 peptide (p53(232-240)) Epitope-specific CTL were then employed to identify syngeneic tumor cell populations expressing that antigenic determinant. Two syngeneic tumor cell lines, MC38 colon carcinoma and MC57G fibrosarcoma, were demonstrated to express the endogenous WT p53(232-240) determinant naturally, as defined by CD8 + CTL recognition. Cold-target inhibition assays confirmed that CTL-mediated lysis was due to immune recognition of the p53(232-240) peptide epitope. The p53(232-240)-specific CTL line did not lyse syngeneic normal cells (i.e., mitogen-activated splenocytes) in the absence of exogenous peptide, suggesting that the WT-p53-specific CTL could distinguish between tumor cells expressing self-TAA and normal host cells. We have demonstrated, for the first time, that the adoptive transfer of WT-p53-specific CTL to mice with established pulmonary metastasis resulted in antitumor activity in vivo. The ability to generate MHC-class-I-restricted CD8- CTL lines specific for a non-mutated p53 determinant from normal, immunocompetent mice, which display antitumor activity both in vitro and in vivo (by adoptive transfer), may have implications for the immunotherapy of certain p53-expressing malignancies.     

Distinction between immunogenicity and tolerogenicity among HBcAg T cell determinants. Influence of peptide-MHC interaction

One purpose of this study was to examine the concept of T cell immunodominance employing a neonatal tolerance model. The extent to which a single T cell recognition site can represent the total T cell response to hepatitis B core Ag (HBcAg) was examined in the B10.S and B10 murine strains. It was shown that the entire B10.S T cell response to HBcAg was focused on a single immunodominant site represented by residues 120-131. This was demonstrated by exposing B10.S neonatal mice to p120-140 or p120-131, which resulted in a state of T cell tolerance to the entire HBcAg. In contrast, p120-140 contained an immunogenic T cell site for B10 mice, p129-140, but this site was nontolerogenic. Similarly, injection of p120-140 into (B10.S X B10)F1 neonatal mice resulted in tolerization of p120-131-specific, I-As-restricted T cells, but not of p129-140-specific, I-Ab-restricted T cells. The second purpose of this study was to attempt to explain the immunologic basis of an immunogenic yet nontolerogenic T cell determinant. It was shown that the p120-131 T cell site, which is immunogenic and tolerogenic in B10.S mice, could be converted into an immunogenic/nontolerogenic T cell site by a single amino acid substitution in either residue 127 or 129. Residues 127 and 129 were previously shown to be involved in interaction with MHC class II molecules (agretopic). These results demonstrated that the relative avidity of a peptide-MHC interaction can influence T cell tolerance induction. Furthermore, the results suggest that a higher threshold of peptide-MHC avidity may be required to induce T cell tolerance as compared to the threshold of peptide-MHC avidity required to immunize T cells.     

Immunogenicity and arthritogenicity of recombinant CB10 in B10.RIII mice

Two major T cell determinants are recognized by I-Ar-specific T cells in CII, the immunodominant CII610-618 (GPAGT AGA R) within CB10 and the subdominant CII445-453 (GPAGP AGE R) within CB8. Although the determinants differ by only two residues, CB8 is capable of inducing collagen-induced arthritis (CIA), while CB10 is not. We, therefore, investigated the structural differences between the two determinants that are critical to inducing arthritis. When the CB10 determinant was mutated to that of CB8 using recombinant techniques, the resulting mutant rCB10T614P,A617E product became arthritogenic. Conversely, when the CB8 determinant was mutated to that of CB10, the resulting mutant CB8P449T,E452A was no longer arthritogenic. Comparison of the epitope specificity of the autoantibodies induced by wild-type CB10 and mutant rCB10T614P, A617E revealed no qualitative differences. T cells from mice immunized with either CB10 or mutant rCB10 produced predominantly Th1 cytokines when cultured with the immunizing Ag. In contrast, when cultured with mouse CII, T cells from mice immunized with the nonarthritogenic CB10 produced predominantly Th2 (IL-4 and IL-10) cytokines whereas the arthritogenic mutant rCB10 induced predominantly Th1 (IFN-gamma) cytokines. We conclude that the T cell cytokine response most critical for the induction of CIA is that induced against the corresponding homologous murine T cell determinant and, further, that the structural differences between the T cell determinants in CB8 and -10 are important in breaking self tolerance and inducing autoimmune response.     

Analysis of TCR V beta gene usage and encephalitogenicity of myelin basic protein peptide p91-103 reactive T cell clones in SJL mice: lack of evidence for V gene hypothesis.

We have analyzed the epitope specificity and encephalitogenicity of peptides that span the C terminus of MBP, p84-103. Our studies show that multiple antigenic epitopes with disease-inducing capacity exist in SJL mice. Three peptides that span this region were examined and found to be immunogenic. However, the mode of immunization (active or passive) determined the incidence and severity of EAE. In our experiments adoptive transfer of p91-103-reactive T cell lines was most consistent in the development of disease. Interestingly, the response to peptides p89-101, p91-103, and p84-102 was absent following immunization with MBP. This suggests that although p91-103 and p89-101 were encephalitogenic they were not the major immunogenic epitopes following immunization with MBP. Analysis of a panel of eight p91-103-reactive T cell clones showed significant heterogeneity in the fine specificity, the TCR V beta gene usage, and in their ability of transfer EAE. These studies suggest that in SJL mice the epitopes involved in the pathogenesis of disease are multiple and there is no clear correlation between encephalitogenicity and TCR V beta gene usage. These observations argue against the presence of a dominant TCR V beta gene in the pathogenesis of EAE in SJL mice.     

CD4+ T cell responses to self- and mutated p53 determinants during tumorigenesis in mice

We analyzed CD4+ T helper responses to wild-type (wt) and mutated (mut) p53 protein in normal and tumor-bearing mice. In normal mice, we observed that although some self-p53 determinants induced negative selection of p53-reactive CD4+ T cells, other p53 determinants (cryptic) were immunogenic. Next, BALB/c mice were inoculated with J774 syngeneic tumor cell line expressing mut p53. BALB/c tumor-bearing mice mounted potent CD4+ T cell responses to two formerly cryptic peptides on self-p53. This response was characterized by massive production of IL-5, a Th2-type lymphokine. Interestingly, we found that T cell response was induced by different p53 peptides depending upon the stage of cancer. Mut p53 gene was shown to contain a single mutation resulting in the substitution of a tyrosine by a histidine at position 231 of the protein. Two peptides corresponding to wt and mutated sequences of this region were synthesized. Both peptides bound to the MHC class II-presenting molecule (Ed) with similar affinities. However, only mut p53.225-239 induced T cell responses in normal BALB/c mice, a result strongly suggesting that high-affinity wt p53.225-239 autoreactive T cells had been eliminated in these mice. Surprisingly, CD4+ T cell responses to both mut and wt p53.225-239 peptides were recorded in J774 tumor-bearing mice, a phenomenon attributed to the recruitment of low-avidity p53.225-239 self-reactive T cells.     

T cells from nonresponder mice. MHC restricted and unrestricted recognition of insulin

Two types of insulin-reactive T cell hybridomas expressing TCR-alpha beta were derived from nonresponder H-2b mice immunized with pork insulin. One type had characteristics of conventional class II-restricted Th cells. These CD4+ CD8- I-Ab-restricted T cells recognized a self determinant, present within the insulin B-chain. This determinant was distinct from the immunodominant A-chain loop determinant that is recognized by the majority of T cells induced after immunization with normally immunogenic beef insulin. Our results suggest that this determinant is readily generated during immunologic processing of insulins, including nonimmunogenic pork insulin and self insulin. A second type of T cell lacking CD4 and CD8 recognized a distinct B-chain determinant of insulin in a class II-dependent, but MHC unrestricted, fashion. These cells may represent a novel subpopulation which has bypassed conventional selection during development in the thymus.     

Identification and characterization of a T cell-inducing epitope of bovine ribonuclease that can be restricted by multiple class II molecules.

An immunodominant epitope of bovine RNase restricted by I-Ek molecules was identified using a T cell hybridoma recognizing RNase. This epitope was localized to the peptide RNase(90-105). Single conservative amino acid substitutions were made at each of the positions 94 through 105. It was found that only at one position, Asn-103, were conservative substitutions not allowed. This residue was shown to be the critical residue in determining T cell specificity. The ability of RNase(90-105) and the well-defined T cell epitope, HEL(46-61) to stimulate mouse strains expressing different independent H-2 haplotypes was examined using a T cell proliferation assay. The response to HEL(46-61) was completely restricted to mice expressing an I-Ak molecule. In striking contrast, 6 of 10 different mouse strains, H-2b,f,k,q,s,u, mounted vigorous T cell responses to RNase(90-105). The response was restricted to both I-A and I-E molecules, including I-Ab, I-Af, I-Ek, I-Aq, and I-As. H-2d mice were nonresponders to RNase(90-105), which was shown to be due to the failure of RNase(90-105) to bind to I-Ad molecules. A variant RNase(90-105) peptide was generated, containing an I-Ad binding motif, that could bind to I-Ad molecules. Despite its ability to bind, this variant peptide was not able to stimulate a response in H-2d mice. This result demonstrates that the ability of a peptide to bind to an Ia molecule is necessary but not always sufficient for a response to occur. Thus, in contrast to the highly restricted HEL(46-61) determinant, the RNase(90-105) determinant is permissive in its binding to Ia molecules. These results show that in the universe of T cell inducing epitopes contains both highly restricted and broadly restricted epitopes are found.     

T cell response to preproinsulin I and II in the nonobese diabetic mouse

Immunization against insulin, insulin B chain, or B chain peptide B(9-23) (preproinsulin peptide II(33-47)) prevents diabetes in the nonobese diabetic (NOD) mouse. Whether or not peptide II(33-47) is the only proinsulin determinant recognized by CD4 T cells remains unclear. Using two peptide libraries spanning the entire sequence of preproinsulin I and preproinsulin II, respectively, we identified T cells specific for four proinsulin epitopes within the islet cell infiltrate of prediabetic female NOD mice. These epitopes were among immunogenic epitopes to which a T cell response was detected after immunization of NOD mice with individual peptides in CFA. Immunogenic epitopes were found on both isoforms of insulin, especially proinsulin II, which is the isoform expressed in the thymus. The autoimmune response to proinsulin represented only part of the immune response to islet cells within the islet cell infiltrate in 15-wk-old NOD mice. This is the first systematic study of preproinsulin T cell epitopes in the NOD mouse model.     

The Dermatophagoides pteronyssinus group 2 allergen contains a universally immunogenic T cell epitope

The use of T cell epitope-containing peptides for the induction of anergy in allergen sensitization is limited by genetic restriction that could be circumvented by using universally immunogenic epitopes. We attempted to identify such epitopes on Dermatophagoides pteronyssinus group 2 allergen (Der p 2), a major allergen of D. pteronyssinus T cells from BALB/c (H-2(d)), C57BL/6 (H-2(b)), C3H (H-2(k)), and SJL (H-2(s)) mice that were immunized with rDer p 2, recognized an immunodominant region encompassing residues 21-35. A synthetic 21-35 peptide (p21-35) induced strong dose-dependent in vitro T cell proliferation with cells of the four mouse strains and required processing for MHC class II presentation. Substitution of Ile(28) with Ala resulted in reduction of T cell proliferation in each strain. Ile(28) could represent an important MHC class II anchoring residue for T cell response to p21-35. An immunodominant T cell epitope of Der p 2 therefore behaves as a universal epitope and could be a suitable candidate for T cell anergy induction.     

The T cells specific for the carboxyl-terminal determinants of self (rat) heat-shock protein 65 escape tolerance induction and are involved in regulation of autoimmune arthritis

Immunization of Lewis rats with heat-killed Mycobacterium tuberculosis H37Ra leads to development of polyarthritis (adjuvant-induced arthritis; AA) that shares several features with human rheumatoid arthritis (RA). Immune response to the 65-kDa mycobacterial heat-shock protein (Bhsp65) is believed to be involved in induction of AA as well as in experimental modulation of this disease. However, the understanding of several critical aspects of the pathogenesis of AA in the Lewis rat has severely been hampered by the lack of information both regarding the level as well as epitope specificity of tolerance to the mammalian self (rat) homologue of Bhsp65, 65-kDa rat heat-shock protein (Rhsp65), and about the functional attributes of the T cell repertoire specific for this self protein. In this study, we established that tolerance to Rhsp65 in the Lewis rat is incomplete, and that the residual T cells primed upon challenge with this self hsp65 are disease regulating in nature. We also have defined the T cell epitopes in the C-terminal region within Rhsp65 that contribute predominantly to the immune reactivity as well as the AA-protective effect of this self protein. Furthermore, the T cells primed by peptides comprising these C-terminal determinants can be efficiently restimulated by the naturally generated epitopes from endogenous Rhsp65, suggesting that self hsp65 might also be involved in natural remission from acute AA. These novel first experimental insights into the self hsp65-directed regulatory T cell repertoire in AA would help develop better immunotherapeutic approaches for autoimmune arthritis.     

T-cell recognition of differentially tolerated epitopes of cartilage proteoglycan aggrecan in arthritis

Proteoglycan (PG) aggrecan, a major macromolecular component of cartilage, is highly immunogenic; it induces arthritis in genetically susceptible BALB/c mice. The present study maps the T-cell epitope repertoire of cartilage PG by identifying a total of 27 distinct T-cell epitopes. An epitope hierarchy, accounting for the different effector functions of PG-specific T cells, and determinant spreading, has been found. T-cell responses to four epitopes were associated with arthritis induction. Some of the T-cell epitopes were full T-cell activators, whereas a number of subdominant and cryptic epitopes proved to be partial activators in vitro, inducing either cytokine secretion or T-cell proliferation, but not both. A few T-cell epitopes of the core protein of cartilage PG were clearly recognized by T cells in PG-immunized arthritic animals, but the corresponding peptides did not induce T-cell responses when injected into naive BALB/c mice; thus these T-cell epitopes were designated as "conditionally immunogenic."     

Identification of cross-reactive peptides using combinatorial libraries circumvents tolerance against Her-2/neu-immunodominant epitope

The majority of the currently defined tumor-associated Ags are often overexpressed products of normal cellular genes. Therefore, tolerance deletes high-affinity T cells directed against the TAAs, leaving only a low-affinity repertoire. We have demonstrated previously that the T cell repertoire against the immunodominant p773-782 A2.1-Her-2/neu-restricted peptide has low affinity in A2xneu mice (Her-2/neu mice crossed with A2.1/Kb mice), compared with A2xFVB mice (A2.1/Kb crossed with FVB-wild-type mice). Immunizations with this peptide have a minor impact in preventing tumor growth in A2xneu mice. Therefore, attempts to expand these responses may be of little clinical value. We hypothesized that if not all possible cross-reactive peptides (CPs) are naturally processed and presented, the possibility exists that T cells against these CPs persist in the repertoire and can be used to induce antitumor responses with higher avidity against native epitopes present on the tumor cells. We have used the positional scanning synthetic peptide combinatorial library methodology to screen the p773-782 T cell clone. The screening data identified potential amino acids that can be substituted in the primary sequences of the p773-782 peptide. The designed CPs induce CTL responses of higher affinity in A2xneu mice compared with the native p773-783 peptide. These CTLs recognize A2+-Her-2/neu(+) tumors with high efficiency. Moreover, multiple immunizations with CPs significantly prolonged the survival of tumor-bearing A2xneu mice. These results have demonstrated that it was possible to circumvent tolerance with the identification of CPs and that these peptides could be of significant clinical value.     

Assessment of the role of determinant selection in genetic control of the immune response to insulin in H-2b mice.

The immune response to insulin is regulated by MHC class II genes. Immune response (Ir) gene-linked low responsiveness to protein Ags can be mediated by the low affinity of potential antigenic determinants for MHC molecules (determinant selection) or by the influence of MHC on the functional T cell repertoire. Strong evidence exists that determinant selection plays a key role in epitope immunodominance and Ir gene-linked unresponsiveness. However, the actual measurement of relative MHC-binding affinities of all potential peptides derived from well-characterized model Ags under Ir gene regulation has been very limited. We chose to take advantage of the simplicity of the structure of insulin to study the mechanism of Ir gene control in H-2b mice, which respond to beef insulin (BINS) but not pork insulin (PINS). Peptides from these proteins, including the immunodominant A(1-14) determinant, were observed to have similar affinities for purified IAb in binding experiments. Functional and biochemical experiments suggested that PINS and BINS are processed with similar efficiency. The T cell response to synthetic pork A(1-14) was considerably weaker than the response to the BINS peptide. We conclude that the poor immunogenicity of PINS in H-2b mice is a consequence of the T cell repertoire rather than differences in processing and presentation.     

A threshold for central T cell tolerance to an inducible serum protein

We report an inducible system of self Ag expression that examines the relationship between serum protein levels and central T cell tolerance. This transgenic approach is based on tetracycline-regulated expression of a secreted form of hen egg lysozyme, tagged with a murine hemoglobin (Hb) epitope. In the absence of the tetracycline-regulated transactivator, serum levels of the chimeric protein are extremely low (< or = 0.1 ng/ml) and the mice show partial tolerance to both Hb(64-76) and lysozyme epitopes. In the presence of the transactivator, expression increases to 1.5 ng/ml and the mice are completely tolerant. Partial tolerance was further investigated by crossing these mice to strains expressing transgenic TCRs. At the lowest Ag levels, 3.L2tg T cells (specific for Hb(64-76)/I-E(k)) escape the thymus and approximately 10% of CD4(+) splenocytes express the 3.L2 TCR. In contrast, 3A9 T cells (specific for hen egg lysozyme(46-61)/I-A(k)) are completely eliminated by negative selection. These data define a tolerogenic threshold for negative selection of Ag-specific T cells by circulating self proteins that are 100-fold more sensitive than previously demonstrated. They suggest that partial tolerance at extremely low levels of self Ag exposure is the result of a restricted repertoire of responding T cells, rather than a simple reduction in precursor frequency; tolerogenic thresholds are T cell specific.     

An MHC class II restriction bias in CD4 T cell responses toward I-A is altered to I-E in DM-deficient mice

The MHC-encoded cofactor DM catalyzes endosomal loading of peptides onto MHC class II molecules. Despite evidence from in vitro experiments that DM acts to selectively edit the repertoire of class II:peptide complexes, the consequence of DM expression in vivo, or a predictive pattern of DM activity in the specificity of CD4 T cell responses has remained unresolved. Therefore, to characterize DM function in vivo we used wild-type (WT) or DM-deficient (DM(-/-)) mice of the H-2(d) MHC haplotype and tested the hypothesis that DM promotes narrowing of the repertoire of class II:peptide complexes displayed by APC, leading to a correspondingly selective CD4 T cell response. Surprisingly, our results indicated that DM(-/-) mice do not exhibit a broadened CD4 T cell response relative to WT mice, but rather shift their immunodominance pattern to new peptides, a pattern associated with a change in class II isotype-restriction. Specifically, we found that CD4 T cell responses in WT mice were primarily restricted to the I-A class II molecule, whereas DM(-/-) mice recognize peptides in the context of I-E. The observed shift in isotype-restriction appeared to be due in part to a modification in the peripheral CD4 T cell repertoire available for peptide recognition.     

Thymocyte antigens do not induce tolerance in the CD4+ T cell compartment.

Thymocytes fail to tolerize the developing T cell repertoire to self MHC class I (MHC I) Ags because transgenic (CD2Kb) mice expressing H-2Kb solely in lymphoid cell lineages reject skin grafts mismatched only for H-2Kb. In this study, we examined why thymocytes fail to tolerize the T cell repertoire to self MHC I Ags. The ability of CD2Kb mice to reject H-2Kb skin grafts was age dependent because CD2Kb mice older than 20 wk accepted skin grafts. T cells from younger CD2Kb mice proliferated, but did not develop cytotoxic functions in vitro in response to H-2Kb. Proliferative responses were dominated by H-2Kb-specific, CD4+ T cells rather than CD8+ T cells. Representative CD4+ T cell clones from CD2Kb mice were MHC II restricted and recognized processed H-2Kb. TCR transgenic mice were generated from one CD4+ T cell clone (361) to monitor development of H-2Kb-specific immature thymocytes when all thymic cells or lymphoid cell lineages only expressed H-2Kb. Thymocyte precursors were not eliminated and mice were not tolerant to H-2Kb when Tg361 TCR transgenic mice were intercrossed with CD2Kb mice. In contrast, all thymocyte precursors were eliminated efficiently in thymic microenvironments in which all cells expressed H-2Kb. We conclude that self MHC I Ags expressed exclusively in thymocytes do not induce T cell tolerance because presentation of processed self MHC I Ags on self MHC II molecules fails to induce negative selection of CD4+ T cell precursors. This suggests that some self Ags are effectively compartmentalized and cannot induce self-tolerance in the T cell repertoire.