Recognition of the Major Histocompatibility Complex Restriction Element Modulates CD8⁺ T Cell Specificity and Compensates for Loss of  T Cell Receptor Contacts with the Specific Peptide

B6, transporter associated with antigen processing (TAP)1^−/− (29), β₂m^−/− (30, 31), and TAP1/β₂m^−/− (32) mice of the H-2^b haplotype were bred at the Microbiology and Tumor Biology Center, Karolinska Institute. Animal care was in accordance with institutional guidelines. TAP1^−/−, β₂m^−/−, and TAP1/β₂m^−/− mice used were back-crossed to B6 background 6, 10, and 7 times, respectively. For thymectomy experiments, β₂m^−/− mice were 735 rad–irradiated and thymectomized or sham-thymectomized, and reconstituted with β₂m^−/− fetal liver hematopoietic cells.

RMA is a subline of the Rauscher virus–induced B6 lymphoma RBL-5, and RMA-S is a TAP2-deficient variant of RMA (33). T2 is a hybrid between the two human cell lines 0.174 and CEM, and has an antigen-processing deficiency due to a deletion on the MHC class II region including the TAP1 and TAP2 genes. T2D^b is an H-2D^b transfectant of T2. All cell lines were maintained at 37°C and 5% CO₂ in RPMI 1640 tissue culture medium supplemented with 5% FCS, 50 μg/ml streptomycin, 100 μg/ml penicillin, and 2 mM l-glutamine. Con A–activated blasts were generated by culturing erythrocyte-depleted splenocytes in 5 μg/ml of Con A for 2 d in tissue culture medium as described above with 10% FCS.

The following synthetic H-2D^b–presented peptides were synthesized using solid phase F-moc chemistry: LCMV GP33 KAVYNFATM (34, 35); LCMV GP 33-4A (GP33-4A) KAVANFATM; LCMV GP 33-34A (GP33-34A) KAAANFATM; LCMV GP 33-348A (GP33-348A) KAAANFAAM; influenza PR8 NP 366-374 (NP366) ASNENMETM (36); and Yersinia enterocolitica Yop51 249-257 (Yop249) IQVGNTRTI (37).

GP33-specific CD8⁺ CTLs were elicited in B6 and β₂m^−/− mice by peptide immunization (38). 100 μg peptide was dissolved in distilled water and mixed with IFA in a 1:1 ratio by sonication, then injected subcutaneously in the base of the tail. 12 d after immunization, 25 × 10⁶ immune spleen cells were cocultured with 25 × 10⁶ 2,000 rad–irradiated B6 or β₂m^−/− splenocytes in the presence of 0.05 μM peptide in 12 ml complete medium (RPMI 1640 supplemented with 10% FCS, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 5 × 10⁻⁵ M 2-ME, 2 mM l-glutamine, and 50 μg/ml streptomycin, 100 μg/ml penicillin) at 37°C and 5% CO₂. 6–7 d later, these cells were used as effector cells in a ⁵¹Cr-release assay. CTL clones were generated by limiting dilution cloning. Long-term CTL clones and lines were maintained in complete medium (see above) based on MEM-α and further supplemented with Hepes buffer and 20 IU/ml IL-2. CTL clones and lines were restimulated in 12-d intervals with 2,000 rad–irradiated splenocytes in the presence of 0.05 μM peptide.

B22-249.1 is a mouse mAb which binds to a conformation-dependent epitope on the α1 domain of the H-2D^b molecule, and Y3 is a mAb which binds to conformed H-2K^b molecules. For flow cytometry with these mAbs, cells were incubated with mAb for 30 min on ice, washed, and then stained with goat anti–mouse Oregon Green conjugate (Molecular Probes) on ice for 30 min. For measurement of CD8 and TCR expression, CTLs were stained for 30 min with the FITC-conjugated anti-CD8α mAb 53-6.7 (PharMingen) and the PE-conjugated anti–TCR-α/β mAb H57-597 (PharMingen), respectively. After washing, analysis was performed using a FACScan^® (Becton Dickinson) with CellQuest software.

CTL activity was measured in a standard ⁵¹Cr-release assay. In brief, peptide-coated target cells were prepared by incubating cells with indicated concentrations of peptide for 1 h at 37°C. Coated cells were labeled with 10 μl 10 mCi/ml ⁵¹Cr for 1 h at 37°C. Titrated amounts of effector cells were incubated with 3 × 10^{3 51}Cr-labeled target cells for 4 h at 37°C, 5% CO₂. After incubation, released radioactivity was measured and specific lysis was calculated according to the formula: % specific release = [(experimental release − spontaneous release)/(maximum release − spontaneous release)] × 100. Cold temperature–treated target cells were generated through culture of RMA-S at 26°C for 12 h (39). For cold target competition experiments, effector T cells and unlabeled (cold) competitor cells were mixed and preincubated at 37°C for 30 min before addition of labeled (hot) target cells. The assay was then run as a standard ⁵¹Cr-release assay for 4 h. For blocking of CTL recognition with the B22-249.1 mAb, titrated amounts of mAb were included in the CTL assay medium.

We first compared the levels of properly conformed H-2D^b molecules on the cell surface of β₂m^−/−, TAP1/β₂m^−/−, and B6 (TAP1/β₂m^+/+) control cells stained with an mAb directed against H-2D^b (B22-249.1) or H-2K^b (Y3) conformation-sensitive epitopes. A substantial level of conformed cell surface H-2D^b molecules was found on β₂m^−/− cells, whereas conformed H-2K^b was detected at lower levels, in line with previously published results (27, 28; Fig. 1). This is compatible with H-2D^b being more independent of β₂m during folding and transport of MHC class I free heavy chains (27, 28). On Con A blasts deficient for both β₂m and TAP1, the staining with B22-249.1 was virtually at background levels, which implies that the pool of conformed H-2D^b molecules on β₂m^−/− cells is dependent on a functional TAP complex.

B6 and β₂m^−/− mice were immunized with antigenic peptides restricted to either H-2D^b or H-2K^b. All tested peptides primed responses in B6 mice (data not shown), whereas only one, the H-2D^b– restricted LCMV GP33, primed a response in β₂m^−/− mice. This is in line with the low but significant levels of folded H-2D^b molecules on the surface of β₂m^−/− cells (27; Fig. 1). CTLs from both B6 and β₂m^−/− mice primed with the GP33 peptide killed RMA-S cells pulsed with the GP33 peptide used for priming but not a control influenza NP366 peptide (Table I). CTL responses were mediated by CD8⁺ cells as determined by mAb- and complement-mediated depletion of effector populations in vitro (data not shown). FACS^® analysis of both polyclonal bulk populations and clones confirmed the TCR⁺CD8⁺ phenotype of the responding CTLs in β₂m^−/− as well as B6 mice (Fig. 2). Cell surface expression levels of both TCR and CD8 were similar in CTLs from β₂m^−/− and B6 mice. However, a marginal increase in CD8 expression levels was detected in some β₂m^−/− CTLs compared with B6 CTLs. This increase may be due to the low levels of H-2D^b expressed during selection and priming in the absence of β₂m. However, lysis performed by β₂m^−/− CTLs did not show increased dependency on CD8 compared with B6 CTLs, as determined by mAb blocking of CD8 (data not shown).

No priming of CTLs was observed in thymectomized β₂m^−/− mice that had been irradiated and reconstituted with fetal liver (Table I). Control mice (receiving irradiation and fetal liver reconstitution without thymectomy) generated a normal antipeptide CTL response, showing that the responding CD8⁺ T cell population in β₂m^−/− mice was dependent on the thymus for development. In addition, a normal antipeptide response was also found in mice that had been thymectomized but not irradiated, showing that the thymus was not necessary during the priming of the response (Table I). The latter control indicated that the generated CTLs were not positively selected in the thymus in response to the peptide injected for immunization. We conclude that β₂m^−/− mice have a peripheral pool of CD8⁺ T cells which are thymus dependent, H-2D^b restricted, and peptide specific despite expressing very low levels of H-2D^b molecules, corresponding to a few percent of those expressed by B6 mice.

Alloreactive CTLs from mice with low MHC class I expression have an increased avidity to self-MHC (23). By priming peptide-specific, self-MHC– restricted CTLs in β₂m^−/− mice it is possible to study the role of self-MHC avidity for T cell specificity in MHC- restricted T cells. To compare the avidity for self-MHC class I in B6 and β₂m^−/− CTLs specific for the GP33 peptide, these CTLs were tested for their ability to kill RMA target cells (MHC^high) in the absence of loaded peptides. B6 CTLs failed to kill RMA cells, whereas β₂m^−/− CTLs efficiently killed these target cells (Fig. 3, A and B, respectively). Cold target competition experiments made with bulk CTLs showed that the β₂m^−/− CTL killing of labeled RMA cells was completely inhibited by RMA-S pulsed with GP33 but not by RMA-S without peptide (Fig. 3 C). This showed that a majority of the clones in the β₂m^−/− CTL population were specific for both self-MHC expressed at high levels and the peptide antigen presented by MHC at low levels.

Three GP33-specific β₂m^−/− CTL clones were generated, all of which also killed RMA cells. However, the clone C10 was more efficient in killing RMA-S loaded with GP33 compared with RMA (Fig. 3 D), whereas clone 27/30 killed RMA slightly better than RMA-S plus GP33 (Fig. 3 E), and clone 3C5 killed both of these target cells to a similar extent (Fig. 3 F). This pattern is compatible with a clonal variation in TCR avidity for peptide versus MHC within the β₂m^−/− CTL population. In line with the results obtained with bulk cultures, the GP33-specific B6 CTL clone 2C10 did not recognize RMA in the absence of GP33 peptide (Fig. 3 G).

To test the reactivity of the CTLs against MHC class I in the absence of peptides, we used RMA-S cells incubated at cold temperature (26°C), which induce high levels of functionally “empty” MHC class I molecules (39), and T2D^b cells expressing D^b molecules mostly devoid of peptides. Clone 27/30 also killed RMA-S cells after cold temperature incubation (Fig. 4 A). Furthermore, T2D^b was recognized irrespective of loaded peptide, whereas T2 control targets were not killed (Fig. 4 B). Thus, the β₂m^−/− CTL clone 27/30 displays an avidity for the restriction element of the GP33 peptide, even in the absence of specific peptide. It should be noted that these CTLs were primed to respond only against GP33, indicating that the ability to recognize self-MHC had been selected for during thymic development and priming. Although these results do not exclude that β₂m^−/− CTLs also have an increased avidity for self-peptides, we use the term “peptide independent” to describe the specificity for the restriction element displayed by the GP33-specific β₂m^−/− CTLs.

We next compared the avidity of β₂m^−/− CTLs for self-MHC with and without added GP33, by blocking of the H-2D^b ligands. We used an mAb specific for properly conformed H-2D^b molecules (B22-249.1), the binding of which is not affected by GP33 (data not shown). RMA target cells were killed at similar levels regardless of the presence or absence of specific peptide (Fig. 5). However, in the absence of GP33 peptide the CTL killing activity was efficiently blocked by 5.0 μg/ml of B22-249.1, whereas virtually no blocking was observed when the RMA target cells were prepulsed with the GP33 peptide at 37°C (Fig. 5 A). FACS^® analysis of B22-249.1 binding to RMA cells did not allow accurate quantitation of the fraction of free H-2D^b ligands at half-maximal lysis, since half-maximal blocking of CTL lysis was achieved at a concentration at which staining was close to maximal (Fig. 5 B). However, these experiments clearly demonstrate that triggering of β₂m^−/− CTLs in the absence of GP33 peptide requires a considerably higher number of H-2D^b ligands.

We next made a similar series of experiments using T2D^b cells as targets. These cells express the B22-249.1 epitope at levels <10% those of RMA (Table II) but are still killed by the β₂m^−/− GP33-specific CTL clone 27/30 in the absence of specific peptide (Fig. 4 B). Lysis of T2D^b cells was blocked already at a 1.0 μg/ml concentration of B22-249.1 in the CTL assay (Fig. 5 C), indicating that a large fraction of cell surface H-2D^b molecules is required for triggering of this CTL clone. Pulsing of T2D^b cells with GP33 peptide prohibited blocking at all mAb concentrations tested (Fig. 5 C). This effect was peptide specific, since neither NP366 nor Yop249 had any effect on avidity as determined by blocking with mAb. By titration of B22-249.1 in FACS^® analysis of T2D^b (Fig. 5 D), the fraction of free H-2D^b ligands on T2D^b at half-maximal peptide-independent lysis was estimated at ∼80% (comparing Fig. 5, C and D). Thus, ∼80% of the H-2D^b molecules on T2D^b were necessary for peptide-independent recognition by the high-avidity CTL clone 27/30, which corresponds to about four times the level of folded H-2D^b on β₂m^−/− cells (Table II). Taken together, these data indicate that β₂m^−/− CTLs use peptide-independent recognition of a high number of H-2D^b ligands to compensate for the absence of specific peptide.

Finally, we analyzed the consequences of self-MHC avidity in GP33-specific B6 and β₂m^−/− CTLs in terms of peptide specificity, when MHC was expressed at lower levels. Neither B6 nor β₂m^−/− CTLs kill RMA-S (MHC^low) in the absence of specific peptide, and RMA-S cells require a pulse of ∼100 pM GP33 peptide to become sensitive targets to both B6 and β₂m^−/− CTLs (Fig. 6). This shows that B6 and β₂m^−/− CTLs require roughly equal amounts of H-2D^b/GP33 complexes to get a triggering signal. However, when using GP33 peptide variants alanine-substituted at one or several TCR contact residues (reference 40, and data not shown), the β₂m^−/− CTLs recognized RMA-S cells pulsed with up to 1,000-fold less peptide than B6 CTLs (Fig. 6). Substitution at one position (GP33-4A) already drastically reduced the efficiency of B6 CTLs against peptide-pulsed RMA-S targets, whereas β₂m^−/− CTLs could still recognize peptides with alanine substituted at two positions. Thus, although β₂m^−/− CTLs were still peptide specific, they were less dependent on the TCR contact residues in the peptide when the peptide was loaded on RMA-S target cells (which express about three times the levels of H-2D^b found on β₂m^−/− cells [Table II]). These data indicate that the avidities for peptide and MHC both contribute functionally to the triggering of CTLs, and that they can be considered separately. Further, increased avidity for the restriction element compensates for a reduced avidity for the peptide.

In this paper we have investigated the influence of TCR interaction with self-MHC in recognition of MHC-bound peptides. To address this issue, we used MHC class I–restricted CD8⁺ T cells selected and primed in an environment with high (B6) or low (β₂m^−/−) MHC expression (25). We find that self-MHC can deliver a triggering signal independently of specific peptide, provided there is an increase in self-MHC density. The low surface density of H-2 expressed on RMA-S cells was not sufficient to trigger either β₂m^−/− or B6 CTLs in the absence of the specific antigen, but increasing the ligand density by cold temperature incubation could sensitize RMA-S to the β₂m^−/− CTL clone 27/30. Furthermore, since T2D^b cells were also killed irrespective of loaded peptide, recognition of syngeneic class I was most probably peptide independent. These results indicate that the interaction between TCR and self-MHC as observed in crystals not only provides the structural framework of specific T cell recognition, it also contributes a part of the avidity required for CTL triggering.

Recognition of MHC^high targets by β₂m^−/− CTLs was blocked by mAb against properly conformed D^b molecules, whereas the presence of specific peptide prohibited blocking. These results indicated that the lack of GP33 was compensated by using an increased number of low-avidity interactions with self-MHC. The results further suggest that a minimum of four times the level of MHC class I expression present during selection and priming in vivo were necessary for activation by self-MHC to occur in vitro. RMA-S expresses only about three times the level of MHC found on β₂m^−/− cells (Table II), which may explain why β₂m^−/− CTLs were peptide specific when tested against RMA-S targets. In the presence of specific GP33 peptide, no mAb-mediated blocking of MHC^high target cell killing was observed. This supports the notion that MHC-restricted CTLs have a high avidity for the peptide antigen and a low avidity for self-MHC.

Interestingly, we found that T cells with an increased avidity for MHC class I molecules had a more relaxed peptide specificity, i.e., they were less dependent on the exact peptide sequence. While both B6 and β₂m^−/− GP33-specific CTLs had similar sensitivity for GP33, β₂m^−/− CTLs had superior sensitivity for peptides lacking one or several TCR contact residues of the GP33 peptide. Thus, increased recognition of the restriction element compensates for lack of TCR peptide affinity in recognition of low- affinity peptide ligands.

Our data argue that recognition of the two entities of the MHC complex, peptide and MHC heavy chain, can be considered separately (41). Increased recognition of MHC could substitute for lack of peptide recognition. The indication that TCR avidity for its ligand can be subdivided in this way opens the possibility of extending the differential avidity model of T cell selection and recognition (42).

The present data can be interpreted within a model for T cell recognition based on the total avidity contributed by TCR affinity for MHC, affinity for peptide, and the number of MHC–peptide complexes. Increased avidity for MHC can compensate for lack of avidity for peptide, suggesting that these binding forces are functionally interchangeable and can be considered separately (Fig. 7 A). This would suggest a variation in the avidities for MHC and presented antigens among mature CTLs (See also Fig. 3, D–G). Interestingly, mice deficient in terminal deoxynucleotidyl transferase (TdT), which lack N-region additions in the TCR, display increased promiscuity in T cell recognition of peptide ligands (43). In this situation, MHC was suggested to provide compensatory avidity in recognition of peptide antigen by TdT^−/− CTLs. Further, the data in this paper support an inverse correlation between the number of MHC–peptide ligands necessary for triggering and TCR avidity for the ligand (for data, see Fig. 5; for model, see Fig. 7 B). Lower avidity for an antigen would then be compensated by an increase in antigen density. This feature becomes most important for CTLs with a relatively high MHC avidity, since self-MHC can be regarded as a high-density and low-affinity ligand.

By combining Fig. 7, A and B, a hypothetical three- dimensional diagram can be generated in which the surface depicts how the TCR affinities for peptide and MHC, and the number of ligands necessary for triggering, are interchangeable (Fig. 7 C). The diagram proposes that when the TCR avidities for both peptide and MHC are low, the number of ligands necessary to achieve triggering will be high. We would like to suggest further that the same model would apply for thymic selection, where the threshold coordinates for positive selection would be considerably lower on all axes while coordinates for negative selection would be closer to the threshold for triggering of effector functions.

Selection of H-2D^b–restricted T cells specific for the GP33 epitope in β₂m^−/− mice is thymus dependent, and results in a peripheral repertoire biased towards recognition of the restriction element expressed with and without self-peptides. However, peptide-independent triggering occurs only when MHC is expressed at levels considerably higher than those encountered by the T cells in vivo. This indicates repertoire calibration during thymic selection to fit the self-MHC ligand density in the periphery. Our results fit very well with the notion that T cells being selected in the thymus view self-MHC as a low-affinity ligand. Self-ligands capable of triggering induce deletion, and in the periphery self-MHC never reaches the ligand density to trigger CTLs in the absence of triggering antigens. It should be noted that in addition to the increased recognition of MHC, the β₂m^−/− CTLs may also display an increased avidity for self-peptides. There may be a clonal variation in recognition of MHC and peptide within the β₂m^−/− CTL population, where the clone 27/30 represents the most peptide-independent phenotype.

Considering the low expression of MHC present during selection in the β₂m^−/− mice, it is possible that the CD8⁺ T cells use upregulation of (co)receptor expression to achieve positive selection. Indeed, we have observed a marginal increase in CD8 expression in some of the β₂m^−/− CTLs, which potentially could also contribute to the elevated recognition of self-MHC. However, in functional experiments based on recognition of GP33 and alanine-substituted variants, we have observed that β₂m^−/− CTLs are less susceptible to CD8 blocking with anti-CD8α mAbs compared with B6 CTLs. This result would suggest, rather, a decreased dependency on CD8 in triggering of β₂m^−/− CTLs (data not shown).

Accumulating evidence suggests that the CD8⁺ T cells present in the β₂m^−/− mice have retained important characteristics of the β₂m^+/+ wild-type concerning development and recognition of MHC. First, β₂m^−/− mice can generate in vivo CTL responses against MHC class I–restricted peptide (44; and this study) and viral (45) antigens, and reject skin grafts over a minor histocompatibility barrier (46). Also, β₂m^−/− CTLs can recognize peptide antigens on β₂m^+/+ targets (this study). Second, development of β₂m^−/− CD8⁺ T cells is dependent on the thymus (this study), and results in a CD8⁺ T cell repertoire biased towards recognition of syngeneic class I (23; and this study). Third, H-2D^b heavy chains expressed in the absence of β₂m are recognized by conformation-dependent mAbs (27, 28, 47; and this study), and expression is TAP dependent (32; and this study). Note also that all detectable β₂m^−/− CTL reactivity against self-MHC could be blocked by an mAb specific for properly conformed H-2D^b molecules, which reduces the likelihood for a role of aberrantly conformed free D^b heavy chains in this system. Fourth, wild-type CTLs can recognize endogenously processed antigens (48) and alloantigens (27, 48) on β₂m^−/− cells. Taken together, these data suggest that T cell selection follows the normal rules in β₂m^−/− mice, although the selection window is dramatically shifted towards low ligand density. The confrontation of such T cells with cells expressing normal MHC levels allows detection of the T cell avidity for self-MHC. However, the functional avidity must be established already during selection and must exist during priming. Thus, we propose that T cells in normal mice have a similar avidity for self-MHC which would be detectable by exposing them to cells with supraoptimal MHC levels.

We have demonstrated that avidity for self-MHC can trigger MHC-restricted T cells independently of specific peptide if ligand density of self-MHC is sufficiently increased. This interaction compensates for lack of TCR affinity for peptide, and it depends on a TCR–MHC interaction of relatively low affinity that requires high numbers of MHC ligands. The data and the avidity threshold model discussed contribute to our understanding of T cell specificity in the periphery and during thymic selection.

We thank Hidde Ploegh, Adnane Achour, Benedict Chambers, and Hans-Gustaf Ljunggren for discussions and reading of the manuscript.

This work was supported by the Swedish Cancer Society, Karolinska Institute, and Robert Lundbergs minnesstiftelse. R. Glas is supported by a fellowship from The Swedish Foundation for International Cooperation in Research and Higher Education (STINT), Stockholm, Sweden, and by Alex och Eva Wallströms stiftelse.

β₂m

β₂-microglobulin

B6

C57Bl/6

LCMV

lymphocytic choriomeningitis virus

TAP

transporter associated with antigen processing