Peptide Variants of Viral CTL Epitopes Mediate Positive Selection and Emigration of Ag-Specific Thymocytes In Vivo¹

Naive T cells expressing a highly diverse TCR repertoire are generated in the thymus from bone marrow (BM)³ lymphoid precursors (reviewed in Ref.1). Upon entering the thymus, T cell progenitors proliferate and undergo a complex series of gene rearrangement events leading to cell surface TCR expression and subsequent differentiation (reviewed in Refs.2 and 3). It has been demonstrated that peptides bound to the MHC molecules within the thymus control both positive and negative selection (reviewed in Ref.4). During selection of the TCR repertoire, thymocytes that carry TCRs having low-affinity interactions with MHC-bound self-peptides are positively selected, and are exported into the pool of mature peripheral lymphocytes. In contrast, thymocytes bearing those TCRs that recognize self-peptides with high affinity are eliminated (3).

Single amino acid substitutions in either the MHC or the peptide dramatically alter recognition by T cells (5, 6). Analysis of crystal structures of αβ TCR/class I MHC complexes have demonstrated that peptide specificity of T cells is primarily determined by the interaction between the CDR of the TCR-Vα and -Vβ domains and the peptide side chains, which protrude of the peptide-binding groove of MHC molecules toward the two TCR-CDR3 loops. Structural studies of peptide/MHC complexes (pMHC) have provided detailed information about the conformation of peptide when bound to MHC molecules (reviewed in Refs.7 and 8). The peptide-binding groove of MHC molecules is composed of two helices on top of an eight-strand anti-parallel β-pleated sheet. The peptide-binding groove contains various binding pockets, the shape and charge of which are dependent on the highly polymorphic amino acids characteristic of a given MHC allele, which in turn selectively determines the spectrum of peptides that may bind to it (reviewed in Ref.9 and references therein).

Much of the recent work on thymic selection was influenced by experiments that examined T cell responses to peptide analogues derived from the antigenic peptide by substitution of amino acid residues involved in interactions with the TCR. Such peptide analogues, so-called altered peptide ligands (APL), can generate qualitatively different T cell responses compared with those produced by the antigenic peptide (10). In particular, some APL were shown to act as TCR antagonists and inhibit T cell responses to the antigenic peptide (11). Several studies have shown that antagonist peptides are capable of positively selecting (12, 13), negatively selecting (14), or otherwise altering (15) selection of thymocytes.

Thymic selection processes have also been addressed in structural terms using TCR-transgenic mice. For example, in N15 transgenic mice carrying a TCR specific for the vesicular stomatitis virus nucleoprotein octapeptide N52–59 (VSV8) in the context of H-2K^b, a weak agonist peptide variant inducing positive selection has been identified (16). This variant is identical with the VSV8 peptide except for substitution of leucine for valine at the P4 peptide residue (L4). The cognate viral peptide ligand, VSV8, triggers negative selection. Another TCR transgenic mouse model, P14, expressing a TCR specific for the D^b-restricted immunodominant lymphocytic choriomeningitis virus epitope gp33–41 has been developed (17). This system has been widely used to study the effect on thymocyte development of mutations in gp33–41 peptides that interact either with the binding pockets of D^b (18) or with the TCR contact residues, using fetal thymic organ culture (FTOC; reviewed in Ref.19). The crystal structure of gp33/H-2D^b shows that conserved single mutations at positions 4 or 6 of the peptide are solvent exposed and presumably function as TCR contacts (20, 21). In yet a third TCR transgenic mouse model, F5, where the TCR recognizes a nucleoprotein peptide of the influenza virus NP366–379 in the context of H-2D^b (22, 23), the peptide antagonist mediated positive selection in FTOC (23, 24), whereas the cognate peptide itself led to deletion of CD4⁺CD8⁺ (double-positive (DP)) thymocytes (25).

Variants of peptides derived from infectious agents or tumor Ags could, in principle, mediate positive selection and export of specific T cells from the thymus. As such, these APL might be candidates for manipulating the thymic repertoire in vivo, controlling the generation of naive and memory T cells within the peripheral lymphoid compartment. This “thymic vaccination approach” would aim to deliver, by parenteral administration, positively selecting APL of cognate Ags to elicit maturation of thymocytes with desired TCR specificities at the level of thymic repertoire development. Expanding repertoire generation has enormous potential in aiding the organism’s fight against infections or in affording tumor immunity. To test this concept, we have designed variants of gp33–41 and VSV8 peptides with substitutions at the amino acid residues interacting with the TCR and examined their effects on thymocyte maturation and emigration in vivo in two well-defined systems.

N15 tg^+/+ RAG2^−/− H-2^b mice were generated as described previously (26). P14 tg^+/+ RAG2^−/− H-2^b transgenic mice were obtained from Taconic Farms (Germantown, NY). Congenic strains C57BL/6 (Ly-5.2) and C57BL/6 (Ly-5.1) were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were sex matched and used at 3–4 wk of age for peptide injections and at 7–11 wk of age for other manipulations. Mice were maintained and bred under sterile barrier conditions at the animal facility of the Dana-Farber Cancer Institute (Boston, MA).

The peptide gp33–41 (KAVYNFATC) and its variants, Y4S/F6A (KAVSNAATC) and A7E (KAVYNFETC), were synthesized by standard solid phase methods with the modification of C to M (C9M) to prevent dimer formation mediated by free sulfhydryl-groups. N52-59 (VSV8, RGYVYQGL; Ref.27) and its variant, L4, (RGYLYQGL) were also made and all peptides purified by reverse phase HPLC (Hewlett Packard HPLC 1100; Hewlett Packard, Palo Alto, CA).

All anti-mouse mAbs were purchased from BD Pharmingen (San Diego, CA). For flow cytometry, single cell thymocyte, splenocyte, or lymph node suspensions were prepared in PBS containing 2% FCS and 0.05% NaN₃. Cells were stained at 1 × 10⁶ cells per 100 μl in PBS (2% FCS and 0.05% NaN₃) containing the Abs at saturating concentrations. Phenotypes and proportions of cell subsets were analyzed by three-color flow cytometry using a FACScan (BD Biosciences, San Jose, CA) and the CellQuest program (BD Biosciences). Dead cells were excluded from the analysis by forward and side scatter gating.

For sorting, BM cells or splenocytes from N15 RAG2^−/− and P14 RAG2^−/− mice were prepared as single-cell suspensions and stained at 30 × 10⁶ cells per milliliter in PBS containing the following mAbs at saturating concentrations: CyChrome-conjugated anti-CD8 and PE-conjugated anti-CD44 Abs were used to sort CD8⁺CD44⁻ splenocytes; for BM, FITC-conjugated anti-CD4, -CD8α, -CD45R/B220, -Ly-6G, and -CD11b were used to sort cells negative for the mixture of the above Abs. Cells were sorted under sterile conditions into tubes containing PBS: 2% FCS and 0.5% gentamicin using a MoFlo (DakoCytomation, Carpenteria, CA) and the Summit program (DakoCytomation).

Sorted splenocytes or BM cells (1 × 10⁶ cells per 100 μl PBS/2% FCS per mouse) were transferred i.v. into irradiated (700 rad, split dose 450 and 250 at a 3-h interval) B6 Ly-5.1 mice several hours after irradiation (¹³⁷Cs source). Peptides (25 μg/100 μl PBS per mouse) were injected i.v. 4 days after the transfer of splenocytes, or 3–4 wk after the transfer of BM cells. N15-specific peptides (VSV8 and L4) were injected once, whereas P14-related peptides (gp33–41_C9M, Y4S/F6A_C9M, and A7E_C9M) were injected once or three times every 24 h.

Tetramers consisting of complexes of biotinylated H-2D^b refolded with the gp33–41_C9M, Y4S/F6A_C9M, or A7E_C9M peptides were produced using the method previously described (28). For immunofluorescence analysis, 1 × 10⁶ cells (thymocytes, splenocytes, or lymph node cells) were incubated with FITC anti-CD8α mAb for 1 h at 4°C, followed by addition of 0.5 μg of PE-labeled tetramers gp33–41_C9M/D^b, Y4S/F6A_C9M/D^b, or A7E_C9M/D^b and CyChrome-anti-CD4 and incubation for another hour. After two washes, cells were analyzed on a FACScan as described above.

Assays for RMA-S H-2D^b stabilization, apoptosis, proliferation, BrdU, CFSE, and intracellular cytokine staining were done as detailed elsewhere (29, 30, 31).

gp33–41 is the cognate peptide Ag of the P14 TCR (Vα2 and Vβ8) and triggers negative selection of P14-bearing DP thymocytes (reviewed in Ref.19). Structural variants of gp33–41 were designed to influence the outcome of thymocyte selection by altering the affinity of the pMHC ligand interactions with the TCR. No change was made in the peptide anchor residues that occupy the binding pockets of H-2D^b, thus ensuring proper peptide presentation in the context of MHC. Indeed, the crystal structure of the gp33–41/H-2D^b complex shows that the side chains of amino acid residues at peptide positions p1, p4, p6, p7, and p8 are exposed to the solvent (20, 21). To design a variant with reduced affinity for the P14 TCR, we have introduced two types of mutations: in one mutant, both centrally disposed p4 and p6 residues have been modified (Tyr (Y) to Ser (S) at p4 and Phe (F) to Ala (A) at p6). In the other, Ala (A) was substituted with Glu (E) at p7. Both variants were synthesized in two alternative forms, one with the natural amino acid Cys (C) at the anchor residue p9, and the other with Met (M) at p9, thus avoiding any potential peptide dimerization mediated by free SH-groups. This modification was previously shown to stabilize the binding of gp33–41 peptide to H-2D^b (32). The sequence of gp33–41 and the variant peptides and the relevant gp33–41D^b structure are shown in Fig. 1 A.

Next, we verified whether these gp33–41 variant peptides were able to bind to H-2D^b molecules using RMA-S cells. To this end, RMA-S cells were incubated with the peptides listed in Fig. 1,A, and the extent of staining with anti-H-2D^b Abs on the surface of peptide-loaded RMA-S cells was measured by FACS. The binding profiles are shown in Fig. 1 B. Variant peptides Y4S/F6A_C9M and A7E_C9M bound equally well to H-2D^b molecules, and in a similar fashion compared with the cognate gp33–41_C9M epitope, suggesting that amino acid substitutions at peptide residues p4, p6, and p7 indeed do not affect peptide binding, and, by extension, peptide presentation to T cells.

To evaluate the functional potential of T cells in mice injected with gp33–41 variant peptides, splenocyte, and lymph node T cell responses to the above peptides were examined for proliferation and cytokine secretion. Both splenocytes (Fig. 1,C) and lymph node T cells (data not shown) proliferated in vitro in response to the gp33–41_C9M peptide, reaching a peak response at 10⁻¹⁰ M. The highest response to the A7E_C9M mutant peptide was achieved at the peptide concentration of 10⁻⁸ M (Fig. 1,C). It should be noted that when a similar assay was performed using the gp33–41 variant peptides, the potency of either of these peptides at the peak of response was reduced by two logs relative to gp33–41_C9M, namely, 10⁻⁸ M for the gp33–41 and 10⁻⁶ M for the A7E mutant (data not shown). In contrast, incubation of T cells with the Y4S/F6A_C9M peptide resulted in essentially no response at any peptide concentration, possibly due to low affinity interactions with the TCR. Consistent with the proliferation data, an assay for intracellular cytokine staining with anti-IFN-γ or -IL-2 Abs showed the highest levels of both cytokines when splenocytes were incubated with the gp33–41_C9M, slightly lower levels in the presence of A7E_C9M and no cytokine secretion in the presence of Y4S/F6A_C9M peptide (Fig. 1 D). Collectively, our results suggest that the Y4S/F6A_C9M variant peptide (and Y4S/F6A, data not shown) does not elicit responses of mature T cells from P14 RAG2^−/− mice.

To examine the effect of gp33–41 variant peptides on thymocyte development in P14 RAG2^−/− mice, we developed a protocol for peptide injection in vivo. Earlier studies using N15 RAG2^−/− transgenic mice showed that a single i.v. injection of VSV8 peptide leads to a severe depletion of DP thymocytes, whereas a variant of VSV8, L4 (V4L mutation at p4) mediates positive selection in FTOC (16). Here, in contrast, a single injection of gp33–41_C9M peptide into P14 RAG2^−/− transgenic mice caused only a modest reduction in the percentage of the DP thymocytes, although the total number of thymocytes was reduced by approximately two-thirds (Fig. 2 A, upper panel). Note that the down-regulation of both CD4 and CD8 on the DP thymocytes due to impending clonal deletion causes “spillover” into a single-positive (SP) CD8 gate, resulting in a higher percentage of SP CD8 cells as compared with the PBS control. However, injection of the double mutant Y4S/F6A_C9M led to an unexpected tripling of total cell numbers, without perturbation in subset distribution. In contrast, A7E_C9M had no effect on the thymocyte number and only a slight increase in the percentages of DP or SP CD8 thymocyte subsets.

To next investigate the effect of variant peptides on the expression of thymocyte surface markers characteristic of maturation and/or activation states, cells were examined by triple-color immunofluorescence using various mAbs. A representative staining profile of SP CD8 thymocytes for the expression of β₇ integrin, a marker linked to thymocyte emigration (33), is shown in Fig. 2,A (lower panel). Injection with the gp33–41_C9M peptide led to a decrease in the β₇ integrin expression level on SP CD8 thymocytes as compared with the control PBS-injected mouse (mean fluorescence intensity (MFI) 50 vs 108, respectively), despite essentially no change in the percentage (40–41%) of anti-β₇ integrin reactive cells. Both Y4S/F6A_C9M and A7E_C9M induced an increase in the percentage of this thymocyte population (61–70%) without changing β₇ integrin levels on individual thymocytes. A complete analysis of thymocyte markers is summarized in Fig. 2 B. Higher levels of CD44 and CD69 were observed on SP CD8 thymocytes injected with the gp33–41_C9M peptide. In contrast, Y4S/F6A_C9M had no effect on the above markers, but Vβ8 (P14-specific TCR), β₇ integrin, and CD8β expression were up-regulated. There was an increase in the expression of CD44, CD8β, and β₇ integrin on SP CD8 thymocytes of mice injected with the A7E_C9M variant. These results indicated that a single injection of gp33–41_C9M and Y4S/F6A_C9M affected both the cell numbers and expression of thymocyte markers, suggestive of early events in thymocyte activation. A7E injection did not alter cell numbers, but affected thymocyte marker expression. Such phenotypic changes may be reflective of molecular up- or down-regulation and/or selection of cellular subpopulations. Although not shown, alterations in cellular phenotypes were evident at the earliest interval examined postinjection (6 h) as well.

While reducing absolute cell number, a single dose of gp33–41 had little influence on the percentage of DP thymocytes in P14 RAG2^−/− mice. Therefore, we have injected gp33–41 variants every 24 h for 3 days and found that under these conditions the DP thymocyte depletion was pronounced, leading to a nearly total elimination of these thymocytes (Fig. 3,A). This observation is consistent with that made in another H-2D^b-restricted TCR transgenic system, F5, where multiple peptide injections were also required (34). In contrast, almost total elimination of N15 RAG2^−/− DP thymocytes was achieved by a single K^b-binding VSV8 cognate peptide injection (Refs.16 and 26 and data not shown). Whether this difference is a result of greater CD8αβ coreceptor binding to H-2K^b vs D^b (28), the higher copy number of peptide complexes with K^b vs D^b molecules (28), or TCR affinity differences remains to be determined. Surprisingly, injection with the Y4S/F6A_C9M mutant resulted in a significant increase in the total number of thymocytes as well as DP thymocyte subpopulation. In contrast, the A7E_C9M variant had no effect on the thymocyte counts (Fig. 3,A). The expression of the examined phenotypic markers on SP CD8 thymocytes followed a similar trend after third injection (data not shown) as compared with a single peptide injection (Fig. 2 B).

The possibility that the unusual increase in the number of DP thymocytes following exposure to the Y4S/F6A peptide might be due to cellular proliferation and attendant DNA synthesis was examined by BrdU incorporation assay as shown in Fig. 3,B. As expected, lower BrdU incorporation was detected in each of the thymocyte subpopulations (double-negative (DN), DP, and SP CD8) of mice injected with gp33–41_C9M, supporting previous observations on negative selection and apoptosis by this ligand (19). More importantly, no significant difference in the amount of BrdU incorporation was found in thymocytes of mice injected with either Y4S/F6A_C9M or A7E_C9M variants, compared with the control PBS-injected mice. As a consequence, we wondered whether Y4S/F6A peptide might increase the DP subpopulation by preventing apoptosis. To test this possibility, staining of cells from mice injected with gp33–41 variants with anti-annexin V mAb was performed (Fig. 4). Indeed, there was an increase in the percentage of annexin V⁺ cells (12%) in mice injected with gp33–41_C9M peptide, suggestive of apoptosis. In contrast, no such phenomenon was observed upon injection of either Y4S/F6A_C9M or A7E_C9M variants; in those cases, values were comparable to the control mouse. We reasoned that if the Y4S/F6A peptide competes with apoptosis-inducing peptides for binding to H-2D^b molecules on the cell surface, then Y4S/F6A may “rescue” thymocytes from undergoing cell death. Competitive binding assays, in which P14 RAG2^−/− mice were injected with the mixtures of the negatively selecting cognate peptide gp33–41_C9M and Y4S/F6A_C9M variant, were performed in vivo. The results in Fig. 5 show that, as predicted, increasing the amount of Y4S/F6A peptide in the injection mixture resulted in a higher number of total and DP thymocytes. Thus, we infer that the Y4S/F6A variant may compete with other negatively selecting peptides for binding to H-2D^b molecules expressed on thymic stroma either by binding to “empty” surface MHC class I molecules or, perhaps, by a cross-presentation mechanism (35).

To study the relative avidity of interactions between the P14 TCR and the gp33–41 variant peptides bound to H-2D^b molecules, we prepared tetramers of H-2D^b with each of the above peptides, and performed quantitative immunofluorescence analysis of thymocytes, splenocytes, and lymph node cells from P14 RAG2^−/− mice. The results of MFI staining of SP CD8 thymocytes from P14 RAG2^−/− mice with the three tetramers at different concentrations are depicted in Fig. 6. The inset shows representative staining profiles at a single comparable concentration of tetramer. Clearly, the strongest binding occurs with the tetramer containing the gp33–41_C9M peptide, as reflected by higher fluorescence intensity levels. Tetramer containing A7E_C9M mutant bound with lower affinity, whereas no detectable binding was observed with the tetramer refolded with the Y4S/F6A_C9M variant peptide. These results in conjunction with functional data (Fig. 1, B–D) suggest that the Y4S/F6A_C9M mutant must interact with the P14 TCR with extremely poor affinity, if at all.

To date, the processes of T cell development involving the interactions between the P14 TCR and the gp33–41 variant peptides have been studied exclusively in the transgenic mouse system. Although the homogeneity of TCR-expressing cells on the RAG2^−/− background is an advantage for specific analysis, it remains difficult to identify the numerically small population of recent thymic emigrants (RTE). To overcome this problem, we have used irradiation chimeras using congenic mouse strains (expressing the CD45.1 marker in B6 and CD45.2 in P14 and N15 transgenic mice). Previously, using N15 transgenic mice carrying a TCR specific for the VSV8 peptide in the context of H-2K^b, a weak agonist peptide variant L4, with the substitution of leucine for valine at the P4 peptide residue, inducing positive selection has been identified (16). We wondered whether interactions between the low affinity ligands, Y4S/F6A and L4, and their specific TCRs would result in thymic positive selection and emigration. Thus, lineage-negative BM precursors of P14 or N15 RAG2^−/− mice (donor) were injected into irradiated congenic B6 mice (recipient) and the development of donor-type cells was monitored weekly by immunofluorescence staining and multicolor FACS analysis. As shown in Fig. 7, donor-type P14-specific SP CD8 thymocytes appeared in the thymus 3–4 wk after BM injection, comprising ∼70% of the SP CD8 subset by wk 5 (Fig. 7, upper panel). In contrast, almost no peripheral donor T cells have been detected in irradiation chimeras at 3–4 wk after BM injection (middle and lower panels, Fig. 7), although such cells are identifiable between 4 and 5 wk after injection. When BM cells from N15 RAG2^−/− mice were sorted and injected into irradiated B6 recipients, and the development of donor-type T cells was examined in an analogous manner, similar kinetics of maturation and emigration of N15 RAG2^−/−-specific SP CD8 T cells were found (data not shown). Therefore, we administered the selecting peptides to the recipient at 3–4 wk after donor BM injection and assessed whether such exposure might influence the subsequent selection and emigration processes of donor thymocytes.

Between 3–4 wk post-BM reconstitution, gp33–41 and its variant peptides were injected daily for 3 days and animals examined 24 h later. Fig. 8,A, upper four panels, shows the anti-CD4 and anti-CD8α profiles of all thymocytes in the recipient (both donor and host). Fig. 8,A, lower eight panels, enumerate the donor (anti-CD45.2 reactive) cells in the DP and SP CD8 subpopulation as gated in the upper panel, with the absolute number of thymocytes given in the Fig. 8 inset (right). The number of thymocytes in irradiation chimeras injected with Y4S/F6A_C9M was highest, whereas that of gp33–41_C9M-injected mice was lowest. This difference recapitulates the effect of gp33–41_C9M variant peptides vs gp33–41_C9M on thymocytes from P14 RAG2^−/− transgenic mice. Fig. 8,A, inset, shows that the lowest DP numbers are in irradiation chimeras injected with gp33–41_C9M peptide, whereas DP numbers are increased in mice injected with the Y4S/F6A_C9M variant. The number of SP CD8 thymocytes was also highest in chimeras injected with Y4S/F6A_C9M variant peptide (Fig. 8 A, inset), suggesting that this ligand mediated positive selection of P14 RAG2^−/−-specific T cells. As the increase in thymocyte numbers exceeds the donor-engrafted population, injection of Y4S/F6A_C9M peptide may lead to the rescue of certain nontransgenic thymocytes from negative selection as well.

To appreciate whether APL might function in an analogous manner in other systems to modulate selection, we have generated N15 RAG2^−/−-B6 irradiation chimeras, injected either VSV8, L4, or PBS and subjected the animals to comparable analysis. Here, as well, the number of thymocytes in irradiation chimeras injected with L4 was highest, whereas that of VSV8-injected mice was lowest (Fig. 8,B, inset), similarly to the effect of gp33–41 variant peptides on thymocytes from P14 RAG2^−/−-B6 chimeras. The CD4/CD8 donor thymocyte profiles (Fig. 8 B, lower panel) showed the lowest percentage of DP and SP CD8 thymocytes in irradiation chimeras injected with VSV8 peptide, compared with mice injected with L4 or with PBS. In contrast, the number of both DP and SP CD8 donor thymocytes was highest in chimeras injected with L4 variant peptide, consistent with positive selection.

To determine whether Y4S/F6A and L4 lead to emigration of SP CD8 thymocytes from the thymus to the periphery, spleens and lymph nodes from the same P14 RAG2^−/−-B6 and N15 RAG2^−/−-B6 irradiation chimeras analyzed in Fig. 8 were examined using triple-color immunofluorescence with anti-CD45.2, anti-CD8α, and anti-Vα2 or -Vβ5 mAbs. The results are represented in Fig. 9,A, where the percentages of donor T cells in the host lymph nodes of P14 RAG2^−/−-B6 irradiation chimeras injected with gp33–41_C9M variant peptides are shown. Note that treatment with gp33–41_C9M leads to activation of the cognate P14 T cells, as judged by their size increase (Fig. 9,A, upper panel) and down-regulation of the TCR (Vα2; Fig. 9,A, lower panel), in line with previous observations in other TCR transgenic models (25). The greatest number of donor-type CD45.2⁺CD8⁺Vα2⁺ T cells (Fig. 9,A, inset) was in the lymph nodes of Y4S/F6A_C9M-injected chimeras, suggesting that donor-type thymocytes developing in the presence of Y4S/F6A_C9M mature and emigrate to the lymph nodes. Similar increase in the numbers of donor-type cells in lymph nodes was observed 9 wk after injection of Y4S/F6A_C9M peptide (data not shown). The functional analysis of donor-type CD8⁺ lymph node T cells in irradiation chimeras injected with the positively selecting Y4S/F6A_C9M peptide showed ∼2-fold higher proliferation levels in response to the cognate peptide gp33–41_C9M in vitro, compared with cells from PBS control-injected chimeric mice, reflecting the 2-fold difference in the number of donor-type CD8⁺ T cells in lymph nodes of chimeras injected with the Y4S/F6A_C9M peptide (data not shown). Note that Y4S/F6A_C9M peptide induces little emigration to spleen relative to the PBS control. In contrast, in N15 RAG2^−/−-B6 irradiation chimeras injected with the L4 variant, higher CD8⁺Vβ5.2⁺ donor-type T cell numbers were observed both in lymph nodes and spleens, several days (Fig. 9 B, inset) or 9 wk (data not shown) after L4 injection. The possible basis for this difference is described below, perhaps related to differential K^b vs D^b peptide presentation.

To investigate the basis for the higher T cell numbers of the donor phenotype in the peripheral lymphoid tissues of mice injected with the Y4S/F6A or L4 peptides, we measured cell divisions in response to the above peptides in vivo. Sorted CD44⁻CD8⁺ splenocytes of P14 or N15 origin were labeled with CFSE before transfer into irradiated recipient congenic hosts. Four days later, the time interval allowing cells to settle in the peripheral organs, peptides were injected, and spleens and lymph nodes taken for FACS staining 3–4 days after the peptide injection. CFSE allowed assessment of the extent of proliferation of individual transferred T cells, as previously described (31). As shown in Fig. 10 A, the percentage of CFSE⁺CD8⁺ cells was lowest in lymph nodes of irradiation chimeras injected with the gp33–41_C9M peptide, with no brightly CFSE⁺ cells remaining, suggesting that naive P14-specific T cells had undergone proliferation and activation-induced cell death (AICD) in response to the gp33–41_C9M ligand. In contrast, the percentage of CFSE⁺CD8⁺ T cells in mice injected with the Y4S/F6A _C9M peptide was similar to the control PBS-injected mice, implying that this peptide caused no T cell expansion. In chimeras injected with the A7E_C9M peptide, the percentage of CFSE⁺CD8⁺ cells was higher than in gp33–41_C9M-injected mice, but lower than in Y4S/F6A_C9M-injected mice, consistent with the A7E_C9M variant inducing some degree of T cell proliferation. It must be noted that because the adoptive transfer is into irradiated recipients, donor-type cell proliferation is significant even in the absence of peptide administration (31, 36), based on the reduction in the intensity of CFSE staining of the control (PBS injected) chimeras (MFI = 121) as compared with the initial CFSE staining intensity of donor cells before injection (MFI = 9000).

CD44⁻CD8⁺ splenocytes from N15 RAG2^−/− mice were similarly labeled with CFSE and transferred into irradiated congenic hosts, followed by injection of VSV8 or L4 peptides (Fig. 10,B). In the presence of VSV8, no brightly CFSE⁺CD8⁺ cells were detected in lymph nodes (Fig. 10 B) or spleen (data not shown), suggesting that naive N15-specific T cells had undergone proliferation and AICD in response to the VSV8 ligand. In contrast, the number of CFSE⁺CD8⁺ T cells in mice injected with the L4 peptide was similar to the control group, implying that this peptide caused neither substantial T cell expansion or AICD. These data suggested that cognate peptide ligands gp33–41 and VSV8, interacting with the TCR with relatively high affinity, compared with their respective APL, induce activation of peripheral T cells, whereas peptide variants Y4S/F6A and L4, which bind TCR with low affinity and mediate positive selection, do not cause cell divisions.

Analysis of Y4S/F6A and A7E peptides in H-2^b mice represents the first examination of the direct effects of amino acid substitutions at the P14 TCR contact residues on thymocyte selection and activation in vivo. The crystal structure of gp33/H-2D^b suggests that single mutations at both peptide positions 4 (p4) or 6 (p6) directly affect TCR contacts (20, 21, 37). Those mutants of gp33–41 mediate positive selection presumably due to weaker pMHC/TCR interactions (19). In contrast to previous studies, in the present work mutations at p4 and p6 of gp33–41 were introduced in the same variant, to generate a positively selecting ligand with reduced TCR affinity to foster emigration of the developing thymocytes to the periphery. Indeed, Y4S/F6A_C9M was found to interact with the P14 TCR on SP CD8 thymocytes with an affinity below our detection limit, as judged by tetramer staining. In contrast to substitutions of amino acids at p4 plus p6, a single mutation at p7 did not significantly affect thymocyte subset numbers. However, the A7E_C9M variant altered the thymocyte phenotype (Fig. 2,B). Our functional data (Fig. 1) suggest that A7E_C9M is a weak agonist of gp33–41. Consistent with the ability of peptides with the mutation at p7 to interact with the P14 TCR, incubation of P14 thymocytes with A7S in vitro in FTOC at high molarity resulted in negative selection (38).

Y4S/F6A_C9M induced up-regulation of the P14 TCR, the CD8β coreceptor, and the β₇ integrin levels on SP CD8 thymocytes, characteristic of positive selection, with no change in the expression of other markers for activation/emigration/positive selection, including CD25, CD44, CD62L, and CD69. Characterization of RTE has been controversial. Up-regulation of several markers on SP CD8 thymocytes undergoing positive selection and emigration, including CD5 (39), β₇ integrin, CD44 and L-selectin (CD62L) has been reported (33, 40), whereas others did not observe these changes (41). Our data further demonstrate the heterogeneity of the phenotypes of SP CD8 thymocyte subpopulations affected by positively selecting ligands, as well as difficulties in the precise characterization of the small subpopulation of RTE. In contrast, gp33–41_C9M led to thymocyte activation and subsequent increase in the expression of CD44 and CD69, in line with similar observations in other models (25, 42, 43), whereas the A7E_C9M led to up-regulation of CD44, β₇ integrin and CD8β (Fig. 2 B).

The kinetics of reconstitution of irradiated hosts by thymocyte progenitors from the BM of P14- and N15-TCR transgenic mice was similar to the findings of Tanchot and Rocha (36). In addition, we show that thymocyte emigration is dependent on the affinity/avidity of pMHC/TCR interactions. Peripheral SP CD8 T cells of the donor phenotype, when transferred into the irradiated hosts later injected with Y4S/F6A_C9M or L4 ligands, did not expand, as judged by CFSE staining. “Background” proliferation did occur in a peptide-independent manner in chimeric hosts due to availability of niches caused by irradiation (31). These results suggest that although the low affinity pMHC/TCR interactions are insufficient to trigger cell divisions, differentiation nevertheless follows.

The nature and the number of APL involved in positive selection of MHC class I-restricted T cells and their relationship to antigenic peptides has been controversial (reviewed in Ref.3 and references therein). Although most of the above studies have been performed in vitro, little is known about mechanisms of positive selection of CD8 T cells in vivo. Affinity measurements support the idea that positively selecting peptide ligand affinities are lower than those of negatively selecting ligands for TCRs, but additionally linked to their MHC binding/stability properties (44). A recent publication described an antagonist of H-2K^b-specific OT-I TCR and a variant of OVA 257-264 peptide, (E1), endogenously expressed by cortical epithelial cells of TAP-deficient mice, which mediated positive selection of CD8⁺ T cells in vivo (45). Our report supports the idea that weak pMHC class I/TCR interactions promote positive selection of SP CD8 thymocytes. Certainly the 10,000-fold weaker functional N15 T cell stimulation by L4 vs VSV8 peptide is consistent with the view (46). However, because Y4S/F6A_C9M in complex with H-2D^b has no detectable binding with the P14 TCR, we cannot exclude the possibility that this pMHC/TCR interaction is no greater than the basal level of P14 TCR binding to D^b in general. Two recent studies in class II MHC-restricted TCR transgenic mouse systems are also consistent with the notion that weak pMHC ligands may foster positive selection (47, 48).

Of importance is the observation that Y4S/F6A_C9M led to an increase in the number of DP thymocytes, a phenomenon which has not been reported to occur during positive selection. However, the binding of Y4S/F6A_C9M to D^b might possibly prevent other endogenous negatively selecting thymic peptides from binding and interacting with the TCR. Consistent with this possibility, we show that Y4S/F6A_C9M competes for binding to H-2D^b with the negatively selecting cognate peptide gp33–41 (Fig. 5).

Collectively, our data show that cognate peptides can be modified to create variants that result in selection, directly or indirectly, of desired TCR specificities at the level of thymic development. This exogenous peptide administration offers a potential of expanding repertoire generation in vivo in a manner useful to the organism. Whether these peptide-specific T cells generate stronger defense mechanisms to fight viral infection or tumors in the normal, nontransgenic mouse, remains to be investigated. In this respect, exploring means of enhancing differentiation of thymocytes bearing desired TCRs, together with the understanding of the mechanism of thymocyte emigration to the periphery, would be of great importance. Several agents have been shown to inhibit thymic export (reviewed in Ref.49), while a recent report described factors mediating emigration from the thymus (50). In the future, combined approach of exposing the subject to a positively selecting APL plus a thymic export-enhancing agent might generate a practical and efficient protective immunity.

We thank Dr. David Harrington (Department of Biostatistical Science, Dana-Farber Cancer Institute) for preparation and statistical analysis of Fig. 3 A, and Dr. Diane Mathis and Dr. Linda Clayton for review of the manuscript.