Effect of MHC Class I Diversification on Influenza Epitope-Specific CD8+ T Cell Precursor Frequency and Subsequent Effector Function

Earlier studies of influenza-specific CD8+ T cell immunodominance hierarchies indicated that expression of the H2Kk MHC class I allele greatly diminishes responses to the H2Db-restriced DbPA224 epitope (acid polymerase, residues 224–233 complexed with H2Db). The results suggested that the presence of H2Kk during thymic differentiation led to the deletion of a prominent Vβ7+ subset of DbPA224-specific TCRs. The more recent definition of DbPA224-specific TCR CDR3β repertoires in H2b mice provides a new baseline for looking again at this possible H2Kk effect on DbPA224-specific TCR selection. We found that immune responses to several H2Db- and H2Kb-restricted influenza epitopes were indeed diminished in H2bxk F1 versus homozygous mice. In the case of DbPA224, lower numbers of naive precursors were part of the explanation, though a similar decrease in those specific for the DbNP366 epitope did not affect response magnitude. Changes in precursor frequency were not associated with any major loss of TCR diversity and could not fully account for the diminished DbPA224-specific response. Further functional and phenotypic characterization of influenza-specific CD8+ T cells suggested that the expansion and differentiation of the DbPA224-specific set is impaired in the H2bxk F1 environment. Thus, the DbPA224 response in H2bxk F1 mice is modulated by factors that affect the generation of naive epitope-specific precursors and the expansion and differentiation of these T cells during infection, rather than clonal deletion of a prominent Vβ7+ subset. Such findings illustrate the difficulties of predicting and defining the effects of MHC class I diversification on epitope-specific responses.

D espite the fact that viruses encode multiple proteins, virus-specific CD8 + CTLs focus on a limited number of peptide + MHC class I (pMHCI) epitopes. The relative magnitudes of Ag-expanded pMHCI-specific CTL populations often sort into reproducible hierarchies ranging from large (immunodominant) to small (subdominant) (1). Although many factors may determine the positioning of a given pMHCI complex within this hierarchy, the first requirement is that the peptide should access the nascent MHCI molecule during the course of infection, then bind in a way that is recognized by the available CD8 + TCR repertoire (1,2). Underlying any repertoire effect is the fact that these same MHCI alleles that present viral peptides at the time of infection have earlier played a key role in TCR selection during thymocyte ontogeny. A number of studies have examined the impact of differences in MHCI haplotype and diversity on the generation of epitope-specific CD8 + T cell responses in mice (3)(4)(5)(6)(7)(8) and humans (9)(10)(11), providing some evidence that the presence of particular MHCI alleles can modify responses to peptides presented by other MHCI molecules. Possible mechanisms to explain this effect include changes in the thymic selection of TCRs specific for an individual epitope (6,10) and altered epitope presentation during infection (11)(12)(13). The latter could reflect modulation of MHCI cell surface expression depending on the combination of MHCI alleles expressed (13,14), or that competition for overlapping peptides during infection has the potential to modify epitope processing or presentation (11,15). Clearly, such MHCI-related effects on response magnitude are of interest as we seek to understand and manipulate CTLmediated immunity in MHC-polymorphic human populations.
An accumulating body of evidence suggests that epitope-specific TCR diversity plays a part in determining the quality of virusspecific CTL responses and the outcome of infection. More diverse repertoires have been shown to include TCRs with a range of structures and avidities, allowing a spectrum of response profiles that provide better protection against virus infection (16) and minimize the possibility of mutational escape from CD8 + T cellmediated immune control (17). This has led to the suggestion that vaccination, particularly to protect against viruses (such as HIV and influenza) that readily generate mutants, should optimally prime memory T cell populations with a high level of TCR diversity. Given the role for MHC alleles in shaping the available TCR repertoire (5,10,16,18), it is important to understand how differences in MHCI haplotype are likely to impact on the frequency, range, and properties of the epitope-specific TCRs selected by immunization, infection, or both.
Early studies of influenza epitope-specific CD8 + T cell responses in mice indicated that expression of the MHCI allele H2K k greatly diminishes the magnitude of H2D b -restricted responses (4,8). Subsequent experiments narrowed this effect down to a particular epitope-specific population, demonstrating that virus-infected B6xC3HF 1 (B6C3F 1 , H2 bxk ) mice mount significantly smaller responses to the D b PA 224 epitope (acid polymerase, residues 224-233 complexed with H2D b ) compared with C57BL/ 6 (B6, H2 b ) mice (3). In contrast, responses to the other immunodominant D b NP 366 epitope (nucleoprotein, residues 366-374 complexed with H2D b ) were equivalent between the two mouse strains after primary infection. The diminished CD8 + D b PA 224 + responses in the F 1 mice looked to be associated with the deletion of a prominent Vb7 + subset of D b PA 224 -specific TCRs, supporting the idea that thymic development of naive CD8 + D b PA 224 + precursors in the presence of H2K k had led to a specific "hole" in this TCR repertoire. This result was recorded at a very early stage of our understanding of the Ag-selected CD8 + D b PA 224 + response, long before we had any capacity to measure naive T cell numbers and accurately compare TCR repertoires. Having much better tools available, we repeated this analysis in B6C3F 1 mice, with the intent of achieving a much more rigorous definition of the impact of MHCI diversification and possible cross-tolerance (during thymic differentiation) on the selection and expansion of D b PA 224specific CD8 + T cells during influenza virus infection.
Although this study confirmed a significant reduction in the magnitude of the D b PA 224 -specific response in H2 bxk F 1 mice, it further identified reduced responses to the subdominant epitopes K b PB1 703 (polymerase B1, residues 703-711 complexed with H2K b ), D b PB1F2 62 (polymerase B1 +1 reading frame, residues 62-70 complexed with H2D b ), and K b NS2 114 (nonstructural protein 2, residues 114-121 complexed with H2K b ) in F 1 compared with B6 mice at all time points examined. Thus, several responses appear to be diminished as a result of MHCI diversification and cross-tolerance. Furthermore, it now seems that our previous conclusion that the CD8 + D b PA 224 + Vb7 + set is substantially missing from the B6C3F 1 response may have resulted from the selective deletion of these high-avidity T cells during the process of generating peptide-stimulated CD8 + D b PA 224 + cell lines for analysis of Vb usage. Although the numbers of naive CD8 + D b PA 224 + precursors were reduced in B6C3F 1 compared with B6 mice, this was not associated with major changes in TCRb repertoire diversity and could not fully account for the diminished response during infection. Overall, it seems that influenza-specific CD8 + T cell responses in B6C3F 1 mice may be impeded by factors that affect both the generation of naive CTL precursors (CTLps) and the expansion and differentiation of these precursors subsequent to infection.

Mice and virus infections
The C57BL/6J (B6, H2 b ), C3H/HeJ (C3H, H2 k ), and B6xC3HF 1 (B6C3F 1 , H2 bxk ) mice were bred and housed in specific pathogen-free conditions at the Department of Microbiology and Immunology at the University of Melbourne (Parkville, VIC, Australia). For analysis of primary responses to influenza virus, mice were anesthetized by methoxyflurane inhalation and infected intranasally (i.n.) with 10 4 PFU of the influenza virus A/HK-x31 (HKx31, H3N2) in 30 ml PBS. Recall responses were examined in mice that were first primed i.p. with 1.5 3 10 7 PFU of the influenza virus A/PR8/34 (PR8, H1N1) at least 6 wk before i.n. infection with the HKx31 virus. Memory responses were analyzed 6 wk after i.p. priming with either the wild-type or recombinant (19,20) PR8 viruses. Virus stocks were amplified in the allantoic cavity of day 10 embryonated chicken eggs, and virus titers were determined by plaque assay as PFU on monolayers of Madin-Darby canine kidney cells (21). All experiments were performed with the approval of The University of Melbourne Animal Ethics Committee.

Tissue sampling and cell preparation
Lymphocytes were recovered from the infected airways of mice by bronchoalveolar lavage (BAL) and incubated on plastic for 1 h at 37˚C to remove adherent cells. Single-cell suspensions were prepared from spleens by pressing through 70 mm cell strainers. In some instances, spleen cells were treated with Tris-buffered ammonium chloride to lyse RBCs.

Kinetics of virus infection
For determination of the kinetics of pulmonary virus infection in B6 and B6C3F 1 mice, either naive or PR8-primed mice were infected i.n. with the HKx31 virus. At various intervals postinfection, lungs were removed and homogenized in RPMI 1640 (Invitrogen) containing 100 U/ml streptomycin, 100 U/ml penicillin (both Invitrogen), and 24 mg/ml gentamicin (Sterisafe). Virus present in lung homogenates was quantified by plaque assay as described earlier.

Tetramer and Ab staining
Lymphocytes were plated at 0.5-2 3 10 6 cells and incubated with the D b NP 366 , D b PA 224 , K b PB1 703 , D b PB1-F2 62 , and K b NS2 114 tetramers (ImmunoID) conjugated to either PE or allophycocyanin for 1 h at room temperature. After washing, cells were stained for CD8a and, in some instances, for CD62L. For analysis of Vb usage, splenocytes from naive mice or tetramer-stained splenocytes from mice sampled on day 10 after HKx31 infection were stained with anti-CD8a-PerCPCy5.5 (53-6.7; BD Pharmingen) and a panel of 14 FITC-conjugated mAbs specific for various Vb families (BD Pharmingen). To identify Ag-experienced CD8 + T cells, we stained splenocytes with anti-CD8a-allophycocyanin (53-6.7; Bio-Legend) and anti-CD11a-FITC (2D7; BioLegend). For analysis of cell surface H2D b and H2K k expression, splenocytes from naive mice were treated to lyse erythrocytes and then incubated in Fc block (spent 2.4G2 supernatant with 0.5% normal mouse serum and 0.5% normal rat serum). Cells were then stained with anti-H2D b -FITC (KH95; BD Pharmingen) or anti-H2K k -FITC (36-7-5; BioLegend). Flow cytometric analysis was performed on a BD FACSCalibur (BD Biosciences), and the data were analyzed by either CellQuest Pro (BD Immunocytometry Systems) or FlowJo (Tree Star) software.
Sorting, single-cell PCR, and sequencing Single CD8 + D b NP 366 + Vb8.3 + or CD8 + D b PA 224 + Vb7 + cells were sorted on a BD FACS Aria (BD Immunocytometry Systems) into the wells of a 96well PCR plate. cDNA was synthesized, and Vb8.3 + or Vb7 + transcripts were amplified and sequenced as described previously (22,23). The TCR gene nomenclature used here is according to Arden et al. (24).

Tetramer elution assay
The TCR avidities of D b NP 366 -and D b PA 224 -specific populations from B6 and B6C3F 1 mice were compared using a tetramer elution assay (30). Splenocytes (0.5-2 3 10 6 cells) were stained with the D b NP 366 -PE or D b PA 224 -PE tetramers for 1 h at room temperature to label epitope-specific populations. Cells were washed and then incubated in the presence of 5 mg/ml anti-H2D b (28-14-8; BD Pharmingen) at 37˚C. At designated time points, cells were removed onto ice, washed, then stained for CD8a expression and analyzed by flow cytometry for residual tetramer staining.

Statistical analysis
All statistical comparisons were performed using an unpaired Student t test.

Results
Reduced H2D b -, H2K b -, and H2K k -restricted responses in H2 bxk F 1 mice These experiments extend our earlier IFN-g intracellular cytokine staining assay comparison (3) of CTL responses after HKx31 influenza A virus infection of naive and "memory" B6C3F 1 (H2 bxk ) and B6 (H2 b ) mice by using tetramer staining reagents representing the D b NP 366 , D b PA 224 , K b PB1 703 , D b PB1F2 62 , and K b NS2 114 epitopes. Analysis of CTLs from the spleen and infected airways (BAL) 10 d after primary infection with the HKx31 virus ( Fig. 1A-D) confirmed (3) that the numbers of CD8 + D b PA 224 + cells were ∼8fold lower in the F 1 spleen and ∼20-fold lower in the F 1 BAL (Fig.  1B, 1D). The "minor" K b PB1 703 , D b PB1F2 62 , and K b NS2 114 responses were also significantly diminished in the F 1 mice, but only by 2-to 3-fold (spleen; Fig. 1B). Thus, of the five H2 b -restricted epitopes examined, only the D b NP 366 response was not significantly reduced in the heterozygotes, although it was repeatedly observed to be slightly lower ( Fig. 1A-D). As such, the immunodominance hierarchy ( Fig. 1A Secondary responses were analyzed in mice immunized i.p. at least 6 wk previously with the H1N1 PR8 virus, then challenged i.n. with the serologically distinct H3N2 HKx31 virus ( Fig. 1G-J) that shares the immunogenic PR8 internal proteins. Although the numbers of D b NP 366 -specific CD8 + memory T cells in B6 and B6C3F 1 mice were not significantly different after PR8 priming (3) (Fig. 1E, 1F), further expansion after secondary challenge established that there was indeed a reduction in the F 1 spleen (2-fold; Fig. 1H) and BAL (4-fold; Fig. 1J). In fact, the recall responses to all epitopes were generally lower in the heterozygotes ( Fig. 1G-J).
A lack of published data on H2 k -restricted epitopes recognized in C3H mice limited our ability to compare the influenza-specific CD8 + T cell responses in C3H and F 1 mice (3,31). However, to demonstrate that F 1 mice generated responses to both H2 b -and H2 k -restricted viral peptides, we compared primary ( Fig. 2A-D), memory (Fig. 2E, 2F), and recall ( Fig. 2G-J) responses to the D b NP 366 , D b PA 224 , and K k NP 50 epitopes for the parental and F 1 mice using an IFN-g intracellular cytokine staining (30) assay because we did not have tetramer reagents available for the K k NP 50 epitope. The percentages of D b PA 224 -and K k NP 50 -specific CD8 + cells were significantly reduced after primary and secondary challenge of F 1 mice; however, as found previously with the tetramers (Fig. 1G-J), this was apparent only for the secondary response to D b NP 366 (Fig. 2G-J).

Pathogenicity of influenza virus infection in parental and F 1 mice
Because B6C3F 1 mice generate a broader CD8 + T cell response targeted toward a greater number of viral peptides (Figs. 1, 2), we reasoned that F 1 mice might control influenza more rapidly than the B6 parents, which may, in turn, lead to reduced epitopespecific CTL responses (32). The kinetics of body weight loss and viral clearance were thus compared after primary or secondary challenge of F 1 and inbred parental mice (Fig. 3). F 1 mice lost significantly less weight compared with both parent strains after primary infection (Fig. 3A). Profiles of weight loss were more similar between the three strains after secondary infection, though B6 mice lost more weight compared with F 1 mice between days 2 and 4 ( Fig. 3B). There was a high mortality rate among C3H mice after primary infection (7/8 mice), which was reduced when the mice were primed with the PR8 virus i.p. before secondary challenge i.n. (2/5 mice) (Fig. 3A, 3B). This susceptibility to influenza infection may be influenced by a mutation in the TLR4 gene of C3H mice (33), which could make these mice more susceptible to secondary bacterial infections (34), particularly in the absence of pre-existing immunity to influenza virus (Fig. 3A, 3B).
Neither peak titers nor the kinetics of virus clearance differed after primary infection (Fig. 3C), indicating that the reduced magnitude of epitope-specific CD8 + T cell responses observed in F 1 mice (Figs. 1, 2) is not accounted for by any divergence in viral load. However, after secondary challenge, although maximal lung virus titers were comparable for B6, C3H, and F 1 mice (Fig. 3D, day 2), the C3H and F 1 mice cleared the virus slightly faster (Fig.  3D, day 8), indicating that there are strain-specific differences in the kinetics viral clearance.
Similar estimates of the total influenza-specific CD8 + T cell response in H2 b , H2 k , and H2 bxk F 1 mice It was possible that the total size of the influenza-specific CD8 + T cell response differed in B6, C3H, and B6C3F 1 mice. With only limited information available on the H2 k -restricted influenza epitopes recognized in C3H and F 1 mice, it was not feasible to enumerate the total influenza-specific CD8 + T responses in these mice via tetramer staining or peptide stimulation and IFN-g detection. We therefore used short-term (5 h) stimulation with anti-CD3 to estimate the size of the total IFN-g-producing CD8 + T cell populations in F 1 and parental mice, reasoning that during the acute stage of the CD8 + T cell response to influenza virus infection, the majority of IFN-g-producing CD8 + T cells in the spleen and BAL would be influenza-specific. Indeed, after primary challenge with the HKx31 virus, the proportion of IFN-g + CD8 + T cells detected in the spleens of B6 mice (18.6 6 5.7%; Fig. 4A) correlated well with the proportion of total influenza-specific CD8 + T cells calculated by adding up all the H2 b -restricted responses measured by tetramer staining (∼15%; Fig. 1A). However, the proportion of IFN-g + CD8 + T cells detected in the BAL of B6 mice (73.4 6 5.5%; Fig. 4C) exceeded the proportion of influenza-specific CD8 + T cells identified by tetramer staining (∼44.3%; Fig. 1C), which could reflect nonspecific recruitment of memory CD8 + T cells into the lung airways during infection (35). After primary infection, similar proportions and numbers of IFNg + CD8 + T cells were detected in the spleen and BAL of B6, C3H, and B6C3F 1 mice (Fig. 4A-D), suggesting that the sizes of the total primary influenza-specific CD8 + T cell responses are roughly equivalent for the three strains.
Another strategy to compare the size of the influenza-specific CD8 + T cell responses in B6, C3H, and B6C3F 1 mice is exami-nation of expression levels of CD11a and CD8a to identify and enumerate Ag-experienced CD8 + T cells (CD11a hi CD8a low ) (7). In naive mice, ,11% of total CD8 + T cells in the spleen displayed a CD11a hi CD8a low Ag-experienced phenotype (Fig. 4E). Primary influenza infection led to a substantial increase in the frequency of CD11a hi CD8a low CD8 + T cells (approaching 40% of all CD8 + T cells in the spleen), and there were no significant differences in the proportion and number of CD11a hi CD8a low CD8 + T cells between mouse strains. Thus, according to at least two measures, B6, C3H, and B6C3F 1 mice appear to generate similar-sized CD8 + T cell responses to influenza virus infection.
Prevalence of epitope-specific CTLps in naive H2 bxk F 1 mice It has been suggested that the number of epitope-specific CTLps can substantially influence response magnitude during infection (5,25,26,36). The recent refinement of enrichment/flow cytometrybased protocols (26) allows the reproducible recovery of epitopespecific CTLps from naive secondary lymphoid tissue. We applied this technique to determine whether differences in the D b PA 224 + CTLp frequency for B6 and B6C3F 1 mice contributed to the reduced response to this epitope in heterozygous mice. Numbers of D b NP 366 + CTLps were also determined, because this was the only epitope-specific response that approached a similar magnitude in Even so, it is important to note that, despite a 2-fold reduction, the number of D b PA 224 -specific CTLps in the heterozygotes is still significantly greater than the number of D b NP 366 + CTLps (p # 0.005; Fig. 5). Furthermore, the overall decrease in magnitude for the splenic D b PA 224 -specific response in B6C3F 1 versus B6 mice was at least 8-fold after primary infection (Fig. 1). Thus, although  the diminished availability of D b PA 224 + CTLps in B6C3F 1 mice may contribute to a reduced response to this epitope during infection, it appears that additional mechanisms are influencing the extent of Ag-driven clonal expansion.

Characterization of TCRVb usage in parental and F 1 mice
Differences in MHCI haplotype and genetic background can impact on the selection of immune repertoires and the availability of epitope-specific TCRs (5,16,37). To determine whether the reduced generation of CD8 + D b PA 224 + and CD8 + D b NP 366 + cells in B6C3F 1 mice was associated with modified thymic TCR repertoire selection and possible deletion of particular TCR clonotypes, we compared the prevalence of mAb-defined TCRVb families for all CD8 + T cells from naive B6, C3H, and B6C3F 1 mice (Fig. 6A). Overall, this global measure of Vb frequency indicated that CD8 + T cells from B6C3F 1 and C3H mice were more likely (than B6 mice) to express Vb2, Vb6, Vb8.1/8.2, Vb8.3, and Vb14, but less likely to stain for Vb3, Vb5.1/5.2, and Vb12. Most striking was the minimal use of Vb5.1/5.2 by C3H and B6C3F 1 mice (1.1 6 0.1% and 0.4 6 0.1%, respectively) relative to the B6 strain (13.6 6 0.7%). Thus, the TCRb repertoire in C3H and B6C3F 1 mice differs in some respects from that selected in the B6 strain, suggesting that either C3H non-MHC background genes or the presence of the H2 k haplotype is in some way influencing the emergence of particular Vb families. However, looking at the results (Fig. 6A) for the Vb8.3 and Vb7 families that are used prominently in the B6 response to D b NP 366 and D b PA 224 , respectively, there were no dramatic differences in prevalence between the F 1 and parent.
Our previous analysis described the absence of Vb7 + cells from in vitro expanded D b PA 224 -specific B6C3F 1 CTL lines, with the conclusion that the lack of response was due to a "hole" in the available B6C3F 1 repertoire (3). A concern with using this approach to measure Vb prevalence is that it relies on stimulation with high doses of peptide that may lead to loss of CTLs with high TCR/pMHC avidity (30). The analysis has thus been revisited using contemporary tetramer and Ab staining to measure Vb usage by D b NP 366 -and D b PA 224 -specific CD8 + cells taken directly ex vivo from influenza virus-infected B6 and B6C3F 1 mice (Fig.  6B, 6C). In B6 mice, Vb8.3 + TCRs are very prominent (42.3 6 37.3%; Fig. 6B) in the D b NP 366 -specific response (22,38), whereas Vb7 predominates (55.1 6 8.7%; Fig. 6C) for the D b PA 224 -specific set (3,23). These Vb profiles were essentially unaltered in the B6C3F 1 mice, with the CD8 + D b NP 366 + Vb8.3 + and CD8 + D b PA 224 + Vb7 + populations representing the major Vb families used (65 6 17.1% and 62 6 03.1%, respectively). Although expression of most subdominant Vb families was similar for CD8 + D b NP 366 + and CD8 + D b PA 224 + cells from B6 and F 1 mice, the prevalence of Vb4 TCRs was increased among F 1 CD8 + D b PA 224 + cells (p , 0.05; Fig. 6C). Thus, at this global level, MHC diversification in B6C3F 1 mice has not substantially perturbed the  Estimates of the overall size of the influenza virus-specific CD8 + T cell responses in B6, B6C3F 1 , and C3H mice. Lymphocytes from the spleen and BAL of HKx31infected mice were stimulated for 5 h with plate-bound anti-CD3ε Ab, and the frequency (A, C) and number (B, D) of IFN-g + CD8 + cells were determined. Data are representative of two independent experiments and show mean 6 SD for five B6 and B6C3F 1 mice, and the mean for two of five surviving C3H mice. Profiles of CD11a upregulation and CD8a downregulation were used to identify Ag-experienced CD8 + T cells in the spleens of naive mice and mice infected with the HKx31 virus 10 d previously (E). The frequency (F) and number (G) of CD11a hi CD8a low CD8 + T cells in the spleens of HKx31-infected mice were compared. epitope-specific TCRb repertoires. The previous interpretation that CD8 + D b PA 224 + Vb7 + cells are deleted in B6C3F 1 mice during development thus appears to be an artifact resulting from the use of high-dose peptide stimulation to generate in vitro CTL lines. Similar CDR3b repertoires for CD8 + D b NP 366 + and CD8 + D b PA 224 + cells from B6 and B6C3F 1 mice It was possible that the D b PA 224 -specific Vb7 + TCRb repertoire in the B6C3F 1 mice might be altered without any obvious change in the overall profile of Vb usage. This was addressed by comparing the findings for single-cell CDR3b sequence analysis of CD8 + D b PA 224 + Vb7 + cells from influenza-infected B6C3F 1 mice with the extensive published data for the B6 parent (19,23,39). The CDR3b repertoire of CD8 + D b NP 366 + Vb8.3 + cells (22), which also establish slightly lower precursor frequencies in B6C3F 1 and B6 mice (Fig. 5), was analyzed concurrently. The results are summarized in Table I Tables I, II). Thus, the reduced frequency of D b PA 224 + CD8 + cells in B6C3F 1 mice does not appear to reflect any impaired selection of Vb7 + D b PA 224 -specific TCRs. Similarly, obvious differences in TCRb usage do not explain the reduced number of D b NP 366 + CTLps in F 1 mice.
Functional and phenotypic differences in influenza-specific CD8 + T cells from F 1 and B6 mice A 2-fold reduction in D b PA 224 + CTLp frequency appeared insufficient to explain the substantial 8-fold reduction in the B6C3F 1 D b PA 224 + CD8 + response after infection. Furthermore, we had failed to identify differences in the F 1 D b PA 224 -specific TCRb repertoire that could account for this discrepancy. We thus asked whether some difference in the quality of the F 1 D b PA 224 + CD8 + cells might correlate with this poor expansion during infection.
Immune CD8 + cells were recovered from the BAL and spleens of F 1 and B6 mice on day 8 postinfection with the HKx31 virus and stimulated with the NP 366 or PA 224 peptides to see whether there was any difference in the cytokine production profiles. Overall, the numbers of CD8 + D b NP 366 + and CD8 + D b PA 224 + cells producing IFN-g (Fig. 7A, 7B) were equivalent to the counts detected by tetramer staining (data not shown), and the level of cytokine production (measured as mean fluorescence intensity) differed only for the D b NP 366 -specific set from spleen (Fig. 7C,  7D). However, looking within the splenic CD8 + IFN-g + population, a greater proportion of both the D b NP 366 -and D b PA 224specific CD8 + cells from B6C3F 1 mice produced TNF-a (Fig. 7E, 7F) and at higher mean fluorescence intensity levels than in the B6 parent (Fig. 7G, 7H). These B6/F 1 differences in TNF-a profiles were much less evident for the highly activated BAL populations (Fig. 7F, 7H).
The CD62L lymph node homing receptor is typically downregulated on influenza-virus-specific CD8 + CTL effectors and  PA 224 -specific CD8 + T cells in unprimed H2 bxk mice. Epitope-specific CD8 + T cell precursors were identified from pooled lymphoid tissues (25,26). Data show total numbers of D b NP 366 -and D b PA 224 -specific CD8 + T cell precursors found in individual B6 and B6C3F 1 mice, with the lines indicating the mean for each group.
reflects the extent of differentiation (40,41). To determine whether levels of CD62L downmodulation were similar between B6 and F 1 mice, we compared CD62L expression on CD8 + cells stained with D b NP 366 and D b PA 224 tetramers from B6 and B6C3F 1 mice on day 8 postinfection (Fig. 7I, 7J). The expression of CD62L on CD8 + D b NP 366 + cells from the spleen and BAL of B6 and B6C3F 1 mice was similar. In contrast, significantly fewer T cells within the smaller CD8 + D b PA 224 + set from the spleens of B6C3F 1 mice had downmodulated cell surface CD62L by day 8 postinfection (Fig. 7I); however, again, this effect was no longer apparent in the more highly stimulated BAL environment (Fig.  7J). Both high TNF-a prevalence (Fig. 7E-H) and the retention of the CD62L hi phenotype (Fig. 7I) are, in fact, characteristic of T cells that have undergone less cycles of division during the course of the host response (40,42, and A.E. Denton, P.C. Doherty, and S.J. Turner, unpublished observations). Thus, the diminished D b PA 224 -specific response may reflect an inability to fully differentiate a mature CTL response.

Functional avidity versus TCR/pMHC binding avidity
To determine whether differences in T cell responsiveness to peptide stimulation were involved in the impaired expansion of the F 1 CD8 + D b PA 224 + set, immune CD8 + D b NP 366 and CD8 + D b PA 224 + cells were stimulated in vitro with graded concentrations of peptides, using IFN-g production as a readout (Fig. 8A, 8B). Although identical profiles were found throughout for the B6 and F 1 CD8 + D b NP 366 + sets (Fig. 8A), the CD8 + D b PA 244 + T cells from B6C3F 1 mice were slightly less sensitive at concentrations ,10 29 M (Fig.  8B). This variation in "functional avidity" for the CD8 + D b PA 244 + T cells (Fig. 8B) was independent of TCR avidity, as the rates of tetramer elution were identical (Fig. 8C, 8D) and the B6 and F 1 TCR repertoire profiles look to be broadly comparable (Fig. 6, Table I). Thus, the CD8 + D b PA 224 + cells from B6C3F 1 mice are slightly less responsive to low concentrations of peptide, an effect that may, again, reflect that they have undergone fewer cycles of clonal expansion (43).

Reduced levels of MHCI allele expression may contribute to reduced responses in F 1 mice
A way in which MHCI diversification in F 1 mice may affect T cell priming is through altered levels of MHCI (and thus peptide) presentation. In fact, cell surface expression of H2D b and H2K k was reduced in B6C3F 1 mice with respect to parental mice (Fig.  9). However, only the D b PA 224 -specific CTL response, and not the D b NP 366 -specific CTL response, is substantially affected in F 1 mice relative to B6 mice (Figs. 1A-D, 2A-D). This may reflect that D b PA 224 presentation is more limited after influenza A virus infection (19,44). If levels of Ag presentation are a key de-terminant of response magnitude in F 1 mice, then increasing levels of D b PA 224 epitope presentation could potentially increase the magnitude of the CD8 + T cell response to this epitope in F 1 mice. To test this hypothesis, we infected mice with a recombinant PR8-NA PA virus (19) that contains the PA 224 peptide inserted into the neuraminidase stalk of the virus. This, in turn, increases the abundance of the PA 224 peptide in the influenza virion and facilitates presentation of D b PA 224 on nonprofessional APCs (19). In 3 (83.6 6 14.9%) 0 (0.0 6 0.0%) a More than 10% of sequences. b A total of 7.9 6 2.5 clonotypes/mouse found in B6 (H2 b ) mice (22). c A total of 20.6 6 3.8 clonotypes/mouse found in B6 (H2 b ) mice (23). d Found in at least four of five mice. line with our previous findings, there was a modest increase in the percentage and number of D b PA 224 -specific CD8 + T cells in B6 mice (relative to mice that received a control virus; Supplemental  Fig. 1A, 1B). However, the same increase in the D b PA 224 -specific response was not observed in F 1 mice (Supplemental Fig. 1A, 1B).
Thus, it appears that the ability to boost the D b PA 224 response is limited by an intrinsic defect within the B6C3F 1 background.

Discussion
This analysis extends our earlier dissection of how H2 k expression influences H2D b -restricted, influenza-specific CD8 + CTL responses (3,4). We demonstrated that H2 bxk F 1 mice generated reduced primary and memory CD8 + T cell responses to at least four of five H2D b -and H2K b -restricted epitopes prominently recognized in the H2 b parent. Moreover, the normal influenza CD8 + CTL immunodominance hierarchy established in the H2 b (B6) mice was altered in the H2 kxb (B6C3F 1 ) heterozygotes because of a defect in the response to the D b PA 224 epitope. These data demonstrate a broad effect of H2 k expression on the magnitude of both immunodominant and subdominant responses. Importantly, reduced primary influenza-specific CTL responses in F 1 mice could not be explained by reduced pulmonary virus titers. Numbers of D b NP 366 + and D b PA 224 + CLTps were reduced in B6C3F 1 mice compared with B6 mice, an observation that could not be explained by deletion of a prominent TCRb subset or noticeable differences in TCR CDR3b repertoire. This was unexpected given previous observations suggesting that CD8 + D b PA 224 + Vb7 + cells were missing from the B6C3F 1 response (3). Immune CD8 + D b NP 366 + and CD8 + D b PA 224 + cells from B6C3F 1 also seemed to be less differentiated, and the CD8 + D b PA 224 + set was slightly less sensitive to stimulation with low concentrations of peptide. Together, these data suggest that the F 1 environment provides a lower quality of stimulation to these epitope-specific CD8 + T cells during infection, which contributes to functional and proliferative differences in these responses compared with the situation for the B6 parent.
The additional MHCI and genetic diversity of F 1 mice can impact epitope-specific responses in two key ways: via differences in TCR repertoire selection that impact the availability of epitopespecific CTLps and through altered levels of epitope presentation. Although differences in the naive TCRVb repertoire were found for B6 and B6C3F 1 mice, and seemed a likely explanation for the decrease in (particularly) D b PA 224 + CTLp frequency, we were unable to identify defined "holes" in the F 1 repertoire that could account for this. Previous observations that the Vb7 + subset of D b PA 224 -specific TCRs was missing in F 1 mice (3) were most likely an artifact of using peptide-stimulated CTL lines, with consequent deletion or lack of expansion of these "high-avidity" TCR clonotypes in culture (30). It remains possible that, if we were to add the analysis of TCRa expression, we might find some evidence for deletion of F 1 CD8 + D b NP 366 and CD8 + D b PA 224 clonotypes. However, apart from the fact that the numbers of naive F 1 CTLps are down by up to 2-fold, there is no a priori reason for thinking that this might be the case.
Interestingly, the mAb-defined profile of Vb usage of CD8 + T cells from naive B6C3F 1 mice more closely resembled the Vb repertoire selected in the C3H parent, rather than representing a mixture of B6 and C3H parental strains. Both MHC and non-MHC gene products are known to skew Vb profiles in different mouse strains (45,46), and in this case, it seems that the C3H background (or H2 k haplotype) is to some extent dominant. It is possible that, as a consequence of these differences in TCRb repertoire selection, other epitope-specific responses found in B6 mice may be relatively diminished in the B6C3F 1 heterozygotes. Indeed, further study will be needed to determine whether such a mechanism is responsible for the decreased responses to K b PB1 703 , D b PB1-F2 62 , and K b NS2 114 in B6C3F 1 mice.
The effect of certain MHCI alleles on Ag-specific TCR usage has been shown for mice and humans. The HLA-B8 response to an epitope from the EBNA3 protein of EBV (FLRGRAYGL, FL9) selects CD8 + cells expressing a public TCR (LC13) (47) that cross-reacts with the alloantigen HLA-B*4402 (48) and is deleted in HLA-B8 + HLA-B*4402 + individuals (10). Even so, EBVinfected HLA-B8 + B*4402 + individuals still mount a significant FL9-specific response, but use a different, more diverse spectrum of TCRs compared with HLA-B8 + B*4402 2 individuals. These TCRs can avoid HLA-B*4402 reactivity by adopting a TCR footprint that is shifted to an area of polymorphism between HLA-B8 and HLA-B*4402 (49). Thus, the TCR repertoire has suffi-FIGURE 8. Comparison of avidity profiles for influenza-specific CD8 + T cells from H2 b and H2 bxk F 1 mice. Profiles of "functional avidity" were determined for splenocytes from mice infected i.n. with the HKx31 virus 10 d previously. The T cells were stimulated with decreasing amounts (100 to 0.0001 nM) of NP 366 (A) or PA 224 (B) peptide for 5 h, and the data are expressed as the percentage of IFN-g-producing CD8 + cells with respect to the maximum IFN-g + CD8 + response after stimulation with the highest concentration of peptide. For measurement of TCR avidity by tetramer dissociation, splenocytes were stained with the D b NP 366 (C) or D b PA 224 (D) tetramers and incubated in the presence of anti-H2D b Ab at 37˚C. The results show tetramer elution over time, represented as the percentage of cells staining with tetramer relative to maximum staining measured in the absence of the anti-H2D b blocking Ab. Mean 6 SD for five mice. **p , 0.01 comparing F 1 and B6. . Lower cell surface expression of H2D b and H2K k on cells from B6C3F 1 mice. Splenocytes from B6, C3H, and B6C3F 1 mice were treated to lyse erythrocytes and block Fc receptors, then stained separately with anti-H2D b -FITC or anti-H2K k -FITC to measure cell surface MHCI expression. Histograms show staining intensities obtained for H2D b (A) and H2K k (B) on total splenocyte populations from the three mouse strains. Histograms are scaled as a percentage of the peak response (% max). cient diversity and versatility to ensure that deletion of prominent Ag-specific TCR clonotypes does not severely compromise the magnitude of the response. In a study of vaccina virus infection of H2 dxb F 1 mice and inbred parents (5), altered epitope-specific TCRVb profiles in the heterozygotes were not necessarily associated with diminished immune responses. Furthermore, the conservation of parental epitope-specific TCRVb profiles in the F 1 mice did not always result in equivalent response magnitudes. Thus, although expression of different MHCI alleles can lead to shifts in intrathymic selection of the TCR repertoire to maintain self-tolerance, this does not necessarily compromise the availability of epitope-specific TCRs or the response magnitude. However, contradicting this, a study of influenza-specific CTL responses in humans demonstrated that certain HLA-A-and HLA-B-restricted responses varied in magnitude depending on the HLA haplotype of the individual (9). Although this highlights the significance of MHC-related effects on epitope-specific responses, it remains to be seen whether these effects are mediated at the level of thymic selection or Ag presentation.
The addition of MHCI alleles and background genes in F 1 mice requires T cells with a greater range of MHCI restriction and peptide specificity to be accommodated without a change in overall CD8 + T cell numbers. The prevalence of epitope-specific CD8 + T cells may thus be reduced in F 1 mice because of increased positive selection of other TCRs. Indeed, it has been suggested that lower levels of cell surface MHCI expression in heterozygous versus homozygous mice may play a role in reducing the positive selection of some epitope-specific TCRs (50), perhaps through a quantitative reduction in TCR signals (51). Differences in MHCI glycoprotein levels for cells from parental and F 1 mice are, in part, a consequence of gene dosage (13). Analysis of cell surface H2D b and H2K k concentrations on splenocytes from B6, C3H, and B6C3F 1 mice showed reduced expression of both of these alleles in the F 1 versus either parent. In this case, perhaps the decreased levels of H2D b result in reduced numbers of positively selected D b NP 366 + CD8 + and D b PA 224 + CD8 + cells and the observed decrease in naive CTLps.
It has been suggested that naive precursor frequency can provide a good prediction of the immunodominance status of epitopespecific responses after Ag-driven clonal expansion (5,36,52). However, this relationship did not obviously determine the magnitude of D b NP 366 -and D b PA 224 -specific responses after influenza infection of B6 and B6C3F 1 mice. In both strains, there are more naive D b PA 224 + CTLps compared with D b NP 366 + CTLps; yet, beyond the first few days, the D b PA 224 -specific response never dominates after normal influenza virus infection (41). Given the reduced H2D b expression in F 1 mice, it is tempting to speculate that the levels of cell surface D b NP 366 and D b PA 224 presentation may be selectively diminished in the F 1 , with consequent reduction in the quantity and quality of stimulation delivered to CD8 + T cells. However, any decrease in D b NP 366 presentation does not appear to substantially affect relative response magnitudes in B6C3F 1 versus B6 mice. This may reflect differences in the presentation profiles of D b NP 366 and D b PA 224 . It has been demonstrated that during infection, a broader array of cell types are capable of presenting the D b NP 366 epitope compared with the D b PA 224 epitope (20,44). Thus, even though expression of H2D b is reduced in F 1 compared with B6 mice, levels of D b NP 366 presentation will likely be greater than D b PA 224 , and therefore sufficient, to prime a response in F 1 mice similar in magnitude to that observed in B6 mice after primary infection. In contrast, the generally lower abundance of the PA protein, coupled with reduced H2D b levels in B6C3F 1 mice, may result in D b PA 224 presentation that is decreased for both duration and concentration, with a consequent decline in clonal expansion and magnitude for responding CTL populations. In this way, a 2-fold lower precursor frequency could be extended to an 8-fold difference in numbers of splenic D b PA 224 + CD8 + T cells after primary infection. Interestingly, others have also found that relatively small differences in epitope-specific CTLp frequency between F 1 mice and inbred parents can result in much larger differences in effector CTL numbers after infection or peptide vaccination (5). Finally, we cannot rule out the possibility that differences in D b PA 224specific response magnitude could reflect differences in Ag processing and presentation between B6C3F 1 and B6 mice. It has been reported that presentation of the PA 224 epitope is exquisitely sensitive to expression of specific immunoproteasome subunits (53). Thus, it remains a possibility that there is limited upregulation of immunoproteosome expression within the B6C3F 1 mice that impacts PA 224 processing and presentation. This intriguing possibility remains a possible avenue of future study.
Consistent with decreased proliferation and maturation of the response [ (40,42) and A.E. Denton, P.C. Doherty, and S.J. Turner, unpublished observations], a greater proportion of CD8 + D b PA 224 + cells from B6C3F 1 mice produced TNF-a (and in higher amounts) and more were CD62L hi . Furthermore, despite equivalent structural TCR avidity, the B6C3F 1 CD8 + D b PA 224 + population was slightly less sensitive to stimulation with peptide. Taken together, these data suggest that responses to D b PA 224 are poorly elicited in the B6C3F 1 environment and fail to reach their full potential for expansion and function. Indeed, several of the very early studies looking into the effect of MHC haplotype on virus-specific CTL responses suggested that the presence of an H2K k environment during infection and not lack of precursors was responsible for limiting the potential of H2D b -restricted, influenza-specific responses (4,8).
As shown in this article, MHCI diversification can substantially affect the selection and expansion of many epitope-specific CD8 + T populations. The explanation of these effects is clearly not as simple as previously thought (3) and involves multiple factors that affect the generation of naive precursors, as well as the proliferation and activation of these precursors during infection. Although this certainly presents challenges for the design of T celldirected vaccines for use in MHCI-diverse human populations, these studies have provided key insights into the complexity of parameters that shape CTL response magnitude where MHCI diversification is apparent.