Preferential HLA Usage in the Influenza Virus-Specific CTL Response1

To study whether individual HLA class I alleles are used preferentially or equally in human virus-specific CTL responses, the contribution of individual HLA-A and -B alleles to the human influenza virus-specific CTL response was investigated. To this end, PBMC were obtained from three groups of HLA-A and -B identical blood donors and stimulated with influenza virus. In the virus-specific CD8+ T cell population, the proportion of IFN-γ- and TNF-α-producing cells, restricted by individual HLA-A and -B alleles, was determined using virus-infected C1R cells expressing a single HLA-A or -B allele for restimulation of these cells. In HLA-B*2705- and HLA-B*3501-positive individuals, these alleles were preferentially used in the influenza A virus-specific CTL response, while the contribution of HLA-B*0801 and HLA-A*0101 was minor in these donors. The magnitude of the HLA-B*0801-restricted response was even lower in the presence of HLA-B*2705. C1R cells expressing HLA-B*2705, HLA-A*0101, or HLA-A*0201 were preferentially lysed by virus-specific CD8+ T cells. In contrast, the CTL response to influenza B virus was mainly directed toward HLA-B*0801-restricted epitopes. Thus, the preferential use of HLA alleles depended on the virus studied.

C ytotoxic T lymphocytes have been shown to play an important role in the control and clearance of virus infections, including those caused by influenza virus (1,2). Although a large number of peptides are generated during processing of viral proteins in infected cells, only some of these peptides are ultimately presented by MHC class I molecules and recognized by specific CTL. This limited recognition of a small number of dominant CTL epitopes has been termed immunodominance (3)(4)(5). Usually immunodominance is defined using a number of synthetic peptides representing known CTL epitopes, without taking into account the full repertoire of potential epitopes. Little information is available on the complete repertoire of CTL epitopes of viral pathogens and the corresponding HLA alleles presenting these epitopes to specific CTL. Using transient HLA expression systems, HLA-A11, HLA-B8, and HLA-B44 were identified as the preferred restriction elements in the CTL response to EBV (6 -9). These studies also demonstrated that the HLA-A1-restricted EBVspecific CTL response is of minor importance.
In most studies addressing the contribution of individual HLA molecules in CTL responses, synthetic peptides representing known CTL epitopes have been used (10 -15). It has been shown recently that the magnitude and specificity of influenza A virusspecific CTL response, using nine peptide analogues of known influenza A virus CTL epitopes, are related to the MHC class I haplotype of individuals (15). In this study, it was demonstrated that the magnitude of the influenza A virus-specific CTL response was highest in HLA-A*0201-positive individuals. This association was thought to be related to recognition of an HLA-A*0201-restricted immunodominant epitope (M1 58 -66 ) from the matrix protein. Thus, it might be expected that if CTL responses were analyzed by individual HLA-A and -B alleles, the majority of CTL would recognize influenza A virus-infected targets in a HLA-A*0201-restricted fashion. Although the use of synthetic peptides provides some information on HLA usage in CTL response, this approach has limitations and does not account for the full repertoire of epitopes presented by infected cells.
In the present study, the contribution of individual HLA-A and -B alleles to the CTL response against a model viral pathogen was investigated. To this end, influenza virus-specific memory CTL response was studied, using C1R cells expressing a single HLA-A or -B allele and influenza virus-stimulated PBMC obtained from three groups of healthy blood donors with defined HLA class I haplotypes. Between groups, the individuals shared two or three of the four HLA-A and -B alleles. Intracellular IFN-␥ and TNF-␣ staining was used for the quantification of virus-specific CTL restricted by a single HLA-A or -B allele, and the lytic capacity of these cells was determined in classical 51 Cr release assays.
We found that individual HLA alleles were not used to the same extent and that in each group of HLA-typed individuals a hierarchy existed between HLA alleles. Especially HLA-B*2705 and HLA-B*3501 were found to be preferred alleles in the influenza A virusspecific CTL response. The preferential HLA usage was found to depend on the type of influenza virus (A or B) studied.

Materials and Methods
Cells PBMC obtained from 13 healthy blood donors (Table I) were isolated between 1999 and 2002 using Lymphoprep (Nycomed, Oslo, Norway) gradient centrifugation and cryopreserved at Ϫ135°C. Three groups of donors, aged between 35 and 50 years of age, were selected according to serological homology within the A locus and B locus of HLA class I molecules (Table I). Genetic subtyping was performed in the laboratory for Histocompatibility and Immunogenetics at the Sanquin Bloodbank Rotterdam using a commercial typing system (GenoVision, Vienna, Austria). All donors had serum Abs against one or more influenza A virus strains, measured by hemagglutination inhibition assay, indicative of one or more exposures in the past. Also, three of the four donors tested for influenza B virus-specific CTL immunity (group II) had influenza B virus-specific Abs.
Hmy2-C1R (C1R) is a human BLCL with low expression of endogenous HLA-Cw4 (17,18), but no expression of HLA-A or -B alleles. Control C1R cells and those transfected with individual HLA-A or -B genes were cultured in R10F. The transfected C1R cells were kindly provided by several investigators: HLA-A1-transfected C1R cell line by P.  (35). Although the genotype of the majority of these HLA-transfected C1R cell lines is known, the cell lines will be depicted by their serotype throughout the manuscript (for example, C1R-B27). Because C1R-A1 cells gradually lose the expression of HLA-A1 on their surface, an enrichment procedure was performed every 2-3 wk using anti-HLA-A1,36 mAbs (BIH0331; One Lambda, Canoga Park, CA) together with DNA-linked anti-mouse Ig beads (Dynal Biotech GmbH, Hamburg, Germany), according to manufacturer's instructions, to ensure a high percentage of HLA-A1 ϩ C1R cells (minimal of 80% HLA-A1 ϩ cells).

Serology
Plasma samples were stored at Ϫ20°C and tested for the presence of influenza A virus (H3N2)-and influenza B virus-specific Abs in a hemagglutination inhibition assay according to standard methods (19,20) using turkey erythrocytes and four hemagglutinating units of virus. The sera were tested for Abs against 11 vaccine strains of subtype H3N2 used since 1968 and 11 influenza B virus strains (from 1979 till 2001). Ferret sera raised against the test Ags were used as positive controls.

In vitro stimulation of PBMC with influenza virus
Stimulation of PBMC with influenza virus was performed, as described previously (15). Cells were resuspended at 1 ϫ 10 6 cells/ml in R10F and infected for 1 h at 37°C with Resvir-9 or B/Harbin/7/94, at a multiplicity of infection (MOI) of 3. Next, the cells were washed once and resuspended in RPMI 1640 medium supplemented with 10% human AB serum, 2 mM glutamine, 100 g/ml streptomycin, 100 IU/ml penicillin, and 20 M 2-ME (R10H), and added to uninfected PBMC at a ratio of 1:1 in a 25-cm 2 culture flask. After 2 days, rIL-2 (final concentration 50 U/ml; Chiron B.V., Amsterdam, The Netherlands) was added and the cells were incubated for another 6 days at 37°C and used as effector cells in a 51 Cr release assay or intracellular cytokine staining (ICS) assay (see below).

Isolation of CD8 ϩ T cells
CD8 ϩ T cells were isolated from the effector cell populations by magnetic sorting, using a CD8 ϩ cell selection kit (Dynal Biotech GmbH). First, the cells were washed once in PBS supplemented with 2.0% FCS (P2F) and finally resuspended in P2F at a concentration of 1 ϫ 10 7 /ml. Capture beads were added to the cell suspension at a bead to CD8 ϩ T cell ratio of 8:1. Following a 30-min incubation at 4°C, the beads/cells were washed six times with 5.0 ml of P2F. The beads, together with the attached cells, were reconstituted in 200 l of RPMI 1640 medium with 1.0% FCS. To detach the cells from the beads, 20 l of DETACHaBEAD (Dynal Biotech GmBH) was added and incubated for 1 h at 20°C. The released cells were isolated, washed once in R10F, and used as effector cells in 51 Cr release assays (22).

Preparation of target cells for 51 Cr release assay
HLA-A-and -B-matched BLCL, C1R, and C1R cells transfected with various HLA genes were used as target cells in 51 Cr release assays. All cells were infected with Resvir-9 at a MOI of 3 in RPMI 1640 medium, containing 0.1% BSA, 2 mM glutamine, 100 g/ml streptomycin, and 100 IU/ml penicillin (R0.1B). Following a 1-h infection at 37°C, the cells were washed once in R10F and incubated in R10F for 16 h at 37°C. The following day, 10 6 cells were washed once in R0.1B medium and incubated for 1 h at 37°C with 75 Ci of Na 2 [ 51 Cr]O 4 . The cells were then washed three times in R10F and used as target cells. Uninfected cells of the HLAmatched BLCL, C1R, and HLA-transfected C1R cell line were included to determine nonspecific lysis of target cells.
The Ag presentation capacity of HLA-transfected C1R cells was compared with cells of two BLCL. To this end, the minimal peptide concentration was determined for which 50% of the target cells were killed (EC 50 ) by epitope-specific CTL clones. After a 1-h incubation with Na 2 [ 51 Cr]O 4 , the cells were washed twice in R10F and distributed in 96-well V-bottom plates. Next, the cells were incubated with a 10-fold serial dilution of the peptides in R10F for 1 h at 37°C, washed once in R10F, and used as target cells in a 51 Cr release assay. The HLA-A2-, -A3-, -B27-, and -B35-restricted influenza A virus-specific CTL clones were previously described (23)(24)(25), while the HLA-A1-and -B8-restricted influenza A virus-specific CTL clones were established, as described previously (25).

Intracellular cytokine staining
Influenza virus-stimulated PBMC were also used for ICS assay. One hundred thousand PBMC were incubated in 100 l of R10F containing Gol-giStop (monensin; BD PharMingen, Alphen a/d Rijn, The Netherlands) and GolgiPlug (brefeldin; BD PharMingen). In addition, 2 ϫ 10 5 influenza virus-infected (Resvir-9, MOI ϭ 3 or B/Harbin/7/94, MOI ϭ 1) and uninfected HLA-matched BLCL cells and C1R cells with or without HLA transgene were also incubated in 100 l of R10F containing GolgiPlug and GolgiStop and used as stimulator cells. After 30 min at 37°C, the stimulator cells were added to the PBMC for 6 h at 37°C. Next, the cells were washed, stained, and fixed, as previously described (26)

ELISPOT assay
Virus-specific CD8 ϩ T cells and their HLA restriction element were also quantified in ELISPOT assays, which were performed as previously described (15). Before stimulation with uninfected and virus-infected C1R cells expressing single HLA class I molecules, for the induction of IFN-␥ production, CD8 ϩ CD16 Ϫ cells were isolated from PBMC using dynabeads.
51 Cr release assay 51 Cr release assays were performed, as described previously (15). Influenza A virus-specific CTL clones were added to 5 ϫ 10 3 51 Cr-labeled target cells at an E:T cell ratio of 10:1 or 5:1. Isolated CD8 ϩ T cells, obtained from influenza A virus-stimulated PBMC cultures, were also used as effector cells at E:T ratios of 10:1 to 1.25:1. After 4 h at 37°C, the culture supernatants were harvested (Skatron Instruments, Sterling, VA) and radioactivity was measured by gamma counting. The percentage of specific lysis was calculated with the following formula: ((experimental release Ϫ spontaneous release)/(maximum release Ϫ spontaneous release)) ϫ 100. The percentage of influenza A virus-specific lysis was calculated from the percentage of lysis of infected cells minus the percentage of lysis of uninfected cells of at least three wells.

Flow cytometry
Following isolation of CD8 ϩ T cells, the purity of the CD8 ϩ T cell population was tested. Approximately 10 5 cells were washed once in P2F, and incubated in 50 l of P2F containing anti-CD8, anti-CD4 (Dakocytomation), and anti-CD3 mAb for 30 min at 4°C. Next, the cells were washed once in P2F, and at least 10 4 events were acquired using a FACSCalibur flow cytometer. The CD8 ϩ T cell purity of the isolated CD8 ϩ T cells was generally Ͼ90%. Before 51 Cr release and ICS assays, target cells were tested for the expression of transfected HLA genes. C1R-A1 and C1R-A3 cells were incubated with biotin-conjugated anti-HLA-A1/36 and anti-HLA-A3 (BIH0269; One Lambda) mAb, respectively, followed by FITC-conjugated streptavidin (Dakocytomation). C1R-A2, C1R-B8, and C1R-B27 cells were incubated with FITC-conjugated anti-HLA-A2,28 (FH0037; One Lambda), anti-HLA-B8 (FH0536A; One Lambda), or anti-HLA-B27 (B27F50X; One Lambda) mAb, respectively. Finally, C1R-B35 were incubated with culture supernatant of hybridoma 4D12 (ATCC HB-178; American Type Culture Collection, Manassas, VA), and subsequently stained with FITC-conjugated anti-mouse Ig Fab (Dakocytomation). The level of HLA expression on C1R and C1R cells with various HLA transgenes was determined by flow cytometry using saturating levels of anti-HLA-ABC Abs (BD PharMingen). All Ab-staining procedures were performed in P2F at 4°C. At least 10,000 cells were acquired with a FACSCalibur flow cytometer and analyzed with CellQuestPro.

Statistical analysis
To identify statistical differences between IFN-␥ responses restricted by individual HLA-A or -B alleles, the Student t test was performed. To investigate preferential HLA-A or -B usage in the three groups of donors, a Friedman test was performed, comparing the response to individual alleles with a theoretical random contribution of the four HLA-A and -B alleles. Differences in ratio IFN-␥ ϩ cells and TNF-␣ ϩ cells were analyzed with a univariate ANOVA post hoc analysis. Ratio of IFN-␥/TNF-␣-positive cells was determined if both proportions of cytokine-positive cells were more than 1.0%. Values of p Ͻ 0.05 were considered statistically significant.

Validation of HLA-transfected C1R cells
Before using C1R cells expressing HLA transgenes as APCs, their HLA expression and Ag-presenting capacity were tested. As shown in Fig. 1A, all C1R cells transgenic for individual HLA genes exhibited HLA class I expression after incubation with an Ab specific for all HLA-A, -B, and -C alleles. The expression was in the same order of magnitude as in normal BLCL cells (solid lines, Fig. 1A). Control C1R cells also exhibited surface expression of HLA class I molecules to a limited extent as a result of endogenous HLA-Cw4 expression (17,18). This expression was at least 10-fold lower than in the HLA-transfected C1R cells. The expres-sion of HLA-A or -B molecules by the respective C1R cells was further confirmed using allele-specific Ab preparations (data not shown). Furthermore, the HLA expression was not reduced after infection of these cells with influenza A viruses (data not shown).
The concentration of peptide analogous of CTL epitopes required to give 50% of the maximum lysis (epitope concentration (EC) 50 values) of target cells by corresponding CTL clones was used as a measure for Ag-presenting capacity of C1R cells. As shown in Fig. 1B, the EC 50 values ranged from 10 Ϫ7 M for the HLA-B*0801-restricted epitope (NP 380 -388 ) to 10 Ϫ10 M for the HLA-B*3501-restricted epitope (NP 418 -426 ). More importantly, the EC 50 values for the HLA-transfected C1R cells were similar or slightly lower than the EC 50 values measured with HLA-matched BLCL cells, indicating that functional presentation of peptides is comparable.
Using HLA-transfected C1R cells infected with influenza A virus, it was demonstrated in a 51 Cr release assay that the natural CTL epitope was liberated from viral proteins and presented to specific CTL clones (Fig. 1C). In addition, the respective CTL clones only recognized HLA-transfected C1R target cells expressing the matching HLA molecule.

IFN-␥ responses in CD8 ϩ T cells restricted by individual HLA-A and -B alleles
The contribution of individual HLA-A and -B alleles to the influenza A virus-specific CTL response was determined by measuring the proportion of IFN-␥ ϩ CD8 ϩ T cells in influenza A virus-specific PBMC cultures, which were restimulated with HLA-matched BLCL or C1R cells expressing individual HLA-A and -B alleles. An example of such an analysis is shown in Fig. 2.
Using this procedure, PBMC of all (n ϭ 13) donors were tested. In donors of group I (HLA-A1, HLA-A2, HLA-B8, HLA-B35), the proportion of CD8 ϩ T cells producing IFN-␥ after stimulation with HLA-matched BLCL cells was 46% (Fig. 3A). Stimulation with HLA-transfected C1R cells showed that on average 16.4% of these cells were restricted by HLA-B*3501. HLA-A*0201-, HLA-B*0801-, and HLA-A*0101-restricted CD8 ϩ T cells accounted for 10.2, 9.2, and 5.5% on average, respectively. No significant differences, summarized in Table II, were observed between the frequencies of virus-specific CTL restricted by single HLA-A or -B alleles in these donors.
We also compared the magnitude of the responses restricted by shared alleles between donors of different groups. The HLA-B*0801-restricted response was significantly lower in donors of group II (2.9%) than in those of group III (7.8%, p ϭ 0.04) and to a lesser extent in those of group I (9.1%, p ϭ 0.1). The HLA-A*0101-restricted response in group I (5.5%) did not significantly differ from that observed in groups II (2.4%) and III (3.5%). Also, no differences were found between groups in the proportion of IFN-␥ ϩ CD8 ϩ T cells restricted by HLA-A*0201 (10.2 vs 10.5%) or HLA-B*3501 (16.4 vs 15.9%).
To demonstrate preferential HLA usage in individual donors, pie charts were constructed (Fig. 3B) In all four donors of group II, a dominant HLA-B*2705-restricted IFN-␥ response was observed in addition to the HLA-A*0201-restricted IFN-␥ response. The HLA-A*0101-and HLA-B*0801-restricted IFN-␥ responses only contributed, to a limited extent, to the influenza A virus-specific CTL response in these donors. The preferred usage of certain HLA class I alleles in these donors was statistically significant ( p ϭ 0.011).
In group III, the most dominant response was restricted by HLA-B*3501 in donors 1, 2, 4, and 5, while in donor 3 this response was only minor (Fig. 3B). An HLA-A*0301-restricted response was observed in all five donors and was found immunodominant in three donors. Also, the HLA-A*0101-and -B*0801-restricted responses were detected in all donors, but were found to be subdominant. Again, the preferred recognition of certain alleles in these donors was statistically significant ( p ϭ 0.005). epitope; and finally, C1R-B35 cells with the NP 418 -426 epitope and used as target cells for CTL clones specific for the respective epitopes (E:T ratio of 10:1). C, HLA-restricted recognition of influenza A virus-infected HLA-transfected C1R cells by virus-specific CTL. HLA-transfected C1R cells, C1R control cells, and three different BLCL cells were infected with Resvir-9 (MOI ϭ 3) and used as target cells in a 51 Cr release assay. Influenza A virus-specific CTL clones (see above) were used as effector cells at an E:T ratio of 10:1. The results are given as the percentage of lysis of each target cell minus the lysis of virus-infected C1R cells. HLA-genotyped BLCL were included as controls to confirm the HLA restriction of the influenza A virus-specific CTL clones. The percentage of lysis, represented by different shades of gray, is (darkest to lightest) (100 -80%; 80 -60%; 60 -40%; 40 -20%; 20 -0%).
The sum of the IFN-␥ ϩ CD8 ϩ T cells observed after stimulation with four different C1R cell lines expressing the four HLA-A and -B alleles was compared with the number of IFN-␥ ϩ CD8 ϩ T cells observed after stimulation with HLA-matched BLCL cells. As shown in Fig. 3B (bar charts), the sum of the responses restricted by individual alleles (Ⅺ) almost accounted for the total influenza A virus-specific response (f) in most cases. In 11 of 13 donors, the difference was less than 10% and only in donor 4 of group II and donor 3 of group III the differences exceeded 10%.

HLA-A-and -B-restricted lysis of target cells by CD8 ϩ CTL
The contribution of individual HLA-A and -B alleles in the recognition of infected target cells by influenza A virus-specific CTL was also studied in 51 Cr release assays, using infected and uninfected HLA-transfected C1R cells and HLA-matched BLCL cells (Fig. 4). The CD8 ϩ effector cell populations obtained from donors in group I recognized HLA-matched BLCL, C1R-A1, and C1R-A2 cells to a similar extent (average percentage of 65%). The average percentages of specific lysis of C1R-B8 and C1R-B35 were 40 and 30%, respectively.
CD8 ϩ effector cells obtained from donors in group III did not exhibit a clear preferred HLA usage in the recognition of their target cells.

IFN-␥ and TNF-␣ expression in influenza A virus-specific CTL restricted by individual HLA-A and -B alleles
The expression of TNF-␣ was determined as a third functional parameter after restimulation of influenza A virus-stimulated PBMC with HLA-matched BLCL and HLA-transfected C1R cells. The ratio between percentage of IFN-␥ ϩ and percentage of TNF-␣ ϩ cells within the CD8 ϩ T cell fraction for each individual for the different HLA-transfected C1R cells and HLA-matched BLCL cells was calculated (data not shown). The ratio of IFN-␥ ϩ and TNF-␣ ϩ cells following stimulation with HLA-matched BLCL cells is on average 1.1, indicating that most virus-specific cells produce both cytokines upon restimulation. However, stimulation with HLA-transfected C1R cells resulted in ratios starting from 1.2 for C1R-B27 to 1.8 for C1R-A2, C1R-A3, C1R-B8, and C1R-B35, while stimulation with C1R-A1 cells resulted in an average ratio of 0.7. This indicates that some HLA-A*0101-restricted influenza A virus-specific CTL produce TNF-␣, but not IFN-␥. This ratio was significantly different from the average ratio of C1R-A2 ( p ϭ 0.011), C1R-B8 ( p Ͻ 0.001), and C1R-B35 ( p ϭ 0.001).

Discussion
The contribution of individual HLA-A and -B alleles in influenza virus-specific CTL responses was determined in groups of HLA class I-matched donors. It was shown that influenza virus-stimulated PBMC of HLA-A-and -B-matched donors preferentially recognized certain HLA alleles, which depended on the type of virus studied. Furthermore, it was demonstrated that cytokine production profiles of CD8 ϩ CTL depended on their HLA class I restriction elements.
Before investigation of CTL responses restricted by individual HLA-A and -B alleles, the use of HLA-transfected C1R cells as APC was validated. The level of expression of HLA molecules was similar in the different C1R cells as measured with saturating amounts of Ab. These cells were also capable of processing and presenting influenza A virus-specific CTL epitopes to a similar extent as EBV-transformed BLCL. In addition, only small differences in susceptibility of the individual C1R cells for infection with influenza A virus were found (between 55 and 80% infected), with the exception of C1R-B35. This, however, did not interfere with our ICS assay, because an excess number of stimulator cells was used to stimulate influenza virus-specific CTL in the in vitro stimulated PBMC cultures. In contrast to the use of BLCL matched for a single HLA allele, the use of C1R expressing single HLA alleles prevents possible competition for processing/presentation between overlapping epitopes (27). In EBV-transformed BLCL cells or C1R cells, competition for available HLA-A or -B alleles between epitopes from influenza virus and EBV may occur. For example, an immunodominant HLA-B*0801-restricted CTL epitope of EBV (EBNA3A 325-333 ) could interfere with the presentation of HLA-B*0801-restricted influenza virus epitopes. However, influenza A virus-infected C1R-B8 cells were readily killed by HLA-B*0801-restricted influenza A virus-specific CTL, indicating that competition between epitopes of these two viruses does not constitute a significant problem.
Using the HLA-transfected C1R cell lines, it was shown that the influenza A virus-specific CTL response in these donors measured by intracellular IFN-␥ staining was dominated by HLA-B*2705and HLA-B*3501-restricted CTL in groups II and I/III, respectively. Based on previous findings with synthetic peptides (15), it was anticipated that the HLA-A*0201-restricted response, presenting the immunodominant epitope M1 58 -66 , would be recognized preferentially. Therefore, the observed hierarchy of virus-specific responses indicates that other yet unidentified HLA-B*2705-and HLA-B*3501-restricted epitopes exist. Recently, an immunodominant HLA-B*3501-restricted epitope (NP 418 -426 ) was identified (25), which may have contributed to the preferred usage of HLA-B*3501 in the CTL response. In addition to the dominance of HLA-B27 and HLA-B35, a large proportion of T cells was specific for epitopes presented by HLA-A*0201 and HLA-A*0301. Most likely the presentation of the immunodominant epitope M1 58 -66 has contributed to the dominance of the HLA-A*0201-restricted response. The large proportion of HLA-A*0301-restricted CTL indicates that more unknown immunodominant epitopes exist, because the HLA-A*0301-restricted NP 265-273 epitope has been shown to be subdominant in response to influenza A viruses (15). Indeed, a novel HLA-A3-restricted CTL epitope was identified recently, which could contribute to the influenza A virus-specific CTL response (28). Finally, the IFN-␥ response specific for epitopes presented in the context of HLA-A*0101 and HLA-B*0801 contributed little to the overall influenza A virus-specific CTL response, which is in agreement with previous work (29). The small contribution of these alleles to the overall CTL response could be caused by the presence of alleles HLA-B*2705 or HLA-B*3501 presenting immunodominant epitopes. This hypothesis was supported by a high proportion of IFN-␥ ϩ HLA-B*0801-restricted cells in an HLA-A*0101, HLA-B*0801 homozygous donor (data not shown). In addition, the HLA-B*0801-restricted response was lower in donors of group II (HLA-B27 ϩ donors) than in donors in the two other groups (HLA-B35 ϩ donors), which is in agreement with the poor recognition of the HLA-B*0801-restricted NP 380 -388 epitope in HLA-B27 ϩ individuals (15,27). These data also indicate that more immunodominant high affinity HLA-B*0801-restricted epitopes are not existing for influenza A virus.
Changes in epitope specificity of the virus-specific CTL response as a result of differences in HLA expression profiles were previously reported in mice infected with influenza A virus (3,11,30). In these studies, the expression of an H-2K b allele (either transgenic or through breeding with C57BL mice) resulted in a reduced H-2 k -restricted response.
A possible skewing of the CTL response in favor of certain HLA restrictions during in vitro stimulation of PBMC in the presence of rIL-2 was excluded, because it was found in a previous study that the hierarchy of peptide-specific CTL responses measured ex vivo in PBMC correlated with the hierarchy of lytic activity measured after in vitro expansion of virus-specific CTL using the same stimulation protocol (15). Furthermore, ELISPOT assays were conducted with CD8 ϩ CD16 Ϫ cells isolated from PBMC of several donors from groups I and II to determine preferential HLA usage ex vivo. In the donors of group II, the same hierarchy of preferred HLA usage was observed ex vivo, as was observed after in vitro expansion of virus-specific CTL. In group I, without preferential HLA usage after in vitro expansion of virusspecific cells, ex vivo no-preferred HLA usage was also demonstrated (data not shown).
In a previous study (15), we reported a lower influenza A virusspecific immune response in HLA-A*0201-negative donors (group III). Also in the current study, the average number of IFN-␥ ϩ cells following restimulation with HLA-matched BLCL is lower in HLA-A*0201-negative donors than in HLA-A*0201-positive donors, although this difference was not statistically significant ( p ϭ 0.28). This is partly explained by the exclusion of two HLA-A*0201-negative donors with an HLA-B*3503 instead of an HLA-B*3501 genotype.   Using influenza B virus for the stimulation of PBMC, we demonstrated that the preferential HLA usage is dependent on the virus studied. In contrast to the influenza A virus-specific CTL response, the HLA-B*0801-restricted response specific for influenza B virus was highly immunodominant, followed by HLA-B*2705. The HLA-A*0101-and HLA-A*0201-restricted responses were shown to contribute little to the overall influenza B virus-specific response. It is difficult to correlate these responses to known (immunodominant) epitopes because only four influenza B virus CTL epitopes are known (31,32), of which three are presented in the context of HLA-B8 (32). These data indicate that the available epitope repertoire determines the outcome of the CTL response and preferred HLA usage.
In contrast to other studies addressing the preferential use of HLA molecules in CTL responses, we used virus-infected human cells expressing a single HLA allele, which accounts for the full repertoire of CTL epitopes presented by these HLA molecules. The preferential usage of certain alleles in the virus-specific CTL response has also been reported for EBV-specific CTL responses. These data demonstrated that certain HLA class I alleles were dominantly recognized, such as HLA-B8, HLA-A11, or HLA-B44, while HLA-A1 was not (7,9). For HIV-specific CTL responses, it was shown with synthetic peptides that the HLA-A2restricted response hardly contributed to the overall HIV-specific response (12,14).
Most of these studies used IFN-␥ production to identify epitopespecific CTL. However, our data suggest that CTL differ in their ability to produce cytokines depending on the epitopes recognized and/or the HLA molecules presenting these epitopes. Influenza A virus-specific HLA-A*0101-restricted CTL produced less IFN-␥ and more TNF-␣ than CTL restricted by other HLA molecules. Therefore, some caution should be exercised in interpreting frequencies of CTL based on IFN-␥ production alone. The preferred HLA usage, as demonstrated by quantification of virus-specific IFN-␥ ϩ CD8 ϩ T cells and the lytic activity, of these cells differed, which also could be attributed to functional differences of HLA-A1-and HLA-B35-restricted CTL in particular. To our knowledge, this is the first study to identify differences in cytokine production in CD8 ϩ CTL effector cell populations. Previous studies identified functional differences between EBV-, CMV-, and HIVspecific CTL, based on perforin and surface marker staining (33,34). These studies also showed reduced killing of two HIV tetramer-positive cell populations in comparison with a CMV tetramer-positive population. Our data provide evidence also that acute viral infections, like influenza virus, induce functionally different CD8 ϩ CTL populations. At present, it is unclear what the underlying mechanism is for differential cytokine expression in virus-specific CTL and how epitope specificity and the HLA molecules control this. Additional studies are required to further characterize these functional differences in CTL function and to investigate the implications of differential cytokine expression.
Thus, collectively, the present study has shown in donors of well-defined HLA genotypes that: 1) in response to virus infection, CTL responses are induced that use certain HLA molecules preferentially, depending on the available repertoire of CTL epitopes; 2) CTL exhibit differential cytokine expression depending on their epitope specificity and/or HLA restriction.