Previously, we have performed a genome-, pathogen-, and HLA-wide search for conserved CTL epitopes in influenza A virus 
. However, this search for conserved
CTL epitopes skewed the selection towards the polymerase and nucleoprotein, whereas the classical antibody targets, HA and NA, were found to contain only a few (8) of the 167 predicted CTL epitopes 
. Instead of searching for conserved CTL epitopes, we attempted in the present study, to select a number of predicted HLA-I binding influenza A CTL epitopes, which constitute a broad coverage of all available influenza A strains. According to this criterion, most of the predicted CTL epitopes from the PB1, PB2, and PA proteins were found to be shared with those tested in our previous publication 
, and for the purpose of novelty, these proteins were therefore excluded from the present study.
To increase the chance for the discovery of new peptide epitopes, we tested the peptides in all HLA-I supertype matched donors available to us. In the present work, we discovered 20 new immunogenic peptides, and confirmed one known, of 131 peptides tested as compared to the discovery of 10 new and 3 known peptides of 167 flu peptides tested in our previous report 
. In the latter case, the peptides were only tested in a few of the available HLA-matched donors.
In recent work on pox-derived epitopes, we found that immune responses of donor PBMC in vitro
, as measured by IFNγ ELISPOT towards HLA-I binding 9 mer peptides, were either CD8+
T cell-dependent and that the latter appeared to be restricted by HLA-II molecules 
. This observation led us to investigate whether the predicted HLA-I binding flu peptides of the present study induce CD4+
T cell-dependent responses. The key finding of the present study is that 16 of 131 9mer peptides derived from influenza A viral proteins induce CD4+
T cell dependent responses in vitro from presumably immune donors. These responses were, with one exception, all blocked by anti-HLA-II antibody. In addition, 8 peptide responses were blocked by an anti-HLA-DR antibody and 6 peptide responses by an anti-HLA-DP antibody. Surprisingly, only 5 peptide responses (including the known peptide PF130 
), were blocked by a HLA-I antibody. For selected peptides, CD4+
T cell depletion experiments showed that CD4+
T cell responses were blocked by anti-HLA-II antibody, whereas CD8+
T cells were blocked by anti-HLA-I antibody. As shown in , the anti-HLA-I antibody in fact showed partially blocking of reactivity for some peptide epitopes (PF106 and PF141) that, according to the cell depletion experiments () induced CD4+
T cell, but not CD8+
T cell, responses. Such partial blocking might reflect anti-HLA-I antibody-mediated apoptosis of activated effector CD4+
T cells as previously demonstrated 
. The fact that CD4+
T cell-mediated responses against PF137 is not inhibited by any of the two anti-HLA antibodies might suggest an excessively high stimulatory binding avidity of peptide specific CD4+
It is generally accepted that HLA class I binding peptides are composed of 8-11 amino acids, whereas HLA class II binding peptides consist of 15–20 amino acids being recognized by CD8+
T cells respectively 
. Both HLA-I and -II molecules bind to primary and secondary peptide anchor motifs covering the central 9-10 amino acids. Thus, considering this common structural basis for peptide binding to HLA-I and –II molecules, the present finding of 9mer peptide binding to HLA-II molecules is not unexpected (also documented in the immune epitope database www.Immuneepitope.org
). As mentioned above, we have previously reported that high-affinity HLA-I binding variola-derived 9mer peptides (KD
<6 nM) induce CD4+
T cell responses ex vivo more than 30 years post-vaccinia virus vaccination, which can be blocked by anti-HLA-II antibody 
. The same phenomenon was observed here: in silico
predicted HLA-I binding 9mer peptides – this time derived from influenza A viral proteins - induce CD4+
T cell mediated responses which appear to be HLA-II restricted as T cell responses are totally blocked by a pan HLA-II antibody. In contrast to our previous study 
, we found no correlation between CD4+
T cell reactivity and HLA-I binding affinity of peptides. The induction of immune responses in the present study was not limited to high-affinity HLA-I binding peptides. Rather, we found examples of peptides with intermediate, low or no binding affinities for its HLA-I allele. Such peptides were all capable of stimulating strong CD4+
T cell responses, again suggesting that these 9mer peptides are presented by HLA-II molecules. As indicated from data in and only half of the HLA-II restricted responses were observed in more than one donor whereas four of the five HLA-I restricted responses were observed in at least two donors. This difference might reflect the relatively low binding affinity of 9mer peptides for HLA-II () as opposed to their binding affinity for HLA-I (), thereby making the ELISPOT assay less sensitive as a readout for antigenic HLA-II binding 9mer peptides.
Using a high-throughput HLA-II peptide binding assay, recently developed in our laboratories 
, we found that 3 of 8 peptides, for which reactivity was blocked by anti-DR antibody, bind to donor HLA-DR subtypes. These numbers are higher than expected since only one quarter of the HLA-DR alleles expressed by the peptide reactive donors was assayed for peptide binding. Regarding the peptides, for which reactivity was blocked by anti-DP antibody, it was quite surprising to find that none of these peptides bind to DPA1*0103/DPB1*0401 although the majority of donors in our material express this common HLA-II subtype. At present we have no explanation for this negative observation.
Frahm et al. 
have recently tried to determine the HLA class I promiscuity of previously well-defined CTL epitopes by testing responses in 100 subjects to a set of 242 HIV and EBV-derived CTL epitopes using PBMC in IFNγ ELISPOT assays, regardless of the individual's HLA type. Fifty percent of all positive responses were detected in individuals who did not express originally described restricting HLA-I allele. The authors concluded that epitope presentation and CTL recognition might occur frequently in the context of alternative HLA class I alleles. Although some alternative HLA-I restrictions were confirmed experimentally, the majority were not identified, but inferred by statistical methods 
. Our previous 
and present results showing HLA-I binding 9mer peptides capable of activating CD4+
T cell dependent responses, may suggest that some of the “alternative restricted” responses described by Frahm et al. 
T cell recognition of epitopes restricted by HLA-II.
Assarsson et al 
have used the similar strategy as ours to identify influenza A virus-derived epitopes, but none of their discovered epitopes matched those discovered in our study. The discrepancy between the two studies might be explained by the fact that Assarsson et al mainly focused on epitopes derived from conserved sequences while we focused on epitopes in influenza virus regardless of their conservancy. Also they used freshly harvested PBMC whereas we used in vitro restimulated PBMC which increases the frequency of influenza virus-specific T cells thereby enhancing the detection sensitivity. Indeed, for some antigenic peptides in 
, the numbers of SFC in the ELISPOT assay are very low. In line with this, some epitopes including two HLA-I restricted (PF-96 and PF-132) and five HLA-II restricted epitopes (PF-103, PF-109, PF-146, PF-152 and PF-PF-141) identified in our study, were also included in 
but failed to show antigenicity. Furthermore, Assarsson et al only discovered 54 epitopes out of 4080 peptides tested (discovery rate: 1.3%), while we identified 21 out of 131 peptides (discovery rate: 16%) by using peptide restimulated T cells. Therefore, the in vitro restimulation might be needed in PBMC of unvaccinated individuals prior to performance of ELISPOT assays in order to increase the detection sensitivity. Intriguingly, Assarsson et al. did not discover any HLA-II restricted epitopes among 38 HLA-I binding epitopes (8–11mer), whereas we identified 16 of 21 HLA-I binding 9mer peptides in our study as HLA-II restricted CD4+
T cell epitopes. However, they excluded some 8–11mer peptides which induced reproducible positive responses, but showed poor binding ability to the relevant HLA-I alleles. It would be interesting to know if the responses induced by these latter peptides are CD4+
T cell dependent.
We propose that studies, which employ ‘reverse immunology’ to monitor HLA class I responses against HLA-I binding peptides by use of IFNγ ELISPOT assay, should take class II-restricted, CD4-dependent T cell responses into account. Our present and previous data 
suggest that HLA-I binding peptides might stimulate CD4+
T cell immune responses restricted by HLA-II molecules. Thus, ELISPOT-based analyses of reactivity against 9mer class I binding peptides should always include either anti-CD4/CD8 blocking or CD4/CD8 T cell subset depletion experiments or, alternatively, perforin- or granzyme B-based ELISPOT analyses to obtain the true phenotype of the antigen-specific T cells.
In the present and previous 
studies we have identified a total of 30 new antigenic flu-derived 9mer peptides (large proteins PB1,PB2,PA,NP,NA,HA and small proteins M1, M2, NS1) potentially recognized by the majority of humans disregarding their HLA allotype and group. We now plan to initiate animal vaccine studies in flu infected HLA transgenic mice to assay for the protective/therapeutic efficacy of the peptides. If a clinical effect is obtained, the peptides might be of use as vaccine candidates in future influenza pandemics.
In conclusion, by the use of PBMC from healthy adult donors, twenty one 9mer peptides derived from influenza A viral proteins were found to induce T cell responses in an IFNγ ELISPOT assay. Only 5 of the peptides induced HLA-I restricted CD8+ T cell responses. The remaining 16 peptides, of which 3 peptides were shown to bind to HLA-DR, induced CD4+ T cell responses apparently restricted by HLA- II molecules.