This study took advantage of the sequences of influenza viruses circulating at different times, to compare the CD4 T cell responses generated by different antigen exposure histories. Most people have encountered influenza at a very early age 
, and it is a rarity to find an adult without an anti-influenza response 
. However, some peptides unique to the recent pandemic CA/09 influenza virus would be expected to not cross-react to previous strains. Our peptide pool strategy used information from different H1N1 influenza strains to selectively stimulate T cells specific for epitopes on viruses circulating for a long time, and to contrast these responses with those against epitopes unique to the recent pandemic CA/09 influenza – these latter responses should be at least partially enriched for primary responses. In both the cross-sectional study and the recent-infection study, qualitatively different responses, i.e. increased numbers of IFNγ−
cells, occurred against the peptides from the recent influenza CA/09. Thus there may be a difference in the quality of recently induced memory CD4 T cells in comparison to memory cells that have persisted over many years and/or may have been multiply boosted.
Despite major advances in predictive algorithms, such as using artificial neural networks or consensus predictions, there are currently no algorithms available that accurately predict the immunogenicity of a peptide across the multitude of HLA class II alleles 
. MHC alleles of the subjects in our studies were not determined, but probably included normal diversity. Therefore a comparative analysis of the possible epitopes within the various peptide pools was not feasible. Accordingly, we initially used a one amino acid difference cutoff to assess T cell function independent of the contributions from HLA type and TcR so that we could identify virus-specific responses within the general population. This method is inclusive 
and should identify many peptides identical between strains. Although some peptides with only slight differences, e.g. one amino acid, may still cross-react antigenically, this method should also provide partial enrichment for non-cross-reactive epitopes for each individual strain of influenza.
In the first (normal donor) study, the responses to conserved (identical) peptide pools should comprise a mixture of long-term memory and recently-boosted memory, depending on the individual infection history. NC/99ni responses should include the same T cell phenotypes, possibly with a bias towards more resting long-term memory because some T cells would recognize epitopes that do not cross-react with recent influenza strains. In contrast, responses to the CA/09ni peptide pool should be a mixture of recently-boosted memory (specific for non-identical but cross-reactive epitopes), and recent primary responses against new epitopes in subjects exposed to or vaccinated with the CA/09 virus. This therefore suggests that the trend towards more IFNγ−IL-2+ cells in the responses to the CA/09ni pool may be due to recent primary responses.
In the second (infection) study, the most divergent (potentially least cross-reactive) peptides of the pandemic H1N1 were identified by ranking, based upon amino acid substitution matrices that take into account the occurrence of pair-wise substitutions and physicochemical properties of each amino acid 
. Soon after pandemic H1N1 infection, responses were higher against all peptide pools, even the NC/99ni pool, suggesting that cross-reaction between NC/99 and CA/09 peptides did occur in this pool. The increased NC/99ni and memory flu peptide responses maintained the bias toward IFNγ+
producing cells seen in the healthy donor study, and the pre-pandemic blood samples. The CA/09ni diff peptides may also have contained cross-reactive peptides, as this pool also induced Th1-biased responses (data not shown). However, the CA/09ni vdiff pool, representing the peptides least likely to cross-react with epitopes in previous influenza strains, revealed a strongly enhanced response biased towards IFNγ−
CD4 T cells after PCR-confirmed CA/09 infection.
The investigation of naïve and recently infected subjects by enrichment of virus-specific sequences and 15-color flow cytometry is an effective method for rapid analysis of an emerging pathogen, as only the protein sequences from a representative isolate are required to generate peptide libraries, and this method identifies the relative quantity and quality of preexisting or newly-induced cell-mediated immunity. The validity of this approach is supported by two sets of results – first, recent infection induced larger expansion of the responses to two pools of CA/09ni peptides than the responses to conserved peptides, and second, the response to the CA/09ni vdiff pool was qualitatively different.
In mice, repeated stimulation of CD4 T cells reduced the percentage of cells secreting effector cytokines in vitro
, and this was associated with reduced protection from influenza challenge. Also in mice, differential numbers and functional phenotypes of memory CD8 T cells were generated by heterologous prime-boost regimens differing in the boosting vector, antigen load, and level of inflammation 
. Interestingly, a higher proportion of IFNγ+
CD8 T cells were maintained following multiple rounds of stimulation, with a gradual loss of IL-2 production 
and changes in the regulation of additional genes. Consistent with these mouse studies, our human data also suggest that the T cell response can further evolve with repeated stimulation.
The difference between the response to the recent CA/09 epitopes and long-term memory may be due to the difference between long-term, multiply-boosted responses versus primary responses, as suggested above. A second possibility is that recent responses have a higher representation of IFNγ−
cells than long-term resting memory, because these cells are reduced in the circulation over time, due to shorter half-lives, conversion, or distribution to tissues. A third possibility is that A/California/09 may generate a different type of immune response than the previously circulating H1N1 strains. Differences in the NS1 protein may alter the production of type 1 interferons and hence other anti-viral genes 
, and changes in the hemagglutinin receptor-binding domain may alter cellular tropism and infectivity 
, potentially modifying antigen capture and functions of dendritic cells, and altering T cell differentiation.
What is the phenotype of the IFNγ−
cells induced by primary CA/09 immunization? These cells may be similar to the IFNγ−
Thpp cells that we have described previously in mice 
and humans 
. In addition to the distinctive cytokine profile, Thpp cells are uncommitted with respect to further acquisition of effector functions, and can differentiate into either Th1 or Th2 cells 
. This uncommitted Thpp phenotype is consistent with the T stem cell phenotype proposed as a long-lived, self-renewing memory population 
, although the commitment state of the influenza-specific IFNγ−
cells has not yet been tested. These cells may also be included within the CD4 central memory (Tcm) population 
. Expression of CD62L, often used as an identifying characteristic of Tcm cells 
, was not tested in the current study because samples were frozen before analysis, and CD62L is variably lost from the cell surface after freezing and thawing. In previous studies, Thpp cells were variable in expression of CD62L. The IFNγ−
influenza-specific cells in the present study did not express CXCR5, and so do not fit a Tfh definition 
or the more stringent Tcm definition used in one study 
The CD4 Tcm, CD4 stem cells and Thpp cells have been proposed to act as a reservoir of memory cells with high proliferative potential and low effector function, that can give rise during subsequent infections to effectors with stronger effector functions but more limited proliferative capacity 
. Thus a response to either vaccination or infection may be most effective when balanced to include both effector memory cells that can immediately respond with strong effector functions during a subsequent infection, as well as Thpp-like cells that provide long-term expansion and preservation of the memory pool, and can differentiate into fresh effector cells.
The balance between effector memory and highly proliferative stem-cell like populations may be particularly important for influenza, as most people are exposed repeatedly to influenza infections, and many are also immunized on a yearly basis. As antibody epitopes are changed rapidly on influenza variants, cross-reactive T cells may be an important component of the response both as helpers for antibody and CTL responses, and also as effectors in their own right 
. As different influenza vaccine strategies are pursued, including adjuvanted vaccines 
and a pan-influenza vaccine 
, it may be important to maintain a balance between effectors and the IFNγ−
producing cells, similar to that observed with tetanus toxoid and other protective protein vaccines