We examined antibody and T cell responses induced by subcutaneous vaccination with two forms of preparation, whole particle vaccine and split vaccine, using immunologically naïve cynomolgus macaques for influenza virus infection. The macaque model enabled evaluation of vaccine efficacy without consideration of pre-existing immunity, which has been difficult to exclude in human studies 
. Since most inactivated vaccines are inoculated subcutaneously or intramuscularly in humans 
, we subcutaneously inoculated vaccines into macaques to reflect results in the present study to human vaccination usage. As a result, the whole particle vaccine derived from our virus library was more effective for conferring memory immune responses and prohibiting virus replication than was the split vaccine.
We previously revealed that pandemic (H1N1) 2009 influenza virus caused viral pneumonia in cynomolgus macaques, which resembled pneumonia in human patients during the pandemic 
. Furthermore, we were able to analyze immune responses in macaques using antibodies against human molecules 
, whereas reagents for immunological analyses were not sufficient in ferrets that showed human-like clinical symptoms by human influenza viruses. Therefore, we considered the macaque models to be suitable for extrapolation of responses and efficacy of vaccines against pandemic influenza virus infection in humans.
As shown in , CD8+
T cells of individual vaccinated macaques responded to different sets of NP peptides since the macaques used in the present study were not inbred. To compare peptide-specific T cell responses between two vaccinations, we required a parameter shared by all studied macaques. Therefore, we selected macaques carrying Mafa-A1*052
02 and identified NP262-270 as a peptide that induced CD8+
T cell responses in all vaccinated macaques. The variety of stimulatory peptides except NP262-270 among macaques might be due to antigen presentation on other Mafa class I molecules, i.e., the second allelic product of Mafa-A1 other than Mafa-A1*052
02 and Mafa-B (). Therefore, we examined CD8+
T cell responses specific for NP262-270 plus Mafa-A1*052
02 to compare memory T cell responses in macaques vaccinated with the whole particle vaccine with those in macaques vaccinated with the split vaccine.
One of the peptides that bound to Mafa-A1*052
02, NP#74, contained a known CTL epitope (ASNENVETM, NP366-374) bound to mouse MHC class I H-2Db
, in which N at position 5 and M at position 9 are known as main anchor residues and N at position 3 is a minor anchor 
. The results presented in showed that anchor positions of NP262-270 were aligned with the same intervals as those seen in NP366-374 bound to H-2Db
. Therefore, it is possible that the Mafa-A1*052
02 binding motif is similar to that of H-2Db
, i.e., hydrophobic or non-polar residues at positions 3, 5 and 9. This was supported by analyses of peptides eluted from Mafa-A1*052
02 (). Based on these results, the Mafa-A1*052
02 binding motif appeared to be hydrophobic residues (L, I or V) at position 3, hydrophobic residues (L or I) at position 5 and hydrophobic or aromatic residues (L, V, M, F or Y) at position 9 as anchors.
Using macaques carrying Mafa-A1*052
02, we showed advantages of the whole particle vaccine in three aspects, immunological memory, immunogenicity and cross-reactivity. Firstly, we proved the advantage of the whole particle vaccine in memory CTL responses. A higher percentage of IFN-γ-positive CD8+
cells specific for the NP262-270 peptide was observed in macaques vaccinated with the whole particle vaccine than in macaques without vaccination and macaques vaccinated with the split vaccine (). Therefore, it is thought that memory T cells were maintained in macaques inoculated with the whole particle vaccine and expanded promptly after challenge infection.
The whole particle vaccine also induced immunological memory in antibody responses more effectively than did the split vaccine. Neutralizing activity of sera against the challenge strain increased 2 days after the challenge infection in macaques vaccinated with the whole particle vaccine, whereas increase of neutralization activity against the challenge strain in sera from macaques vaccinated with the split vaccine was detected on days 8 to 10 after challenge infection (). This was only 2 days earlier than that in macaques without vaccination. Therefore, the whole particle vaccine generated immunological memory in both cellular and humoral responses more effectively than did the split vaccine. Since recall memory B cell responses were impaired in MyD88-deficient mice, Toll-like receptors (TLR) stimulated by whole particle vaccines would play a critical role in generation of memory B cells 
Secondly, as previously shown in mouse and human studies 
, the whole particle vaccine induced higher titers of neutralizing antibody in macaques than did the split vaccine (). This indicates that inactivated whole viral particles were more immunogenic than soluble split antigens in macaques. Several studies have indicated that a virus-like particle (VLP) antigen was superior to a soluble protein in inducing Th1 and CTL responses with effective antigen cross-presentation by dendritic cells 
. We also showed that one immunization with the whole particle vaccine induced neutralizing antibody at a level similar to that induced by two immunizations with the split vaccine. This result suggests that we could reduce the dose of vaccination by using whole particle vaccines to save vaccine amounts and distribute vaccines to more people than by using split vaccines.Not only formulation of antigens but also additional molecules included in the whole particle vaccines were thought to affect the immunogenicity. IL-12 and tumor necrosis factor (TNF)-α production by human dendritic cells was enhanced by whole particle vaccines but not by subunit vaccines 
. This response might be partly mediated by TLR signals 
. Inactivated whole virus particles showed adjuvant effects by activating the TLR7-MyD88 pathway in plasmacytoid dendritic cells 
. Type I interferon production by plasmacytoid dendritic cells was crucial for induction of primary B cell and CD4+
T cell responses and cross-presentation of antigen by dendritic cells, resulting in polarization to Th1 cells and activation of CTL 
. The TLR signal also stimulated naïve B cells directly and enhanced B cell responses 
. Concordant with the previous reports, we showed that IgG1 responses, one of the hallmarks of Th1 responses in human immune responses, were induced more effectively in cynomolgus macaques after vaccination with the whole particle vaccine than with the split vaccine (), indicating that whole particle vaccines including virus RNA as a natural adjuvant enhanced immunogenicity and induced Th1 responses 
. Therefore, whole particle vaccines do not necessarily need inoculation with an additional adjuvant.
Thirdly, sera from macaques inoculated with the whole particle vaccine showed neutralizing activity against not only the vaccine strain but also the challenge strain ( and ), resulting in lower virus titers after challenge in macaques vaccinated with the whole particle vaccine than in macaques vaccinated with the split vaccine (). In the present study, similarities of HA and NA between the vaccine strain and the challenge virus strain were 89% and 83% at the amino acid level, respectively. These results indicated that whole particle vaccines induced not only a larger amount of specific antibody but also a more broadly cross-reactive antibody than did split vaccines 
. This might be explained by results of previous studies showing that whole particle vaccines and split vaccines activated different B cell repertoires (clones) 
and that two formulations of vaccines induced differentiation of distinct T helper cells such as Th1 and Th2 as discussed above, resulting in differences in the number of reacting B cells and affinity maturation 
. Furthermore, since sequences of NP including CTL epitopes were conserved in influenza A viruses compared with HA and NA, CTL specific for NP and other internal proteins might react to heterosubtypic influenza viruses 
. Thus, whole particle vaccines would be effective even if viral antigens were changed by gene mutations within a certain range. Therefore, our virus library containing 144 combinations of 16 HA and 9 NA would be useful when pandemic vaccines are prepared as whole particle vaccines to induce both antibody and CTL responses with broad cross-reactivity 
In summary, we showed that inactivated whole virus particles from the virus library were effective against pandemic (H1N1) 2009 influenza virus infection using the cynomolgus macaque model. Whole particle vaccines were superior for induction of memory antibody and CTL responses, which were confirmed at the level of peptide-specific responses. We are examining other Mafa class I and class II types and we are breeding macaques carrying specific MHC haplotypes. These macaques will be useful for evaluation of vaccination and analysis of both cellular and humoral immune responses in future studies.