Influenza virus-like particle vaccines described in this report are non-replicating particle-based vaccine candidates for influenza based upon a influenza A H5N1 clade 1 and 2 isolates. Our results show that BALB/c mice immunized with VLP vaccines were effectively protected from disease and death when challenged with viruses with antigenically distinct HA and NA proteins from H5N1 influenza viruses. These results highlight the potential of VLP vaccine as an effective immunogen and delivery system for influenza antigens, particularly to the respiratory tract. Our VLPs have the advantage of inducing strong humoral and cellular immune responses against multiple influenza viruses without the need of a supplemental adjuvant. The inclusion of the highly conserved M1 protein is also advantageous, since CD8+ T cells against conserved epitopes within M1 can contribute to protection against morbidity and mortality from influenza 
. Pools of peptides representing M1 were used to identify cellular responses elicited by our VLPs (). In addition, these vaccines have the potential to elicit protective immune responses as effectively as live-attenuated influenza without safety issues associated with the isolation, production, and delivery of live vaccines 
Currently licensed seasonal influenza vaccines elicit immunity that is subtype- and sometimes strain-specific and do not protect against avian H5N1 viruses with pandemic potential. Our influenza VLPs are a new generation of egg-independent candidate vaccines expressed from insect cells infected by a recombinant baculovirus that encodes genes for three influenza virus proteins, HA, NA, and M1 
. These VLPs may have an advantage over HA-only based vaccines by the inclusion of these additional viral proteins, especially against evolving H5N1 isolates from various clades. These VLPs elicited anti-NA antibodies against an N1 NA protein that was not matched to the vaccine. Antibodies against the same subtype of NA, but not the exact NA molecule, can contribute to protective immune responses 
. When delivered via parenteral or mucosal routes, VLPs may be particularly effective immunogens at priming T cells and targeting antigen-presenting cells, as described for other VLPs 
, as well as inducing high titered antibody responses 
. Another key factor is the authentic presentation of surface HA and NA in native, three-dimensional conformation. Recent clinical trials of human papillomavirus (HPV) VLPs have led to FDA approval 
and therefore, this may bode well for the approval of additional VLP-based vaccines, including influenza VLPs. Our influenza VLPs are easy to develop, produce, and manufacture. They are not labor-intensive and they do not require costly production schemes typically associated with manufacturing vaccines in eggs. VLP vaccines, like other recombinant influenza vaccines, are particularly advantageous to address future pandemics because these vaccines 1) need shorter lead times for development of vaccines matched to circulating strains of viruses, 2) use recombinant DNA technology to alleviate safety restrictions and bottlenecks associated with dependence on live viruses, 3) use cell culture based methods with disposable bioreactors to provide rapid response and higher yields (scalable and transferable) for improved surge capacity.
In studies reported here, all mice vaccinated intramuscularly with either vaccine or intranasally with the VLP vaccines survived challenge with lethal doses of reassortant viruses. However, intranasal delivery of the VLPs did elicit a broader immune response than the same vaccine delivered intramuscularly. Cellular responses, in particular, were reduced in the lung mucosa in mice vaccinated intramuscularly compared to intranasally vaccinated mice. The ability to elicit mucosal immune responses in the respiratory tract, including the lungs, is desirable for an influenza vaccine. Neutralization of influenza by pre-existing sIgA and IgG in the lung reduces infection of susceptible epithelial cells 
and thereby reduces the deleterious effects induced by elevated cytokine levels, which typically lead to the development of fever and respiratory symptoms 
. H5N1 infection in humans activates cytokine/chemokine secretion resulting in the occurrence of a “cytokine storm” that may contribute to the severity of disease by these viruses 
. The levels of these pro-inflammatory cytokines, triggered by influenza gene products, are higher during H5N1 virus infection compared to seasonal influenza virus infection 
. Therefore, vaccines, such as VLPs studied here, that prevent infection by antibody or quickly clear infected cells by cell-mediated immune responses 
in the lung mucosa may blunt the activation of this deleterious immune activation by reducing viral replication.
Compared to particulate antigens, intranasal vaccination of soluble proteins, in the absence of an adjuvant, induces low or undetectable immune responses in rodents and primates 
. In the nasal mucosa, VLPs are most likely phagocytosed by microfold epithelial cells (M cells) in the nasal lumen and then directly deposited to the NALT (nasal associated lymphoid tissue) via M cell transcytosis 
, which preferentially drains into lymph nodes. This process induces strong local (NALT) and distant immune responses in both peripheral and mucosal immune compartments 
. In contrast, soluble antigens can penetrate the nasal epithelium and directly interact with dendritic cells, macrophages and lymphocytes and then these antigens are transferred to posterior lymph nodes 
. Soluble antigens can bypass the NALT and be directly fed into superficial lymph nodes by antigen presenting cells in the nasal lumen resulting in a lower local immune response 
. Therefore, VLP immunogens can potentially interact directly with the mucosal immune system to elicit high titer immunity.
Several approaches are in progress to develop vaccines against H5N1 viruses. To date, products tested in humans have not been effective at producing a strong immune response in a large percentage of subjects tested in clinical trials 
. However, many of these previous H5N1 vaccine candidates were derived from clade 1 or clade 3 isolates that required multiple doses and/or the use of various adjuvants to achieve levels of antibodies believed to correlate with seroprotection in a majority of subjects tested 
. Results presented in this report indicate that our A/Indonesia/05/2005 (clade 2) VLP vaccine elicited higher HAI antibody titers than the A/Viet Nam/1203/2004 (clade 1) VLP vaccine without the use an adjuvant and elicited a robust and broadly reactive immune response following two vaccinations. Interestingly, a single immunization was able to protect mice from virus-induced death, albeit mice administered lower doses of VLPs or rHA had viral replication in the lungs and transient weight loss. These results are similar to recent live attenuated vaccines that required two vaccinations to prevent weight loss, but were able to protect ferrets following a single vaccination 
. The ability to protect humans using a “one-shot” vaccination regimen is highly desirable for a vaccine against influenza isolates with pandemic potential. Following an outbreak, vaccines that reduce viral titers in the lung and nasal mucosa may slow the transmission of the virus among humans; there may not be sufficient time for a booster shot of vaccine to achieve optimal antibody titers. The VLP vaccine described in this report demonstrates that a one-dose regimen is potentially possible in rodents using a non-replicating immunogen that can elicit cross-clade protective immune responses and is worthy of evaluation in the clinic.
One of the challenges faced by influenza vaccine developers is the ability to protect populations in the face of a spreading pandemic. The next influenza pandemic may be caused by a H5N1 virus and if so, it is not known which clade or subclade will be responsible. Correlates for protection from infection by H5N1 isolates have not been determined. Historically, the HAI assay is the most widely used serological assay for monitoring influenza immunity and is the accepted standard for measuring functional influenza-specific serum antibodies to the hemagglutinin following vaccination. An HAI titer that is greater than 1:40 (≥40) against a seasonal influenza strain is believed to be protective for ~50% of the vaccinated population 
. However, this does not appear to hold true for avian H5N1 viruses, since no correlation between HAI titer and protective efficacy against H5N1 infection has been reported in animal or human systems. Therefore, new correlates may be necessary to assess the efficacy of potential H5N1 vaccines. One interesting finding in this study was the correlation between the slower disassociation rates of the VLP-elicited antibody to HA compared to antibodies produced in response to rHA vaccines. In addition, antibodies elicited to the homologous clade 2 rHA had faster association rates compared to antibody binding the heterologous clade 1 rHA. The increase in antibody association rates in vivo
could bind HA on viruses quickly and perhaps decrease the number of infected cells in the lung, and thus could act to reduce the amount of viral replication to allow the immune system opportunity to better control the infection. In addition, antibodies that are slow to dissociate from the virion may continue to reduce the ability to uncoat and thus restrict the virus post-infection. Further analysis is needed; however, the use of antibody association/dissociation rates may be a more accurate assessment of vaccine efficacy that could potentially correlate with enhanced efficacy.
One of the more interesting findings in this study was the severity of disease induced by the 6:2 H5N1/PR8 reassortant viruses in the BALB/c mouse model. Previous publications using 1997 H5N1/PR8 clade 3 reassortant viruses generated by Subbarao and colleagues were not lethal for mice 
. Therefore, we were expecting the clade 1 and clade 2 reassortant viruses to also be non-lethal and to be able to culture viruses from lungs of infected mice to compare the efficacy of each vaccine based only upon reduction of virus titers. Instead, we found both the clade 1 and clade 2 reassortant viruses to cause precipitous weight loss and to be lethal for mice. However, lethality was only observed in mice infected under deep anesthetic, since mice infected under lighter anesthesia conditions showed less dramatic weight loss (~7%) and no mortality (data not shown). Therefore, we speculate that these reassortant viruses are lethal under conditions when the virus is allowed to infect the lower respiratory tract. Interestingly, we found that not all clade 2 H5N1/PR8 reassortant viruses were lethal to mice, since mice administered a similar high dose (106
pfu/ml) of the PR8 reassortant viruses with HA and NA proteins from A/Anhui/1/2005 (clade 2.3) did not cause mortality. In addition, none of these reassortant viruses were lethal in a ferret model (data not shown), which may be a reflection of the difference in anatomy, sites of replication, and distribution of viral receptors between BALB/c mice and ferrets. We acknowledge that wild-type H5N1 isolates may result in more pathology in mice compared to the H5N1 HA/NA reassortant viruses used in this study. Many components of avian H5N1 isolates have been attributed to the highly pathogenic nature of these viruses, including the H5N1 PB2 and NS1 proteins, which are not included in the H5N1-PR8 reassortant viruses used here. Nonetheless, reassortant viruses can be useful tools to evaluate immune responses raised to H5N1 HA and NA components in the absence of a high containment facility and special permits required to work with wild-type H5N1 viruses. We will soon have additional capacity to be able to do future immunogencity studies against highly pathogenic wild-type H5N1 isolates.
This is the first report of an H5N1 VLP vaccine derived from a clade 2 influenza isolate. These clade 2 H5N1 VLP and rHA vaccines elicited protection against a lethal challenge from an antigenically similar reassortant virus strain without including an adjuvant. However, rHA only vaccines, administered at the same dose, did not prevent morbidity and weight loss following a cross-clade challenge even though all animals did eventually recover. In contrast, a 3 µg dose of VLPs provided cross-clade protection against a H5N1 challenge with little to no observed weight loss. It is reasonable to expect that use of VLP-based vaccines will provide substantial clinical protection and reduce mortality in humans and appropriate clinical studies should be initiated to further evaluate the potential of VLP vaccines for both seasonal and pandemic influenza.