As a result of the magnitude and severity of recent and ongoing outbreaks of HPAI (H5N1), interest in the development of high-quality vaccines against avian influenza has been heightened. Control programs designed to prevent, manage, or eradicate HPAI which have been developed by OIE/FAO in collaboration with WHO now include the use of vaccines against avian influenza (31
). Such a vaccine should ideally meet three criteria: (i) it must be safe in both host and the environment; (ii) its use must be economically feasible; and (iii) it must be efficacious. Here, we have used reverse genetics to generate a panel of eight candidate live attenuated vaccine viruses. The viruses were attenuated in mammalian substrates and grew to high titers in embryonated chicken eggs. Furthermore, all eight vaccine candidates proved effective in protecting mice and one candidate vaccine which was chosen to be tested in chickens was also found to protect this species against challenge with highly pathogenic avian influenza (H5N1) viruses.
The high pathogenicities of currently circulating H5N1 strains present two significant obstacles to vaccine production using conventional methods: (i) high virulence prevents growth of the viruses to maximum titer in 10-day-old embryonated eggs, the industry standard substrate for the production of influenza vaccines, and (ii) high containment facilities are required in order to protect workers and the environment from possible exposure to the virus. Both of these issues have been addressed herein through the design of reverse genetics-based vaccines. The cleavability of the HA protein of avian influenza viruses is a critical determinant of virulence (11
); low-pathogenicity influenza A strains of subtypes H5 and H7 have each been reported to gain a high-pathogenicity phenotype in birds following the acquisition of a polybasic cleavage site in the HA protein (12
). In order to generate H5 viruses with sufficiently attenuated virulence to allow growth to high titer in 10-day-old embryonated chicken eggs, viruses were reverse engineered to remove the nucleotides encoding the polybasic cleavage site in the HA. Each of the recombinant viruses used in the study grew to maximal titer between 24 and 48 h, a time shorter than the time to death of the embryonated eggs used for growth, demonstrating that 10-day-old eggs can be conveniently used for the production of the vaccine. Furthermore, the use of cDNA to generate the viruses de novo allowed the codon usage at the HA cleavage site to be altered in a way that should reduce the risk of a multibasic cleavage site being reintroduced. Specifically, the number of adenosine nucleotides present at the cleavage site was minimized (Fig. ). We reasoned that by removing adenosines we would lower the possibility of a reintroduction of the polybasic cleavage site by limiting the opportunity for polymerase stuttering to occur. Our strategy has proven successful: sequencing of each of the eight recombinant viruses used in this study, following multiple rounds of growth in embryonated eggs (at least seven passages for each virus), has shown no alteration to the engineered sequence introduced into the HA segment. Furthermore, sequencing of three independent clones of the NS 1-73 vaccine virus isolated from vaccinated chickens showed that, in each case, the PB2 627E polymorphism was retained, there were no alterations from the low-pathogenicity cleavage site motif of the HA gene and no reversion from the truncated form of the NS1 gene. In one clonal isolate, there was an amino acid change in the NEP protein (I32T); the significance of this mutation, if any, is currently unknown. Thus, we have shown that the candidate vaccine viruses are stably attenuated in an avian substrate and that the stability of a low-pathogenicity-type cleavage site can be increased through rational engineering of the genome. In addition to attenuating virulence in eggs, we showed that removal of the polybasic cleavage site leads to reduced pathogenicity in a mammalian model. The removal of the polybasic cleavage site from A/Viet Nam/1203/04 virus results in an increase in the mouse LD50
value by approximately 10,000-fold (data not shown). A similarly high level of attenuation in a mouse model was observed by Hatta et al. using a strain of H5N1 from the original outbreak of highly pathogenic avian influenza in Hong Kong in 1997 (9
). While the potential for human infection with live attenuated poultry vaccines remains a formal possibility, our data suggest that the stable removal of the polybasic cleavage site from the HA of recombinant viruses used in this study greatly reduces the risk they would pose for vaccine production workers or the environment.
In addition to growth to high titers and safety, a basic requirement of a vaccine is adequate protective efficacy. In order to achieve attenuation in virulence while retaining sufficient immunogenicity to induce a protective response, we attempted to cover a broad range of attenuation phenotypes by creating multiple vaccine candidates. We introduced attenuating mutations (in addition to removal of the polybasic cleavage site) into a panel of viruses to produce genotypes which combined (i) incremental truncation of the NS1 protein and (ii) either lysine or glutamic acid at amino acid position 627 of the PB2 protein. The truncations of the NS1 protein were expected to attenuate the recombinant viruses in any host species, while the presence of a glutamic acid in the PB2 protein would attenuate viruses in mammalian hosts only. We hypothesized that a virus could be selected from our panel with the correct balance of attenuation and immunogenicity to provide complete protection from disease in birds, while not inducing pathogenicity on vaccination.
Whether in the background of a virus encoding K or E at PB2 627, truncations of any length in the NS1 open reading frame achieved similar levels of attenuation relative to the corresponding virus encoding a full-length NS1. All viruses encoding NS1 truncations were shown to exhibit attenuated growth in vitro and in ovo and, as expected, the degree of attenuation correlated with increased induction of IFN observed in infected cell cultures. In the mouse model, we saw that vaccination with each of the viruses encoding an NS1 truncation led to protection.
As has been shown before in highly pathogenic influenza virus backgrounds, the presence of E at PB2 627 was found to decrease pathogenicity of the candidate vaccine viruses in a mouse model (9
). Attenuating effects of PB2 627E on viral growth would be expected in a mammalian host due to the decreased polymerase activity and viral growth observed at low temperature (33°C) for viruses encoding 627E (15
). Indeed, Hatta et al. have reported that viral growth in the upper respiratory tract of mice is restricted by the presence of E at PB2 627 (10
). A recently proposed mechanism to explain this difference in activity involves the presence of an inhibitory factor in mammalian cells which reduces the activity of influenza virus polymerase complexes encoding a PB2 627E polymorphism (16
). Toward the aim of increased safety for humans, we introduced a glutamic acid at PB2 position 627. Although this site is subject to rapid mutation upon introduction of avian influenza viruses into mammalian hosts, the initial presence of 627E is desirable and, furthermore, the attenuated replication of the NS1-truncated viruses reduces the probability of reversion.
All eight of our candidate vaccine viruses were protective against lethal challenge in mice. Each virus with a truncated NS1 open reading frame and E or K at PB2 627 was also highly attenuated in mice. Thus, viruses with a desirable balance between attenuation and immunogenicity were obtained. Based on its genotype and its attenuated phenotype in mice, the VN HALo/627E/NS 1-99 virus was chosen to be tested in chickens.
The NS1 1-99 virus also proved to be attenuated in the chicken model. Two-week-old chickens vaccinated with 106 PFU of virus exhibited no mortality or morbidity. Crucially, chickens challenged with a homologous clade 1 A/Viet Nam/1203/04 virus subsequent to the single vaccination dose were entirely protected from signs of disease and mortality. We further tested the efficacy of the vaccine by challenging immunized animals with a heterologous clade 2.2 H5N1 virus, A/egret/Egypt/01/06, the HA protein of which possesses approximately 7% amino acid divergence with VN1203 in the HA1 portion. Even after this heterologous challenge, greater than 80% of the vaccinated animals were protected from signs of disease and death. In spite of the increase in antibody titers after the heterologous challenge, none of the vaccinated animals shed detectable amounts of virus, emphasizing the efficacy of the VN HALo/627E/NS 1-99 vaccine in poultry.
Importantly for the feasibility of poultry vaccination, our results show that a single dose of live attenuated vaccine is sufficient to produce a protective immune response in both mouse and chicken models. This observation is consistent with previous results using alternative live attenuated influenza vaccines in chickens (30
). In contrast, clinical trials using inactivated H5N1 influenza vaccines have shown that, for the production of a significant immune response, two doses containing large quantities of antigen are needed (4
), and furthermore, the use of adjuvants may be required to achieve optimal results (2
). Using the live attenuated NS1-truncated H5N1 vaccines, mice and chickens were both shown to be completely protected from disease after a lethal challenge of 1,000 MLD50
or 100 CLD50
4 weeks after administration of a single vaccination dose. One consequence of the application of the NS1 truncation strategy to the generation of chicken vaccines is that interferon produced in the substrate (chicken's eggs) will be active in the vaccinated animals. The impact (positive or negative) of interferon present in vaccine preparations has not been evaluated in the present work. Nevertheless, our results suggest that live attenuated vaccines based on truncations of the NS1 protein are an economical and efficient way to achieve immunization.
One current limitation in the use of conventional vaccines against AIV is that vaccinated poultry frequently cannot be differentiated from naturally infected birds based on serology, making surveillance difficult to perform. However, it has been demonstrated that differentiation between vaccinated/exposed animals and vaccinated/unexposed animals is possible, using a strategy known as “differentiating infected from vaccinated animals” (DIVA) (2a
). The DIVA strategy requires a vaccine which lacks one or more antigens present in the circulating avian influenza strain. Thus, in principle, a live attenuated virus lacking the C-terminal domain of NS1 could meet the DIVA criterion. Animals vaccinated with an NS1-truncated virus would not produce antibodies against the C terminus of NS1.
In response to the need for effective H5 subtype vaccines in poultry, we have generated and characterized a panel of recombinant viruses encoding attenuating mutations in the HA, PB2, and NS1 proteins. Vaccination of chickens achieved homologous and heterologous protection from challenge with highly pathogenic H5N1 influenza viruses. Safety and protective efficacy in the mouse model furthermore suggested that these vaccines may be suitable for use in mammalian hosts.