NIH Live H5N1 Avian Flu Virus Vaccines Show Protection in Animal Studies
National Institute of Allergy and Infectious Diseases (NIAID), Sept 12, 2006
When tested in mice and ferrets, experimental vaccines based on live, weakened versions of different strains of the H5N1 avian influenza virus were well-tolerated and protected the animals from a deadly infection with naturally occurring H5N1 flu viruses. The findings, which appear in the September 12 issue of PLoS Medicine, are also encouraging, the researchers say, because they demonstrate the ability to create a vaccine based on one particular strain of the H5N1 flu virus that could potentially protect against different emerging H5N1 flu strains.
Senior investigator Kanta Subbarao, M.D., M.P.H., and co-chief Brian Murphy, M.D., both of the Laboratory of Infectious Diseases at the National Institute of Allergy and Infectious Diseases (NIAID), part of the National Institutes of Health (NIH), led the research. The study was the result of a cooperative research and development agreement between NIAID and MedImmune Inc., of Gaithersburg, Md.
"This is an excellent example of the NIH and industry working together to find scientific solutions to potential public health problems," notes NIH Director Elias A. Zerhouni, M.D. "Developing a vaccine that could protect against a potential influenza pandemic is a top priority for all of us."
"If an influenza pandemic were imminent or under way, we would need a vaccine that could stimulate immunity quickly, preferably with a single dose," says NIAID Director Anthony S. Fauci., M.D. "The encouraging findings of this study suggest that vaccines based on live but weakened versions of the H5N1 avian influenza virus may quickly stimulate protective immunity. We are further exploring this live, attenuated vaccine strategy as one of several tools that we hope to have available in the event of an influenza pandemic."
As of September 8, 2006, there have been 244 confirmed human cases of H5N1 infection and more than half of those were fatal, according to the World Health Organization (WHO). Public health officials worry that the H5N1 virus will evolve to become easily transmissible among people, potentially sparking an influenza pandemic, because humans have no pre-existing immunity to the H5N1 viruses.
The NIAID and MedImmune research team created three vaccines by combining modified proteins derived from virulent H5N1 flu viruses with proteins from an artificially weakened (attenuated) flu strain. The virulent H5N1 viruses were isolated from human cases in Hong Kong in 1997 and 2003, and Vietnam in 2004. The attenuated flu vaccine strain, which also serves as the basis for MedImmunefs FluMist influenza vaccine, was lab-grown in progressively colder temperatures ("cold-adapted") to prevent the resulting vaccine viruses from spreading beyond the relatively cool upper respiratory tract. Large quantities of the resulting cold-adapted viruses were grown in chicken eggs.
The safety of the vaccine viruses was evaluated in chickens and mice. In chickens, the H5N1 vaccine viruses were not lethal, while each of the three strains of the "wild-type" (naturally occurring) H5N1 viruses were. Similarly, the vaccine viruses were not lethal in mice, but the 1997 and 2004 strains of the wild-type H5N1 viruses were. The 2003 strain of the H5N1 wild-type virus was not tested in mice because the researchers found that the virus was lethal in those animals only at very high doses. Because the wild-type H5N1 viruses have been shown to replicate in animal lungs and brains, the researchers tested the ability of the 1997 and 2004 strains of the vaccine viruses to replicate in mice and ferrets as an additional safety measure In mice, the vaccine viruses replicated in the respiratory tract but did not spread to the animalsf brains. In ferrets, the H5N1 vaccine viruses did not replicate in the lungs or the brain.
To evaluate the protective ability of the vaccines, the researchers gave the mice a single dose of vaccine virus via nose drops. All of these mice survived infection with the 1997 and 2004 H5N1 wild-type viruses, including two more recent strains of the H5N1 virus found circulating in Vietnam and Indonesia in 2005. Further, mice that received a second dose of vaccine 28 days after the initial inoculation demonstrated a stronger and more rapid immune response and almost complete protection from respiratory infection when exposed to the naturally occurring H5N1 viruses. Ferrets exhibited similar results when given two doses of the vaccine viruses.
"It is impossible to predict how the H5N1 virus will evolve or which strain, if any, will cause an influenza pandemic. To be prepared, we need to select a vaccine capable of inducing an effective human immune response against a range of H5N1 viruses that may emerge in the future. This study shows that such cross-protection can be achieved in small animals," says Dr. Subbarao. "The next step is to evaluate in people the safety and immune response induced by these vaccines to see if they produce cross-reactive antibodies that are likely to protect against different H5N1 viruses."
In June 2006, NIAID and MedImmune launched a Phase 1 study to evaluate the safety and immunogenicity of a live, attenuated H5N1 vaccine based on the 2004 H5N1 virus strain. The study, which is being performed in an isolation unit at Johns Hopkins Bloomberg School of Public Health Center for Immunization Research in Baltimore, is evaluating the safety and immunogenicity of the vaccine in approximately 20 healthy individuals between the ages of 18 and 49. Results from that study are not yet available.
The concept of using cold-adapted flu viruses to create flu vaccines, as detailed in the study in PLoS Medicine, was developed by scientists at NIAID and the University of Michigan School of Public Health (http://www3.niaid.nih.gov/news/focuson/flu/research/prevention/flumist.htm).
NIAID is a component of the National Institutes of Health. NIAID supports basic and applied research to prevent, diagnose and treat infectious diseases such as HIV/AIDS and other sexually transmitted infections, influenza, tuberculosis, malaria and illness from potential agents of bioterrorism. NIAID also supports research on basic immunology, transplantation and immune-related disorders, including autoimmune diseases, asthma and allergies.
The National Institutes of Health (NIH) - The Nation's Medical Research Agency - includes 27 Institutes and Centers and is a component of the U.S. Department of Health and Human Services. It is the primary federal agency for conducting and supporting basic, clinical and translational medical research, and it investigates the causes, treatments, and cures for both common and rare diseases. For more information about NIH and its programs, visit www.nih.gov.
Reference: A Suguitan Jr. et al. Live, attenuated influenza A H5N1 candidate vaccines provide broad cross-protection in mice and ferrets. PLoS Medicine. DOI: 10.1371/journal/pmed.0030360 (2006).
Live, Attenuated Influenza A H5N1 Candidate Vaccines Provide Broad Cross-Protection in Mice and Ferrets
Volume 3 | Issue 9 | SEPTEMBER 2006
Amorsolo L. Suguitan Jr.1, Josephine McAuliffe1, Kimberly L. Mills1, Hong Jin2, Greg Duke2, Bin Lu2, Catherine J. Luke1, Brian Murphy1, David E. Swayne3, George Kemble2, Kanta Subbarao1*
1 Laboratory of Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland, United States of America, 2 MedImmune Vaccines, Mountain View, California, United States of America, 3 Southeast Poultry Research Laboratory, Agricultural Research Service, United States Department of Agriculture, Athens, Georgia, United States of America
Recent outbreaks of highly pathogenic influenza A H5N1 viruses in humans and avian species that began in Asia and have spread to other continents underscore an urgent need to develop vaccines that would protect the human population in the event of a pandemic.
Methods and Findings
Live, attenuated candidate vaccines possessing genes encoding a modified H5 hemagglutinin (HA) and a wild-type (wt) N1 neuraminidase from influenza A H5N1 viruses isolated in Hong Kong and Vietnam in 1997, 2003, and 2004, and remaining gene segments derived from the cold-adapted (ca) influenza A vaccine donor strain, influenza A/Ann Arbor/6/60 ca (H2N2), were generated by reverse genetics. The H5N1 ca vaccine viruses required trypsin for efficient growth in vitro, as predicted by the modification engineered in the gene encoding the HA, and possessed the temperature-sensitive and attenuation phenotypes specified by the internal protein genes of the ca vaccine donor strain. More importantly, the candidate vaccines were immunogenic in mice. Four weeks after receiving a single dose of 106 50% tissue culture infectious doses of intranasally administered vaccines, mice were fully protected from lethality following challenge with homologous and antigenically distinct heterologous wt H5N1 viruses from different genetic sublineages (clades 1, 2, and 3) that were isolated in Asia between 1997 and 2005. Four weeks after receiving two doses of the vaccines, mice and ferrets were fully protected against pulmonary replication of homologous and heterologous wt H5N1 viruses.
The promising findings in these preclinical studies of safety, immunogenicity, and efficacy of the H5N1 ca vaccines against antigenically diverse H5N1 vaccines provide support for their careful evaluation in Phase 1 clinical trials in humans.
Funding: This research was supported in part by the Intramural Research Program of the NIH, NIAID. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
Academic Editor: Joseph S. M. Peiris, The University of Hong Kong, China
Received: April 25, 2006; Accepted: August 1, 2006; Published: September 12, 2006
This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.
Abbreviations: AA, influenza A/Ann Arbor/6/60; att, attenuation; BSL, biosafety level; ca, cold-adapted; CEF, chick embryo fibroblast; CPE, cytopathic effect; EID50, 50% egg infectious dose; HA, hemagglutinin; HAI, hemagglutination inhibition; HPAI, highly pathogenic avian influenza; LD50, 50% lethal dose; MDCK, Madin-Darby canine kidney; NA, neuraminidase; p.i., postinfection; pfu, plaque-forming unit(s); SE, standard error; SPF, specific pathogen-free; TCID50, 50% tissue culture infectious dose; ts, temperature sensitive; wt, wild-type
Reverse genetics technology [23,40,41], used to generate influenza viruses from cells cotransfected with plasmids expressing influenza virus gene segments, has been applied to vaccine development since 1998 [11,12,15]. The technology is especially valuable for the generation of vaccines against HPAI viruses because it permits modification of gene(s) to remove specific virulence motifs such as the highly cleavable multibasic amino acid motif in the HA protein. The generation of a candidate vaccine for human use requires transfection of cells that are qualified for use in generation of vaccines for use in humans; subsequent amplification and biological cloning are carried out in SPF embryonated hen's eggs. The H5N1 viruses that were targeted for vaccine development were selected in consultation with public health authorities and represent viruses isolated in 1997, 2003, and 2004 from human cases of H5N1 infection. These wt viruses were antigenically distinguishable from each other using postinfection ferret antisera . Live, attenuated vaccine candidates were generated against each of the three wt viruses using a common strategy. The HA gene of each of the live, attenuated vaccine candidates was modified to remove the multibasic virulence motif as described in earlier publications [11,12,15], and the six internal protein genes were derived from the highly attenuated AA ca donor virus, as described in . The H5N1 ca candidate vaccine viruses displayed the ts and att phenotypes that are specified by the internal protein genes of the AA ca [26,27], and the viruses failed to plaque in CEF cells in the absence of trypsin, consistent with the absence of the multibasic cleavage site in the HA.
The safety of the H5N1 ca vaccine viruses was established in mice and ferrets. The H5N1 1997 ca and H5N1 2004 ca viruses were not lethal for mice, and when compared with the corresponding H5N1 wt viruses, they were restricted in replication in the upper and lower respiratory tract of mice. The acquisition of the six AA ca internal protein genes by the H5N1 1997 wt and H5N1 2004 wt viruses attenuated these viruses for mice. The level of restriction of replication of the H2N2 and H5N1 wt viruses specified by the six AA ca internal protein genes was comparable, indicating that the acquisition of these genes reproducibly attenuated the H5N1 avian influenza viruses for mice. This has also been seen for an H9N2 avian influenza virus . Thus, the observation that the H5N1 2003 wt and ca viruses replicated to similar levels in mice was unexpected. Since the H5N1 2003 wt virus itself replicated to a low level and exhibited minimal lethality, it is reasonable to suggest that this H5N1 wt virus contains sequences that restrict its replication in mice. If these sequences occur in one or more of the internal protein genes of the virus, it is not surprising that substitution of these attenuating sequences with the six internal protein genes of the AA ca virus did not result in further attenuation. This suggestion will need experimental verification. The att phenotype was also seen in the ferret model: the H5N1 2004 wt and H5N1 1997 wt viruses replicated to high titers in the upper and lower respiratory tract, while the H5N1 2004 ca and H5N1 1997 ca viruses were attenuated in the nasal turbinates and were undetectable in the lungs and brain of ferrets. The similar levels of attenuation displayed by H5N1 2004 ca and H5N1 1997 ca in mice and ferrets indicate that either animal model could be used to demonstrate the attenuation of reassortant viruses bearing internal protein genes from the AA ca donor virus. Thus, the H5N1 ca viruses were safe for the respiratory tract of mice and ferrets, and additional studies indicated that the H5N1 2004 ca virus did not spread from the respiratory tract of mice or ferrets to the brain, while the H5N1 2004 wt virus readily infected the brains of these animals. The H5N1 ca viruses, in contrast to the H5N1 wt viruses, were not lethal for chickens in the standard intravenous pathogenicity test or when administered intranasally, indicating that these viruses should not pose a threat to agriculture. However, the H5N1 ca viruses would not be suitable candidate vaccines for use in poultry, because they fail to replicate and to induce an antibody response in chickens.
Despite a high level of replication in the respiratory tract of mice, the H5N1 ca candidate vaccine viruses were weakly immunogenic in mice following a single dose. Neutralizing antibody titers induced following a single dose continued to rise for 1-2 mo following immunization, indicating that the functional antibody response develops slowly in mice. A second dose of vaccine resulted in a significant boost in antibody titer and in cross-reactivity of the antibody with antigenically distinguishable H5N1 viruses. If similar antibody kinetics are observed in clinical trials, a single dose of vaccine may suffice to confer protection, but the protection may be slow to develop. A boost in antibody titers followed a second dose of vaccine despite the fact that replication of the second dose of vaccine was not detected. Thus, a delayed maturation of functional antibody occurred in response to the H5N1 ca vaccine viruses compared to that seen with other influenza A viruses [32,42]. The delayed antibody response had a correlate in efficacy studies.
A single dose of H5N1 ca vaccines provided complete protection from lethality following challenge with homologous and heterologous H5N1 wt viruses and prevented systemic spread of the wt challenge viruses to the brain. A single dose of H5N1 ca vaccines also resulted in statistically significant restriction of replication of the wt challenge viruses in the lungs that greatly exceeded that induced by infection with the heterosubtypic AA ca (H2N2) virus, thus indicating that this protection was HA- and NA-specific. However, a single dose of vaccine did not induce complete protection against replication of challenge virus in the respiratory tract. In a clinical setting, this partial protection in mice may translate to protection from severe illness and death in humans. Two doses of the H5N1 ca candidate vaccine viruses induced high titers of antibody and provided complete protection from pulmonary replication of homologous and heterologous H5N1 wt viruses. Replication of the second dose of vaccine virus could not be detected, but antigenic stimulation was sufficient to provide a boost in immune response. Two doses of the H5N1 ca vaccine in mice and in ferrets provided complete protection from pulmonary replication of homologous and heterologous H5N1 wt viruses, including A/Indonesia/05/2005 (H5N1). In mice, two doses of the H5N1 ca vaccine provided complete protection from systemic dissemination of challenge virus. Virus was detectable in the brains of ferrets following challenge with a high dose of H5N1 wt virus, but viral titers were statistically significantly reduced compared to mock-immunized animals. Similar to the observations in mice immunized with the AA ca virus, the protection in ferrets against pulmonary replication of challenge virus afforded by the H5N1 ca vaccines exceeded that observed when ferrets were immunized with an H1N1 ca virus bearing the same AA ca internal protein genes, which suggests that the protection is specific to the H5 HA.
It is not possible to predict the evolution of the H5 HA or to predict which strain, if any, will become pandemic. Therefore, an H5N1 vaccine should elicit a cross-protective immune response against a range of H5N1 viruses, including newly emerged strains. In addition, a pandemic vaccine needs to be preselected before the pandemic virus emerges in the population and characterized in humans for safety and immunogenicity, since it is highly unlikely that a vaccine can be generated, manufactured, and characterized rapidly enough to thwart spread of a pandemic strain. The assumption underlying this approach is that the preselected pandemic vaccine virus can prevent disease caused by antigenically and genetically divergent viruses belonging to the same subtypes that appear in nature. In the present study, we were able to directly evaluate this assumption, and our data on the H5N1 ca vaccines are very encouraging in this regard. A single dose of an H5N1 ca vaccine containing the 1997 H5 HA provided complete protection from lethality in mice following challenge with homologous and heterologous H5N1 wt viruses, including a clade 2 virus isolated from Indonesia in 2005. Two doses of the H5N1 ca vaccine provided complete protection from pulmonary replication in mice and ferrets and systemic dissemination of homologous and heterologous H5N1 wt viruses in mice. The high level of efficacy of the H5N1 1997 ca virus against challenge with antigenically and genetically diverse H5N1 wt viruses isolated over a span of 8 y suggests that the H5N1 viruses are evolving in nature to infect wild birds and domestic poultry, and not predominantly to evade antibodies as they do in humans. If the vaccine candidates described in this paper elicit a broadly cross-reactive protective immune response in humans, they would support the approach of developing one or two pandemic vaccine candidates for each subtype (H4 through H16) that are able to induce a broadly protective immune response to wt viruses within that subtype. The present findings serve to contradict recently expressed concerns about such a strategy [5,6].
All humans are immunologically naive with respect to the HA at the start of a pandemic, and newly emerged pandemic viruses contain at least an antigenically novel HA and may also contain a novel NA subtype. Live, attenuated influenza virus vaccines potentially offer three major advantages over inactivated vaccines for rapid immunization of an immunologically naive population . First, in naive populations, live virus vaccines induce higher levels of antibody than inactivated virus vaccines [21,44]. Second, live virus vaccines may induce protective immunity more quickly and would require fewer doses than inactivated vaccine . In immunologically naive populations, two doses of inactivated virus vaccine delivered 1 mo apart are required for induction of immunity , and significant immunity is not achieved until the response to the second dose is established, which occurs at about 40 d after immunization. Even at this time, mucosal immunity, especially that in the upper respiratory tract, is weak. In contrast, live influenza virus vaccines against human influenza viruses can induce a high level of immunity following a single dose, and immunity is achieved as early as 7-10 d following initial immunization, as indicated by the clearance of the vaccine virus from the respiratory tract. A live influenza virus vaccine induces both CD8+ T cell  and mucosal IgA antibodies in addition to serum antibodies , and therefore is the vaccine of choice for an immunologically naive population. Third, the combined humoral and mucosal immune response generated by live virus vaccines in naive populations results in broad protection against antigenically drifted strains. For example, the live, attenuated H3N2 human influenza vaccine containing an A/Wuhan/359/95 (H3N2)-like strain was effective in protecting against the A/Sydney/5/97 (H3N2)-like antigenic drift variant . This may be a particularly useful feature in the event of a pandemic, in which a vaccine generated from the emergent pandemic virus strain is not available. Whether these theoretical advantages will be seen with live, attenuated H5N1 vaccines in humans remains to be established.
In summary, the modified H5N1 ca candidate vaccine viruses were immunogenic in mice, attenuated in mice, ferrets, and chickens, and protective in mice and ferrets against subsequent challenge with homologous and heterologous wt H5N1 influenza viruses. It is not known whether investigational live virus vaccines bearing avian influenza HA and NA genes in the AA ca background will be overattenuated in humans or will be associated with some residual virulence. Live, attenuated vaccines must be able to replicate to levels that elicit a protective immune response without causing disease in the host, so a balance between attenuation and immunogenicity must be achieved. A live, attenuated H5N1 vaccine would be administered to the general population only if an influenza pandemic caused by a virus of the H5N1 subtype were imminent, with confirmation of human-to-human transmission, or was already underway in the United States, and it would be used only on the recommendation of public health authorities. However, it is important to generate and carefully evaluate candidate live, attenuated H5N1 vaccines in clinical trials in humans because of their potential advantages over other vaccine approaches.
On several occasions since 1997, outbreaks of highly pathogenic avian influenza (HPAI) H5N1 infections have occurred in poultry and in humans, fueling public health concerns over their potential to ignite a pandemic that would cause significant morbidity and loss of life. The most recent H5N1 outbreak in poultry, which began in late 2003, affected at least ten Asian countries initially, but recent reports indicate that H5N1 viruses have been isolated in wild birds and poultry in several countries in Asia, Europe, and Africa. More importantly, human cases of H5N1 infections have been reported since December 2003 in nine countries, with a total of 194 laboratory-confirmed cases and 109 fatalities as of 12 April 2006 (http://www.who.int/csr/disease/avian_influenza/country/cases_table_2006_04_12/en/). Although antiviral drugs are valuable for the prevention and treatment of H5N1 infections, they are limited in supply; in addition, resistance to these antivirals-which include drugs that target neuraminidase (NA) as well as those that block M2 ion channel function-has been described in multiple isolates [1-3]. The development of an H5N1 vaccine is recognized as the primary strategy to protect humans against a possible H5N1 pandemic. Infections with H5N1 influenza viruses in avian species and in humans have been occurring sporadically since 1997 and phylogenetic and antigenic analyses of H5N1 viruses collected over this period indicate that they have evolved into different sublineages or clades: 2004 viruses are designated clade 1, 2003 viruses clade 1_, some 2005 isolates clade 2, and the 1997 viruses clade 3 [4,5]. Antigenic differences among influenza viruses are usually identified by analysis of specific postinfection ferret antisera in reciprocal hemagglutination-inhibition (HAI) assays. Reports of genetic and antigenic changes observed in the H5N1 viruses isolated between 1997 and 2005 have raised concern about the ability of a vaccine generated from a single selected strain to protect against a diverse set of viruses [5,6].
The two modalities that are licensed as vaccines in the U.S. against human influenza viruses are inactivated split virus and live, attenuated virus vaccines. The development of effective vaccines against HPAI H5N1 viruses, however, presents several challenges. Due to their high pathogenicity, handling of the wild-type (wt) viruses requires biosafety level 3 (BSL-3) containment. Moreover, these viruses are highly pathogenic in chickens and are also lethal for chick embryos, so they cannot be propagated efficiently and safely in eggs. Several strategies have been explored in attempts to overcome these obstacles. These include the use of a low-pathogenicity H5N3 avian influenza virus that is antigenically related to circulating strains , the use of recombinant H5 hemagglutinin (HA) expressed in baculovirus , the use of recombinant adenoviruses expressing the H5 HA [9,10], and the production of attenuated seed viruses with the H5 HA modified by means of reverse genetics [11,12].
Nicholson and colleagues  investigated the feasibility of using an antigenically related, low-pathogenicity H5N3 influenza virus (influenza A/duck/Singapore/97) inactivated virus vaccine to protect against HPAI H5N1. The vaccine was safe and well tolerated but was immunogenic only when administered with the adjuvant MF59. Putative protective antibody levels were not detectable in study participants 16 mo after receiving two doses , but significant boosting of antibody responses was seen upon administration of a third dose of vaccine. Although the geometric mean titers were low in all participants, higher seroconversion rates and cross-reactive antibodies were seen among participants who received the adjuvanted vaccine . These studies demonstrated that an inactivated virus vaccine against H5N1 influenza viruses generated from antigenically related, low- pathogenicity avian viruses is weakly immunogenic but it can induce broadly reactive antibodies when administered in multiple doses with adjuvant [7,13,14].
A baculovirus-expressed recombinant H5 protein vaccine was evaluated in Phase 1 clinical studies in humans . The vaccine was well tolerated but was poorly immunogenic; only 52% of participants who received two doses of 90 _g of vaccine developed a neutralizing antibody response to the vaccine.
Two adenoviral vectors expressing full-length H5 HA were tested in mice and poultry and, after two doses, were found to be immunogenic and protective against challenge with homologous and heterologous H5N1 wt influenza viruses [9,10]. These vaccine candidates elicited both humoral and cellular immune responses, and the protection against heterologous challenge was presumed to be mediated by the cellular immune response.
Multiple basic amino acid residues adjacent to the HA1-HA2 proteolytic cleavage site of HA of HPAI H5 and H7 viruses constitute a virulence motif that makes the HA cleavable by intracellular and extracellular proteases, thereby altering virus tropism and its ability to spread systemically (reviewed in ). Several investigators have developed experimental H5N1 influenza vaccines in which the HA cleavage site sequence is modified, resulting in the generation of a low-pathogenicity seed virus that can be handled under standard BSL-2 laboratory containment used for the production of current interpandemic human influenza vaccines [11,12]. Li and colleagues  reported the generation of two live, attenuated, ca, temperature-sensitive (ts) vaccine viruses based on two 1997 human H5N1 influenza isolates. The vaccine viruses were immunogenic in ferrets and retained the antigenicity of the wt parental viruses, demonstrating that production of live, attenuated recombinant H5 viruses containing a modified HA and the A/Ann Arbor/6/60 ca (AA ca) internal protein genes is a useful approach to the generation of vaccines against HPAI viruses. However, these vaccines were not tested in mice or humans. Subbarao and colleagues  reported the use of the same general approach to produce an inactivated experimental H5N1 influenza vaccine based on a 1997 H5N1 virus. The seed virus was generated by reverse genetics in 293T cells, a substrate that is not acceptable for use in human vaccines. Subsequently, Webby and colleagues  generated a recombinant H5N1 influenza virus in Vero cells, in which the modified HA and NA genes were derived from A/Hong Kong/213/2003 (H5N1) and the internal protein genes were derived from A/Puerto Rico/8/34 (H1N1). This vaccine was immunogenic in mice when administered with incomplete Freund's adjuvant and provided a dose-dependent protection against heterologous wt H5N1 viruses . A similarly generated inactivated subvirion vaccine, based on A/Vietnam/1203/2004 (H5N1) has been evaluated in Phase 1 studies in healthy adult volunteers . Although a correlation was observed between the dose of vaccine and immunogenicity, the vaccine was poorly immunogenic and only 54% to 58% of individuals immunized with two 90 _g doses of the vaccine developed antibody responses that had been predicted to be protective in two assays. The vaccine is now being evaluated in children and in the elderly. Evaluation of strategies to improve immunogenicity of the inactivated subvirion vaccine with adjuvants and dose-sparing regimens such as intradermal delivery, are ongoing. Thus, to date, the immunogenicity of inactivated virus vaccines that were evaluated in clinical trials is suboptimal and require at least two doses and/or the use of adjuvants.
We have been developing live, attenuated influenza virus vaccines because they have properties that make them attractive vaccines for the prevention of pandemic influenza in humans, including (1) stimulation of an immune response in influenza A virus-naive individuals following a single dose of vaccine; (2) safety for fully susceptible seronegative individuals; (3) induction of cross-reactive immune responses; (4) poor transmissibility; and (5) genetic and phenotypic stability [18,19,20,21]. In the present study, three candidate live, attenuated H5N1 vaccine viruses were generated following rescue in Vero cells. Each of these vaccines, produced in eggs, contains a different H5 HA (from 1997, 2003, and 2004 H5N1 viruses) with the multibasic motif deleted, an accompanying avian N1 NA, and the set of six internal protein genes of the master donor virus AA ca. The three H5N1 ca viruses bear the phenotypes predicted by the modification engineered in the HA gene (trypsin dependence) and those specified by the internal protein genes of the AA ca virus (ts and attenuation [att]). The candidate H5N1 ca vaccine viruses were evaluated for immunogenicity in mice and for efficacy in mice and ferrets. The breadth of immunity that could be induced by immunization with live, attenuated H5N1 vaccines was determined in cross-protection studies using genetically and antigenically diverse H5N1 viruses isolated over an eight-year period representing different clades for challenge.
Generation of H5N1 ca Viruses
Reassortant viruses with a modified H5 HA gene and N1 NA gene derived from H5N1 1997, 2003, and 2004 viruses and remaining gene segments from AA ca were generated by reverse genetics . The HA gene was modified by replacing the multibasic amino acid motif with the sequence seen in avian influenza viruses that are not highly pathogenic in chickens (Table 1) . The nucleotide sequence of each gene segment of the three viruses was identical to the sequence of the corresponding gene insert in the parent viruses, with the exception of a single coding change each in the PA gene segment of the 2004 ca and 1997 ca viruses. The five mutations in the AA ca virus that specify ts and att phenotypes were present in the H5N1 ca viruses [26,27]. The H5N1 ca vaccine viruses were antigenically similar to their respective wt parent viruses by HAI test with postinfection ferret antisera (unpublished data). The phenotypic characteristics of the H5N1 ca viruses were also evaluated in vitro and in vivo in chickens, mice, and ferrets.
In Vitro Phenotypes: Trypsin Dependence and Temperature Sensitivity
The HA gene in the H5N1 ca viruses differed from the HA gene in the corresponding wt parent viruses by the absence of the multibasic amino acid cleavage site. The presence of this motif in an influenza A virus of the H5 or H7 subtype confers the ability to plaque efficiently in CEF cells in the absence of trypsin [28-30]; the H5N1 1997 wt, H5N1 2003 wt, and H5N1 2004 wt viruses demonstrated this property (Table 1). The H5N1 ca viruses failed to plaque in CEF cells in the absence of trypsin, consistent with the absence of the multibasic cleavage site motif in the HA protein (Table 1) .
Extensive experience with H1N1 and H3N2 reassortant viruses and an H9N2 reassortant virus that derived the six internal protein gene segments from the AA ca virus has shown that the ts and att phenotypes are specified by these gene segments [19,26,27,31,32]. The H5N1 ca viruses and the AA ca parent virus were restricted in replication at 39 C (ts); in contrast, the wt H5N1 parent viruses replicated equally well at 33 C and 39 C (Table 2). Thus, the three H5N1 ca viruses demonstrated the ts phenotype.
Attenuation In Vivo in Chickens, Mice, and Ferrets
Level of attenuation in chickens following intravenous administration. The H5N1 1997 wt, H5N1 2003 wt, and H5N1 2004 wt viruses were highly pathogenic for chickens, findings that are consistent with the presence of multibasic amino acid motif in the HA [28-30] (Table S1). The H5N1 ca viruses were no longer lethal for chickens in the standard intravenous pathogenicity test, consistent with the absence of the multibasic cleavage site motif in the HA protein (Table S1) .
Level of attenuation in chickens following intranasal administration. In order to determine whether H5N1 ca viruses would infect and cause disease in chickens exposed to the virus through a more natural route, the H5N1 wt and ca viruses were administered to chickens intranasally. The H5N1 wt viruses were recovered from cloacal and oropharyngeal swabs on day 3 following virus administration, and the infection was lethal for 8 of 8 chickens (Table S1). The H5N1 ca viruses were not lethal for chickens and did not appear to infect the chickens following intranasal administration, as indicated by the absence of detectable infectious virus in cloacal or oropharyngeal swabs and the absence of an antibody response to the virus (Table S1). The poor infectivity of the H5N1 ca viruses in chickens is likely the result of an inability of these ts viruses to replicate at the high body temperature of chickens (40-41 C) and suggests that their use would not pose a threat to the poultry industry.
Lethality in mice. When administered intranasally, the H5N1 1997 wt and H5N1 2004 wt viruses were highly lethal for mice (LD50 values of 102 and 100.4 TCID50, respectively), whereas the H5N1 2003 wt virus was lethal for a few mice at a very high dose (LD50 ≥ 106 TCID50). Lethality for mice was completely abrogated in the three H5N1 ca viruses; none of the mice died even at the highest dose of virus administered (LD50 > 107 TCID50).
Level of replication in the respiratory tract of mice. The level of replication of the H5N1 1997 ca and H5N1 2004 ca viruses was lower than that of the corresponding H5N1 wt viruses in the upper and lower respiratory tract of mice (Table 3), and the differences in titer were similar to or exceeded the difference between the titers achieved in mice infected with the AA wt and ca viruses. The level of replication of the H5N1 2003 wt and ca virus did not differ significantly in the respiratory tract (Table 3 and unpublished data for days 2 and 4). The reasons for this observation are not clear, but it may be the result of a partial attenuation of the H5N1 2003 wt virus, which does not replicate as efficiently in the respiratory tract of mice as the H5N1 1997 wt and H5N1 2004 wt viruses, and may also partly explain why the H5N1 2003 wt virus was not lethal for mice in some studies , although other investigators have found the H5N1 2003 wt virus to be lethal in mice .
Level of replication of H5N1 2004 ca virus in the brains of mice following intranasal administration. Previous studies have established that some of the H5N1 wt viruses can be detected in the brains of mice following intranasal administration [33-38]. In order to determine whether the modifications in the HA gene and the presence of the internal protein gene segments from the AA ca virus altered the neurotropism of the H5N1 2004 ca reassortant viruses, we administered the H5N1 2004 ca virus, the H5N1 2004 wt virus, and an H1N1 ca virus (as a control) to mice by the intranasal route and determined the ability of these viruses to replicate in the respiratory tract or brain. A single dose of 106 TCID50 of each virus was administered intranasally, and eight mice from each group were sacrificed on alternate days from day 2 to day 12 p.i. (Figure 1). The H5N1 2004 wt virus was detected in the brains of mice on days 2 and 4 p.i. with mean titers of 101.8 and 104.9 TCID50/g, respectively, but the H5N1 2004 ca virus and the ca virus were not. It should be noted that although the H5N1 2004 ca virus and the H1N1 ca virus were not recovered from the brain, they were isolated from the lungs and nasal turbinates (unpublished data), indicating that the H5N1 ca and H1N1 ca viruses replicated in the respiratory tract, but did not spread to the central nervous system of mice. Further studies are warranted to delineate the relative contribution(s) of the modified HA and the internal genes of AA ca to the observed lack of neurotropism of the H5N1 2004 ca virus. A similar lack of neurotropism of the H5N1 1997 ca virus was also observed (unpublished data).
Level of replication of H5N1 2004 ca and H5N1 1997 ca in ferrets. The levels of replication of the H5N1 2004 ca and H5N1 1997 ca viruses compared to their respective wt parental H5N1 viruses in ferrets were determined. The H5N1 2004 ca and H5N1 2004 wt viruses replicated in the upper respiratory tract of ferrets but the H5N1 2004 ca virus was not detected in the lower respiratory tract or brain of any of the ferrets. In contrast, the H5N1 2004 wt virus replicated in the lungs and brains of ferrets (Table 4). The H5N1 1997 wt virus replicated efficiently in the upper respiratory tract of ferrets, and the H5N1 1997 ca virus was detected in the nasal turbinates of two out of three ferrets. The H5N1 1997 ca virus was not detected in the lower respiratory tract or brain of ferrets, unlike the H5N1 1997 wt virus, that was isolated from the lungs and brains of all three ferrets (Table 4). The H5N1 1997 ca and H5N1 2004 ca viruses were highly restricted in replication in the upper respiratory tract of ferrets compared to the corresponding wt viruses. The attenuation of the H5N1 2004 ca and H5N1 1997 ca vaccine viruses in the respiratory tract of ferrets was consistent with the observations in mice.
Thus, the H5N1 ca viruses demonstrated the properties predicted by the genetic modifications of removing the multibasic amino acid motif in the HA and the introduction of the internal gene segments from the AA ca virus. Studies were next undertaken to determine the immunogenicity and efficacy of these viruses as candidate vaccines.
Immunogenicity of the H5N1 ca Virus Vaccine Candidates
Each of the H5N1 ca candidate vaccines was administered intranasally to mice at a dose of 106 TCID50, and serum antibody responses at 28 d postimmunization were measured by HAI and microneutralization assays. The nonlethal H5N1 2003 wt virus was administered to a group of mice as a positive control. A single dose of vaccine was poorly immunogenic in mice, with low antibody responses in both assays (Table 5). However, ELISA antibodies were detected in the serum against recombinant H5 HA (unpublished data). The titer and cross-reactivity of the neutralizing and HAI antibody response was greater 4 wk after a second dose of vaccine administered 28 d after the first dose (Table 6), although replication of the second dose of vaccine was not detected in the respiratory tract (unpublished data). The H5N1 2003 ca and wt viruses elicited higher antibody titers against the homologous virus than the H5N1 1997 ca and H5N1 2004 ca viruses; these findings are consistent with previous observations that identified the amino acid asparagine at position 223 in the HA as a determinant of enhanced immunogenicity of the H5N1 2003 viruses .
An additional study was performed to determine the immunogenicity of one versus two intranasal doses of vaccine. In this study, one group of eight mice received a dose of vaccine on day 0, a second group received vaccine on days 0 and 28, and a third group was mock-immunized on day 0 and received vaccine on day 28. Serum was collected from all of the mice on days 27 and 54. The HAI antibody response 4 wk after a single dose of the H5N1 2004 ca vaccine was low. Responses to a single dose of vaccine continued to rise over 2 mo, however, and titers achieved 4 wk after a second dose of vaccine were higher than in mice that received a single dose of vaccine 8 wk earlier (Table 7). The pattern of HAI responses was the same whether the assay was performed with turkey or horse erythrocytes, but the HAI assay was more sensitive when performed with horse erythrocytes (Table 7).
Efficacy of the H5N1 ca Virus Vaccine Candidates
Protection from lethality in mice. All of the mice immunized with H5N1 ca vaccine viruses survived challenge with 50, 500, or 5,000 LD50 of homologous or heterologous H5N1 wt viruses, despite an undetectable neutralizing and HAI antibody response in mice 28 d following a single intranasal dose of 106 TCID50 of the vaccines (Figure 2). Mock-immunized mice died between days 6 and 10 p.i. The AA ca virus that was the source of the six internal protein genes of the H5N1 ca vaccine viruses provided partial heterosubtypic protection from lethality following challenge with the H5N1 1997 and H5N1 2004 wt viruses. This protection decreased with increasing doses of challenge virus (Figure 2). The mechanism of protection from lethality that was observed in the absence of detectable serum neutralizing antibodies against H5N1 viruses is under investigation.
Protection from challenge virus replication in the respiratory tract of mice. Although mice that received a single dose of vaccine were protected from lethality, they were not fully protected from replication of the H5N1 1997 wt or H5N1 2004 wt virus in the respiratory tract (Figure 3A and 3B). The titer of the challenge virus detected in the respiratory tract of mice immunized with the H5N1 2004 ca vaccine was significantly lower than in mock-immunized mice or mice immunized with the AA ca virus (Figure 3A and 3B), and the challenge virus did not spread to the brain (unpublished data). However, complete protection from pulmonary virus replication was not seen following a single dose of vaccine when mice were challenged with 106 TCID50 or even 102 TCID50 of H5N1 1997 wt or H5N1 2004 wt virus (unpublished data). Incomplete protection from pulmonary virus replication is consistent with the low neutralizing and HAI antibody response 28 d following a single dose of vaccine (Table 5). Interestingly, a single dose of the H5N1 2004 ca virus provided complete protection from replication of the H5N1 2003 wt virus in the respiratory tract (Figure 3A and 3B). In consonance with a higher level of neutralizing and HAI antibody response following two doses of vaccine (Tables 6 and 7), complete or near-complete protection from virus replication was observed in the respiratory tract of mice following two doses of vaccine (Figure 3C and 3D).
Efficacy of the H5N1 ca Virus Vaccine Candidates against Antigenic Variant wt H5N1 Viruses Isolated in 2005
Efficacy in mice. The H5N1 ca vaccines were also evaluated for their efficacy against challenge with newly emerged and antigenically distinct H5N1 strains, A/VN/JPHN30321/2005 (H5N1), designated a clade 1 virus, and A/Indonesia/5/2005 (H5N1), a clade 2 virus by the CDC . A single intranasal dose of 105 TCID50 of the H5N1 2004 ca vaccine provided complete protection from lethality following challenge (105 TCID50) with A/VN/JPHN30321/2005 (H5N1) and A/Indonesia/5/2005 (H5N1) (unpublished data). Moreover, mice that received two doses (106 TCID50/dose) of the H5N1 ca vaccines were protected from viral replication in the lungs and brains following challenge with 105 TCID50 of A/Indonesia/5/2005 (H5N1) (Figure 4). This protection from viral replication was associated with high neutralizing antibody titers against A/Indonesia/5/2005 (H5N1) that were detected in the sera of all immunized mice that received two doses of the vaccines (geometric mean neutralizing antibody titers of 90, 174, and 104 for mice that were immunized with two doses of the H5N1 1997 ca, H5N1 2003 ca, and H5N1 2004 ca, respectively).
Efficacy in ferrets. Ferrets that received two doses (107 TCID50/dose) of the H5N1 ca vaccines were completely protected from viral replication in the lungs following homologous or heterologous challenge; the latter included challenge with 107 TCID50 of A/Indonesia/5/2005 (H5N1) (Figure 5). Virus titers in the nasal turbinates of ferrets immunized with H5N1 2004 ca or H5N1 1997 ca were statistically significantly lower than those of the mock-immunized ferrets following challenge with H5N1 2004 wt or A/Indonesia/05/2005 (H5N1). Virus was detected in the brains of these ferrets at statistically significantly lower titers than in mock-immunized animals. A possible explanation for the presence of virus in the brains of these ferrets is that challenge inoculum was present in the olfactory bulb. Since different areas of the brain were not processed separately, this cannot be confirmed. There was a strong correlation between the titers of virus in these tissues and the immune response at day 18 after the second immunization (p < 0.000001 and p = 0.0035, respectively).
Thus, immunization with two doses of an H5N1 ca vaccine possessing the surface glycoprotein genes of the 1997 virus provided a high level of protection in mice and ferrets against H5N1 viruses that had evolved in nature over 8 y and that belonged to different phylogenetic clades.