Broadly neutralizing DNA vaccine with specific mutation alters the antigenicity and sugar-binding activities of influenza hemagglutinin

Abstract

The rapid genetic drift of influenza virus hemagglutinin is an obstacle to vaccine efficacy. Previously, we found that the consensus hemagglutinin DNA vaccine (pCHA5) can only elicit moderate neutralization activities toward the H5N1 clade 2.1 and clade 2.3 viruses. Two approaches were thus taken to improve the protection broadness of CHA5. The first one was to include certain surface amino acids that are characteristic of clade 2.3 viruses to improve the protection profiles. When we immunized mice with CHA5 harboring individual mutations, the antibodies elicited by CHA5 containing P157S elicited higher neutralizing activity against the clade 2.3 viruses. Likewise, the viruses pseudotyped with hemagglutinin containing 157S became more susceptible to neutralization. The second approach was to update the consensus sequence with more recent H5N1 strains, generating a second-generation DNA vaccine pCHA5II. We showed that pCHA5II was able to elicit higher cross-neutralization activities against all H5N1 viruses. Comparison of the neutralization profiles of CHA5 and CHA5II, and the animal challenge studies, revealed that CHA5II induced the broadest protection profile. We concluded that CHA5II combined with electroporation delivery is a promising strategy to induce antibodies with broad cross-reactivities against divergent H5N1 influenza viruses.

DNA vaccines have been considered as an appealing option against pandemic diseases, such as influenza. We have reported previously that a hemagglutinin (HA)-based DNA vaccine can generate protective immunity against viral challenges in a preclinical model of influenza (1). Many groups have made efforts to develop vaccines and immunotherapeutics based on plasmid DNA (2, 3). For influenza viruses, DNA vaccines based on conserved antigens from viral internal proteins offer the potential for eliciting cell-mediated immune responses, which is believed to provide cross-protection and clearance abilities (4, 5).

Influenza A virus remains a threat to humans even though the vaccine against a specific strain provides a useful prophylactic protection. The virus envelope protein HA is the most abundant protein and the main player in inducing immune response (6). Therefore, current efforts in the design of flu vaccine often involve the use of either HA proteins or recombinant virus carrying a specific HA. Although the vaccine protection against the homologous strain is satisfactory, cross-protection against heterologous strains is limited; i.e., the vaccine often has low immunity against different influenza virus strains. Influenza viruses continuously evolve by increasing the mutations in epitopes (antigenic drift) or by reconstituting the genome with other strains (antigenic shift) (7). As a result, influenza vaccines need to be updated annually depending on the circulating strains or upon the emergence of a pandemic. Understanding the contribution of amino acids on HA to antigenicity would facilitate the design of a universal vaccine that confers broad cross-reactivity to various influenza virus strains.

The antigenicity of HA is mainly localized in the globular region that ranges from amino acid 91–261 (H3 numbering) (8). A low frequency of virus variant that displays multiple point mutations in the globular region has been found to succeed in escaping vaccinated immunity (9). Furthermore, the globular region of HA is also the main region recognized by neutralizing antibodies (10). The mutation(s) at antigenic sites may decrease the antibody-neutralizing ability, which blocks interaction between HA and sialosides on the host cell membrane (11, 12).

The HA protein is not only involved in antigenicity, but also responsible for receptor binding for viral transmission (10). Human influenza strains mainly recognize α2,6 sialosides, whereas the avian strains exhibit high-affinity binding to α2,3 sialosides (13). Several studies showed that the receptor-binding specificity and the avidity of the HA protein can be altered by specific amino acid substitutions in the globular region. For example, for H3 influenza virus, two amino acid substitutions Q226L and G228S can alter the binding specificity from α2,3 sialosides to α2,6 sialosides (14, 15). A similar trend was observed in H1 influenza virus; E190D and D225G substitutions shift receptor-binding specificity of the 1918 pandemic strain (16, 17). The antigenicity and receptor-binding activities localize in the same globular region of HA. Whether these two characteristics are correlated remains to be elucidated. Recent reports suggested that specific HA amino acid mutations in the receptor-binding domain (RBD) might be related to the sensitivity of the serological assay (18). These findings suggest that the influenza virus tends to induce a small number of amino acid mutations to increase receptor-binding avidity and reduce the antigenicity of mutated HA (19), consistent with the studies on recent human pandemic strains (12, 20), that the amino acid mutations at RBD could affect the receptor-binding avidity and/or specificity of HA, which leads to changes in the HA immunogenicity and antigenicity and finally the appearance of the immune-escaping strain. Monitoring receptor-binding avidity of circulating viruses may facilitate the accurate prediction of the mutants of epidemic potential, and further provide references for vaccine strain decision.

We previously reported that a consensus DNA vaccine showed full protection against most of H5N1 influenza viruses, yet moderate protection against certain currently circulating clade 2 viruses (1). In this report, we modified CHA5 by incorporating the critical residues from clade 2 viruses or by including more recently circulating viruses to generate the second-generation consensus vaccine CHA5II. The broadness of neutralization profiles of the induced antiserum was evaluated and compared to the antiserum induced by the HA from representative H5N1 strains. The efficacy of the selected candidate was further confirmed with animal challenge studies.

Results

Design of Hemagglutinin Vaccine with Improved Broadness of Protection.

It was previously reported that the antiserum induced by CHA5 exhibited moderate neutralization activity against A/Indonesia/2005 (ID05, clade 2.1), A/Anhui/1/2005 (AN, clade 2.3), and A/duck/Fujian/1734/05 (FJ, clade 2.3). The design was further improved using two approaches in order to elicit broader protection activities.

The first approach was to modify CHA5 with changes specific to clade 2.3 viruses. Amino acid alignment analysis of HA RBD in Fig. 1A revealed several clade-specific mutations between CHA5 and individual strain(s). For example, A102V, S145L, and K228R are specific to clade 1 viruses, whereas D110N, S140D, and K205R are specific to clade 2 viruses that are susceptible to CHA5-induced antiserum. For the insusceptible strains ID05, AN, and FJ, the only common change is S157P. In addition, V190I and P197S are specific to clade 2.3 viruses.

Fig. 1.

(A) Genetic alignment of the globular region of H5N1 HAs. Protein sequence alignment of the HA-H5N1 globular region. The mutations S157P, V190I, and P197S, highlighted in red, are present in A/Indonesia/5/2005, A/duck/Fujian/1734/05, and A/Anhui/1/2005 HA, but not in pCHA5. Furthermore, D110N, S140D, and K205R substitutions (highlighted in blue) were observed in clade 2 viruses and in pCHA5II. (B) Phylogenetic analysis of the HA used in the study.

The second approach was to redesign the consensus sequence by including the most recently circulating strains. The updated consensus DNA vaccine pCHA5II was located closer to the center of the phylogenetic analysis of HA from H5N1 viruses (Fig. 1B). As shown in Fig. 1A, the pCHA5II has different amino acids at D110N, S140D, and K205R.

Evaluation of Antigenicity and Immunogenicity in RBD-Mutated Hemagglutinin.

The impact of the 157th, 190th, and 197th residues in the HA globular region on immunogenicity and induction of neutralization activities was evaluated using a pseudotyped-virus platform. We revised CHA5 to include the S157P, V190I, or P197S mutation and then used these constructs as the vaccine antigens. The vaccine-elicited antisera were then subsequently evaluated for their neutralization profile. Compared to CHA5, the CHA5 S157P-induced antiserum can better neutralize clade 2.3 viruses as the EC₉₀ values were three times higher. On the other hand, CHA5 S157P antiserum exhibited reduced neutralization activity to clade 1 and clade 2.2 viruses (Table 1). To confirm the effects of these residues, we also asked whether the reversal of the specific amino acid mutations in HA from the clade 2.3 viruses FJ and AN would make the resulted pseudotyped virus easier to be neutralized by the serum. Therefore, the 157th, 190th, and 197th residues of HA of FJ and AN were modified to have the same amino acid present in other virus strains: i.e., P157S, I190V, and S197P. The modified HA were then used to produce pseudoviruses; the resulting pseudoviruses were then analyzed for the neutralization activities of CHA5-induced antiserum. Table 2 shows that CHA5 antiserum can successfully neutralize the pseudotype viruses containing the mutations, with two- to threefold increases in titer when compared to the activity to neutralize the wild-type pseudoviruses.

Table 1.

Effect of the amino acid mutations in the globular region on CHA5 vaccine immunogenicity

Antisera†	EC90 of HA pseudotyped virus *
	VN1194‡ (clade 1)§	ID05 (clade 2.1)	TK (clade 2.2)	E319-02 (clade 2.3.2)	FJ (clade 2.3.4)	AN (clade 2.3.4)
CHA5	252 ± 46	89 ± 59	400 ± 67	159 ± 32	63 ± 15	79 ± 18
CHA5 S157P	32 ± 9	100 ± 22	25 ± 8	32 ± 9	200 ± 38	173 ± 22
CHA5 V190I	178 ± 34	72 ± 41	429 ± 85	122 ± 21	59 ± 18	81 ± 23
CHA5 P197S	224 ± 41	63 ± 15	389 ± 62	171 ± 38	63 ± 15	79 ± 18
CHA5II	252 ± 46	800 ± 124	1,600 ± 217	800 ± 120	200 ± 90	224 ± 99

TK, A/turkey/Turkey/1/2005.

^*The geometric means of EC₉₀ neutralization titer from three independent experiments.

^†The DNA vaccine antigen was used to elicit antiserum.

^‡The HA of the virus was used to produce HA-pseudotyped virus.

^§The clade to which the virus belongs.

Table 2.

Effect of amino acid residues on the virus susceptibility of neutralization by CHA5- and CHA5II-induced antiserum

Virus *	EC90 neutralization titers †
	CHA5 ‡	CHA5II
FJ	63 ± 15	200 ± 90
FJ P157S	200 ± 39	800 ± 132
FJ I190V	162 ± 27	598 ± 124
FJ S197P	158 ± 31	634 ± 103
AN	79 ± 18	224 ± 99
AN P157S	200 ± 37	500 ± 81
AN I190V	155 ± 32	462 ± 73
AN S197P	125 ± 26	317 ± 58

^*The HA of the virus was used to produce HA-pseudotyped virus.

^†The geometric means of EC₉₀ neutralization titer form more than three independent experiments.

^‡The DNA vaccine antigen was used to elicit antiserum.

The results confirmed that a single amino acid, proline 157, may determine the susceptibility of the circulating clade 2.3 viral strains to CHA5-induced immunity; that is, the influenza virus may only need a S157P mutation to change its immunogenicity profile and thus evade CHA5-induceed immunity. Moreover, proline 157 can dramatically shift the immunogenicity of a vaccine strain specific to clade 2.3 viruses.

Evaluation of Immunogenicity and Neutralization Activities of pCHA5II.

The activities of pCHA5II were compared parallelly. As shown in Table 1, CHA5II-elicited antisera showed higher neutralization activities against clade 2 viruses than the CHA5-induced antisera did. Indeed, CHA5II showed better activities than CHA5 containing individual mutation that is specific to clade 2.3 viruses. Furthermore, the CHA5II-induced antiserum also showed a two- to fourfold increase in neutralization titer against viruses pseudotyped with modified HA as shown in Table 2.

Receptor-Binding Preference of the Hemagglutinin Variants.

The amino acid mutation in the globular region of HA proteins was analyzed to determine receptor-binding specificity and avidity to sialosides in a direct glycan receptor-binding assay (17, 21, 22) (the structure of sialosides is shown in Fig. S1). CHA5, similar to H5 from virus strains, showed preference toward 2,3 sialosides, but not 2,6 sialosides. Glycan binding analyses of CHA5II, which includes the three mutations most commonly found in clade 2 strains (D110N, S140D, and K205R), exhibited binding specificity and avidity similar to those of CHA5 (Fig. 2A). Interestingly, introduction of the 157P mutation to CHA5 (CHA5 S157P) can significantly increase the receptor-binding avidity (Fig. 2B). CHA5II and CHA5 S157P exhibited binding specificity similar to that of CHA5.

Fig. 2.

Glycan microarray analyses of various HA proteins. The cellular surface receptor-binding ability of HA proteins was determined by glycochip-binding assay against α2,3 sialosides (1–17, structures shown in Fig. S2). (A) CHA5II is represented as an RBD-mutated HA with the D110N, S140D, and K205R mutations. (B) The S157P substitution was introduced to CHA5. The glycan-binding profile was compared to that of CHA5.

It was suggested that glycosylation of the HA protein is likely to generate steric hindrance to mask the antigenic site and to prevent immune system detection (23), resulting in altered virus receptor binding (24) as well as manipulation of the protein structure. Therefore, whether the increases in the sialosides binding activities were caused by the changes in the glycosylation sites or the secondary structures was further investigated. The results confirmed that CHA5, CHA5II (containing D110N, S140D, and K205R), and the modified CHA5 containing S157P have the same glycosylation sites (Table S1) and similar secondary structures (Fig. S2). We concluded that S157P substitution increased the binding avidity of HA protein for cell-surface glycan receptors but did not alter glycosylation or secondary structure of the protein. Thus, the mutations introduced in the HA globular region had various effects on receptor binding, antigenicity, and immunogenicity among clade 2.3 HAs.

Comparison of CHA5II-induced neutralization activities to the activities induced by H5 from representative viruses.

It was clear that CHA5II could induce broader neutralization activities among the artificial HA that we have so far. Whether CHA5II or other HAs from circulating virus strains can be a better vaccine candidate remains to be investigated. Therefore, we optimized the HA DNA sequences from all representative strains as DNA vaccines and analyzed the induced antiserum for their neutralization activities with our in vitro neutralization platform. Not surprisingly, homologous or intraclade neutralization to the HA from the viruses in the same clade was always the strongest (Fig. 3). HAs in clade 1 vs. clade 2 have the most distinct serotype characteristics, because the induced antiserum did not have good cross-clase neutralization activities. The results also showed that the clade 2.3 viruses (FJ and AN) appeared to have a tight neutralization profile, suggesting that the current vaccine strains from other clades would not have the ability against clade 2.3 viruses. Likewise, the antiserum from HA of clade 2.3 viruses showed little cross-neutralization activity to other clades, even to homologous strains. Recently, the vaccine based on AN also showed lack of protection against most influenza viruses (25).

Fig. 3.

Cross-neutralization activities of antisera from various HA (serotypes) against a set of H5N1 pseudotyped viruses (genotypes). The neutralizing titer EC₉₀ was presented by the heatmap that was generated using the web tool of HIV databases. The green spot means that the EC₉₀ endpoint titer elicited by specific HA was significantly higher than 95% confidence interval (> 95% C.I.) compared to other antisera against the same pseudovirus. The EC₉₀ endpoint titer lower than 25, i.e., no neutralization activities, was indicated as “< 25.”

Consistent with our previous studies, CHA5-induced antiserum showed lower neutralization activities toward clade 2.1 (ID05) and clade 2.3 (FJ and AN) viruses. The updated consensus HA-based DNA vaccine CHA5II could induce antiserum that showed the neutralizing superiority to the antisera induced by other optimized HA DNA sequences from the representative strains (Fig. 3). The results indicated that, by shifting the DNA vaccine candidate toward the center position of the phylogenetic tree, pCHA5II indeed conferred broader protection profile than pCHA5, which was located at the outer most of the tree and between clade 1 and clade 2. In addition, comparing pCHA5 and pCHA5II, there are three mutations (D110N, S140D, and K205R) in the globular region. The introduction of these mutations to CHA5 not only enabled the mounting of an effective immune response to clade 2 H5N1 viruses, but also preserved the cross-neutralization ability to the rest of the viruses.

The pCHA5II Conferred Cross-Clade Protection in Virus-Challenged Mice.

To confirm that pCHA5II retained the activities against clade 1 viruses while overcoming the limited protection of CHA5 against A/Indonesia/5/2005/RG2, we challenged the pCHA5II-immunized mice with lethal doses of wild-type A/Vietnam/1194/2004 (Fig. 4 A and B) and RG2 viruses (Fig. 4 C and D). As shown in Fig. 4, pCHA5II protected 90% and 100% of the mice from the threats of the influenza H5N1 viruses in clade 1 and clade 2 viruses, respectively. Moreover, pCHA5II could better protect mice from body weight loss during the course of A/Indonesia/5/2005/RG2 virus challenges.

Fig. 4.

Vaccine protection against lethal challenge by wild-type A/Vietnam/1194/2004 H5N1 virus (A and B) or Indonesia/5/2005/RG2 virus (C and D). BALB/c mice were immunized with two injections of pCHA5, pCHA5II, or the control plasmid (pVAX) using intramuscular immunization with electroporation. The immunized mice were intranasally challenged with the wild-type (A/Vietnam/1194/2004, clade 1) or reassortant (RG2, clade 2.1) H5N1 viruses (n = 10 per group). After virus challenge, survival (A and C) and body weight (B and D) were recorded for 14 d.

Discussion

In this report, two approaches were initially evaluated for improving the protection profiles of our first-generation DNA vaccine, pCHA5. Based on the alignment of the HA sequence of CHA5 immunity-resistant strains, we identified unique amino acid mutation(s) in the HA RBD domain of these insusceptible strains. The HA globular region of influenza virus is exposed to the outside environment and described as a viral spike (10, 21, 26). The changes in the HA globular region would thus allow the influenza virus to alter the efficiency for virus entry as well as for evading host immunity. The phylogenetic comparison of H5N1 HAs between clade 1 and clade 2 showed that there are multiple antigenic variations proximal to the RBD (11, 27). Our evaluation using CHA5 harboring individual mutations concluded that S157P is the most critical one to influence HA antigenicity. Based on the evidence from a monoclonal antibody-neutralizing assay, the H5 HA protein is believed to comprise five distinct neutralizing epitopes (28) and three immune escape-related antigenic sites (11). The S157P mutation affects the group 5 epitope (27) but direct evidence of any correlation with immune escape was lacking. Herein, we proved that the mutation at position 157 of HA in clade 2.3 viruses conferred resistance to CHA5-induced immunity. Furthermore, the 157P mutation altered CHA5 immunogenicity from cross-clade immunity to clade 2.3-specific immunity. The S157P mutation was also able to increase the receptor-binding avidity of HA protein, which is important for influenza virus to evade host immunity. This evidence suggests that residue 157 and related epitopes merit further investigation.

We hypothesized that certain H5N1 viruses might be able to increase receptor avidity and alter specificity through N-linked glycosylation. This phenomenon is especially true for H5N1 clade 2 strains; viruses in clade 2.2 have evolved to possess receptor-binding specificity to α2,6 sialic acid, due to the loss of glycosylation at residue 158 and acquisition of the 193R residue (29). Some evidence suggested that shorter glycan chain on the HA protein could increase receptor binding and immunogenicity (30). In this report, however, we showed that the 157P residue in recent circulating clade 2.3 viruses was critical for immune escape but did not modulate N-linked glycosylation or secondary protein structure.

The second approach is to redesign the consensus HA. We revised our prototype DNA vaccine pCHA5 to pCHA5II after rededucing the consensus sequence from 1,192 full-length H5 sequences. Because pCHA5II was deduced from all HA sequences by 2007, rather than May 2006 for pCHA5, these residues represent the new mutations that emerged in circulating clade 2.3 strains. There are five mutations from pCHA5 to CHA5II (F8L, D110N, S140D, K205R, and I529T). Among the five mutations, three of the mutations are in the receptor-binding subdomain (S140D and K205R) or proximity vestigial esterase domain (D110N). S140D appears to be in the H5-escape-related antigenic site 1, an exposed loop (Ca2 of H1/HA3 140 to 145/H5 136 to 141) for antibody binding (31, 32). Another mutation, K205R, corresponding to the antigenic site D of H3, may influence class I cytotoxic T lymphocyte recognition and facilitate viral escape from CHA5 immunity (33, 34). Furthermore, D110N is located in a well-defined MHC-II-presenting epitope in H1 HA (S1; amino acids 107–119) (35); mutation of this residue may increase antigenicity. Indeed, our serotyping results showed that CHA5II could confer broader neutralizing activity than CHA5. It was surmised that the amino acid mutations surrounding the RBD correlated not only to T-cell epitopes but also to B-cell epitopes, as well as to vaccine immunogenicity. Further research to evaluate the effects of these distinctive mutations on vaccine immunogenicity is ongoing.

To identify potential DNA vaccine candidates for an H5N1 influenza epidemic, one can also screen for the best candidate that shows the broadest protection profile among the circulating virus strains. Our screening results confirmed that our pCHA5II can elicit antiserum that provides the broadest protection profile. It is not surprising to observe significant intraclade protection, but it is rather exciting to learn that some specific HAs have cross-clade neutralization activities. It was noticeable that ID05, categorized to clade 2.1, can confer cross-neutralization activity to clade 2.2 and 2.3 viruses. Furthermore, HA from Turkey (pTK), which was in clade 2.2 and close to the central position in HA phylogenetic tree, could induce a similar cross-neutralization pattern as pCHA5 did. The results shown in Fig. 3 are consistent with the World Health Organization (WHO) suggestion and the cross-neutralization results from many groups, in which ID05 and TK in clade 2.2 viruses were suggested to be vaccine strains for an H5N1 epidemic (36, 37). The ID05-based virus-like particles were also proven to induce protective immunity against other clade 1 viruses (38, 39).

The neutralization trend shown in Fig. 3, where the pCHA5II, compared to pCHA5, can elicit higher cross-neutralization activities against all clade 2 viruses and retain admirable neutralization activity for clade 1 viruses, correlated well with the animal challenge results in Fig. 4. Thus, the neutralization assay against HA-pseudotyped virus provides a reliable platform for determining H5N1 virus antigenicity and can be an effective platform for identification of adequate vaccine strains for a flu pandemic. The WHO as well as the Food and Drug Administration criteria for assessing pandemic influenza vaccines in immunological naïve populations (40) are based on nonrandomized noncontrol HI antibody studies, though they still have some significant implications. Furthermore, the HI assay may underestimate human immune responses to influenza viruses (41). To improve the prediction of the protection profiles of a given vaccine, a convenient, sensitive, and quantitative assay that enables wide-spectrum analysis to probe the cross-protection profiles could facilitate the vaccine development process. Using the pseudotyped-virus neutralization analysis platform, we have successfully evaluated the correlation of genotypes and genotypes of H5N1 viruses. The platform is flexible to deal with any viruses without the safety issues and can be easily standardized; therefore, our pseudovirus neutralization platform may provide a better alternative for evaluating the immune response generated by influenza vaccine as well as for evaluating the cross-neutralization activities critical for vaccine development. In addition, pseudovirus-based neutralization platform like this will also provide clade information of a suspected virus strain. As a matter of fact, our result based on serological analysis using pseudotyped-virus assay implied that ID05 is antigentically related to clade 2.2 viruses, rather than clade 2.1 viruses. A similar suggestion has indeed been claimed by WHO.

The results in Fig. 3 also provided an answer to a long-term question, whether the codes for cross-neutralization lied in the sequences of HA itself or cross-neutralization was elicited by the great enhancement in immunogenicity by the electroporation strategy. As shown in Fig. 3, various optimized HAs can induce antiserum with differential neutralization activities, confirming that the sequence of HA indeed dominated the immunogenicity. Our results also showed that CHA5II might provide a promising and facile approach for a prophylactic that needs minimal prediction and annually updating.

Materials and Methods

Viruses.

The attenuated reassortant H5N1 influenza viruses A/Indonesia/5/2005 (RG2) was obtained from the Center for Disease Control in Indonesia. All viruses were cultivated in the allantoic cavity of specific-pathogen-free (SPF) embryonated eggs, titered in Madin–Darbey canine kidney (MDCK) cells, and expressed as 50% tissue culture infective dose (TCID₅₀). The LD₅₀ in mice was determined for each virus before use in challenge experiments.

Vaccine and Plasmid Construction.

CHA5II was generated from 1,192 full-length HA sequences by the end of 2007, using the same method as described previously (1). Using pCHA5 as a template, CHA5II and HAs from various H5N1 viruses were constructed by site-directed mutagenesis (Multi Site-Directed Mutagenesis Kit, Stratagene). Furthermore, the neuraminidase gene from influenza virus A/Vietnam/1194/2004 was also optimized, synthesized, and cloned into pVAX (pNA) for use in the production of HA-pseudotyped viruses.

Vaccination.

Five- to 6-wk-old female BALB/c mice were immunized with endotoxin-free pCHA5II or other HA constructs, prepared with GenElute™ HP Endotoxin-Free Plasmid Maxiprep Kit (Sigma-Aldrich). The DNA vaccine was given 30 μg per mouse at weeks 0 and 3 by an intramuscular administration of plasmids followed by instantaneous electrical stimulation (TriGrid Delivery System, Ichor) (1, 42). The spacing of the TriGrid electrode array was 2.5 mm, and the electrical field was applied at an amplitude of 250 V/cm of electrode spacing for a duration of 40 ms three times over a 400 ms interval. After immunization, the mice were housed in the SPF animal facility at the Institute of Cell Biology, Academia Sinica, Taiwan. Two weeks after the second immunization, the immunized mice were bled for HA-specific antibody analysis and for neutralization assay to assess vaccine efficacy. All animal experiments were evaluated and approved by the Institutional Animal Care and Use Committee of Academia Sinica, Taiwan.

HA-Pseudotyped-Virus Assay.

All the procedures were described previously (1). Briefly, human 293T cells were cotransfected with three plasmids: pNA, pNL4-3.Luc.R-E- (National Institute of Allergy and Infectious Diseases, National Institutes of Health) (43), and HA of interest. After overnight incubation, the transfected cells were washed with PBS and incubated with fresh culture medium for another 24 h. The supernatant was then harvested and was used for neutralization assay. The mixtures containing 50 × TCID₅₀ of HA-pseudotyped viruses and various dilutions of the antiserum were incubated at 37 °C for 30 min and added to MDCK cells at 37 °C. The plates were washed after 4 h incubation and replenished with fresh medium. After 48 h, luciferase activity in the cells was determined by the Luciferase Assay System (Promega). The relative luminescence values determined in the wells containing cells and HA-pseudotyped virus were defined as 0% neutralization; the values in the cells-only wells were defined as 100% neutralization. The maximum antiserum dilution fold for 90% neutralization was defined as EC₉₀ endpoint titer.

Generation of a Heatmap for HA Serotyping Using Virus Neutralization Titers.

The HA serotyping analysis was generated based on the EC₉₀ endpoint titer determined in pseudotyped-virus neutralization assays. Geometric mean titer and 95% C.I. for given pseudoviruses or antiserum were calculated from the EC₉₀ titers determined in three independent experiments. The EC₉₀ titer data were displayed by the heatmap using the web tool available in HIV databases (http://www.hiv.lanl.gov/content/sequence/HEATMAP/heatmap.html).

Virus Challenge Experiments.

Two weeks after the second immunization, the immunized mice were anesthetized and intranasally challenged with a 50LD₅₀ of the reassortant RG2 and 100LD₅₀ of the wild-type A/Vietnam/1194/2004 H5N1 virus. After infection, the mice were observed daily for 14 d, and survival and clinical parameters such as body weight were recorded. The challenge experiments were performed under biosafety level-2-plus (for RG2) and level-3 (for wild-type Vietnam/1194 H5N1 virus) enhancement conditions.

Statistical Analysis.

The animal experiments to evaluate neutralization activities of the serum were repeated at least three times (n = 3 per group). The virus challenge experiments were conducted with n = 10 per group. Amino acid ClustalW multiple alignment and the threshold frequency for inclusion in consensus sequences were conducted by BioEdit Sequence Alignment Editor version 7.0.5.3 (44).