Discordant antigenic drift of neuraminidase and hemagglutinin in H1N1 and H3N2 influenza viruses

Antigenic Characterization of the HA and NA of H1N1 and H3N2 Viruses.

H6 reassortant viruses containing NA of historical as well as recent H1N1 and H3N2 vaccine viruses (Table S1) were used to measure NI titers of strain-specific ferret antisera. There was minimal NI cross-reactivity between the phylogenetically distant early NAs and antisera raised against recent seasonal strains of the classical H1N1 (Table S2) and H3N2 (Table S3) human lineages, demonstrating extensive antigenic drift since introduction of these subtypes. Ferret serum raised against A/California/7/2009 (CA/09), a representative 2009 H1N1 pandemic (H1N1pdm) virus, demonstrated a robust homologous NI antibody titer with very low cross-reactive titers against NAs of the long-established human seasonal H1N1 lineage (Table S2).

Phylogenetic trees, genetic maps based on amino acid sequences, and antigenic maps were generated for HA and NA of H1N1 (Fig. 1) and H3N2 (Fig. 2) viruses recommended for inclusion in 1996–2009 seasonal influenza vaccines. Since NI data had not previously been used to generate an NA antigenic map, we first established that end-point NI titers provided sufficient sensitivity for the analysis. This process was done by comparative analysis of end-point titers (inverse of the serum dilution that inhibits NA activity ≥50%) and the precise 50% inhibition titer (IC50), the inverse of the serum dilution that results in exactly 50% inhibition determined by nonlinear regression analysis (14) of ferret antisera generated in response to H3N2 infection. The maps generated using these datasets were very similar, with an excellent correlation of distances between antigens on the map, r² = 0.99 (Fig. S1). End-point NI titers were therefore used in subsequent analyses. We also showed that there was good correlation between antigenic distance determined using NI data in tables and the antigenic distance determined from map location (Fig. S2) (r² = 0.95 for N1 and r² = 0.86 for N2 antigens), providing confidence that the antigenic maps are representative of the raw data. To confirm the location of antigens on the map, NI assays were repeated using a larger number of antisera. There was excellent correlation between the location of antigens generated by analyses of the first and second datasets (Fig. S3) (r² = 0.98 for both N1 and N2 antigens).

The raw NI data as well as the antigenic map show that the NA of human H1N1 viruses was in antigenic stasis for over a decade, as antisera to vaccine strains A/Texas/36/1991 (TX/91), A/New Caledonia/20/1999 (NC/99), and A/Solomon Islands/3/2006 (SI/06) each had similar NI titers against all three corresponding NAs, even though there had been substantial genetic diversity between them. In contrast, an abrupt antigenic change was observed for the NA of A/Brisbane/59/2007 (BR/07), the H1N1 virus recommended for use in Northern Hemisphere trivalent vaccines during the 2008/09 and 2009/10 seasons (Fig. 1 A, C, and E, and Table S2). Antisera to the older strains had at least 16-fold lower NI titers against BR/07 than against the homologous antigen. Antigenic differences in the HA for these strains were quite the opposite: HAs of TX/91, NC/99 and SI/06 were distinct from one another, but the antigenic difference between HAs of SI/06 and BR/07 was less pronounced (Fig. 1F and Table S2). The antigenic distance between the NA of BR/07 and the previous vaccine strain, SI/06, is 4.6 units (Fig. 1E), which is greater than the antigenic distance between the HAs of these viruses (2.0 units). From this observation, we speculate that the abrupt antigenic change of N1 NA may have been a factor contributing to the dominance of H1N1 in many parts of the Northern Hemisphere during the 2007/08 season (15).

The antigenic drift of NA was also not proportional to number of amino acid changes for H3N2 viruses isolated from the mid-1990s [A/Wuhan/359/95 (WU/95)] to the 2009/10 Northern hemisphere formulation [A/Uruguay/716/2007 (UY/07), an A/Brisbane/10/2007-like virus], even though phylogenetic analysis of the NAs of these viruses shows a chronologic progressive evolution at the nucleotide level (Fig. 2A). The antigenic analyses (Table S3) and maps (Fig. 2 E and F) of H3N2 vaccine viruses show that the antigenic drift of HA and NA of this subtype is also discordant. A major step in HA antigenic drift between WU/95 and A/Sydney/5/1997 (SY/97) was well-recognized previously (10, 16). However, there is almost no genetic or antigenic difference between the NAs of WU/95 and SY/97. Despite this antigenic similarity in the NAs, effectiveness of the 1997/98 inactivated vaccine was poor for healthy adults (17), suggesting that NA in the vaccine did not induce levels of antibodies with capacity to protect against laboratory-confirmed influenza-like illness.

The recommended H3N2 component for seasonal vaccine formulations from 2000/01–2003/04 was A/Moscow/10/1999 (MW/99). Because of difficulty in generating a high-growth MW/99 reassortant and based on HI data, A/Panama/2007/1999 (PM/99) was widely used for vaccine production as a MW/99-like virus. At the time, the difference in MW/99 and PM/99 NA sequences was of some concern (18), but it was not known whether this corresponded with a difference in antigenic properties. Our results show the NAs of MW/99 and PM/99 are antigenically distinct, with the NA of MW/99 positioned closer to the subsequent A/Fujian/411/2002-like vaccine strain, A/Wyoming/3/2004 (WY/03), than to PM/99 (Fig. 2E).

There tends to be an asymmetrical aspect to the reactivity of antisera with NA; antisera raised against old strains reacted weakly with more recent viruses, whereas reactivity of sera prepared against more recent viruses was often retained in NI tests against older strains. For example, antisera to BR/07 had good inhibition of TX/91 NA, and antisera to WY/03 had excellent inhibition of WU/95 NA (Table S2 and S3). An exception to this trend was the antigenic profile of PM/99's NA. This virus had weak reactivity with all sera generated against subsequent viruses.

Sera from Vaccinated Individuals Discriminate Between Viruses with Drifted NAs.

Whether NA antigenic properties determined from reactivity with ferret antisera are highly relevant predictors of human antibody titers is an important question. Ferret sera differ from human sera in the complexity of past history with influenza virus antigens, particularly because ferrets are screened to ensure seronegative status before inoculation with reference viruses, making their influenza immunity monospecific. In a previous study, we analyzed human sera collected before and after immunization with 2006–07 seasonal influenza vaccines for NI titers against the homologous NC/99 N1 and A/Wisconsin/67/2005 (WI/05) N2 antigens (14). From the small cohort in that study, we identified subsets of vaccinees with increases in serum NI titer against N1 or N2 that were robust enough to be consistently measured across repeated assays. Serum pairs that showed distinct increases in NI titer against N1 or N2 [in each case, three volunteers vaccinated with live attenuated influenza vaccine (LAIV) and three with trivalent inactivated vaccine (TIV)] were reassayed against homologous NA antigens plus NA antigens of the H3N2 or H1N1 vaccine viruses of subsequent vaccine campaigns.

There were two new H1N1 vaccine strain recommendations after the 2006–07 clinical trial: SI/06 and BR/07. Prevaccine NI titers against NC/99 and SI/06 NAs were similar in magnitude, and for most serum pairs, NI titers against NA of the vaccine strain, NC/99, and SI/06 increased to similar levels postvaccination (Fig. 3A). Both observations are consistent with the data from NI tests with use of monospecific ferret reference sera (Fig. 1C), which showed high antigenic similarity between these two NAs. In contrast, most volunteers’ prevaccine sera showed significantly less reactivity against the more recent BR/07 NA. After vaccination, NI titers against the BR/07 NA remained significantly lower than titers against the two earlier strains. The mean fold-increase in titer against BR/07 NA (1.2) was less than that against SI/06 NA (1.75) and homologous NC/99 NA (2.2), although differences were not statistically significant (P = 0.16). These data from human sera mirror the trend observed in assays with monospecific ferret sera, where BR/07 NA marked a substantial antigenic drift in the human H1N1 influenza lineage after great similarity among NAs of TX/91, NC/99, and SI/06.

The 2006 recommended H3N2 vaccine strain, WI/05, was updated in 2008 to include UY/07. All of the prevaccine serum samples possessed detectable NI activity against the NA of WI/05 and UY/07 (Fig. 3B). After vaccination with LAIV or TIV, there were significantly greater increases in NI titer against WI/05 NA than against UY/07 NA (Fig. 3D) (P < 0.05). Thus, the modest antigenic distinction between two N2 NAs that was indicated by assaying ferret reference antisera was confirmed with sera of human vaccinees, showing that it is reasonable to correlate ferret serologic responses with human responses.

Single Amino Acid Change Is Largely Responsible for the Antigenic Drift of the SI/06 NA.

Studies of NA's antigenic structure with monoclonal antibodies have been used together with molecular analysis of escape variants to identify epitopes in N2 NA (19, 20), providing clues as to which sequence changes in circulating viruses signify antigenic variation. After identifying globular head-region amino acids that differ between N1 NA proteins of BR/07 and both NC/99 and SI/06, we used site-specific mutagenesis to generate constructs containing the NA of BR/07 in which specific amino acids from the corresponding sequence of SI/06 and NC/99 were introduced (Fig. 4A). All of the amino acid changes are surface residues at the opposite side to the tetramer interface (Fig. 5) and are thus potentially accessible to antibodies. Amino acid changes at positions 222, 329, and 344 (using amino acid numbering for the N1 sequence) correspond to antibody-binding determinants previously identified for N2 proteins (20, 21). Three of the variant residues (222, 249, and 344) are located on the periphery of the enzyme active site, and another (329) is located on a loop more distant from the active site. None of these changes is predicted to add or remove a glycosylation site. The reactivity of ferret anti-BR/07, anti-SI/06, and anti-NC/99 sera was tested against the mutated antigens (Fig. 4 B–D). Antisera generated against NC/99 and SI/06 could inhibit the BR/07 NA when amino acid 329 was changed from Glu to Lys. Some reactivity with antiserum to NC/99 was restored when amino acid 222 (Gln) was changed to Arg, 249 (Lys) was changed to Gly, or 287 (Ile) was changed to Thr. Of note, none of the mutants lost reactivity with antiserum against BR/07, reflecting the retention of reactivity with previous viruses noted earlier in this report and supporting the idea that other subdominant conserved epitopes are present. Most importantly, sera from human volunteers previously noted as responding to the vaccine antigen NC/99 regained full or partial reactivity with BR/07 that included the single E329K mutation (Fig. 4 E and F). Although the precise structural determinants required for antibody binding are unclear, the impact of the E329K substitution suggests that a charge interaction is important for this NA–antibody interaction. Our results demonstrate the impact of a single point mutation on NA drift and its consequence on recognition by human sera.

In summary, our data provide confirmatory evidence of antigenic drift of NA that is often discordant with the antigenic drift of HA across contemporary H1N1 and H3N2 vaccine viruses. The finding that a single point mutation in NA is predominantly responsible for the antigenic drift between SI/06 and BR/07 highlights the importance of evaluating antigenic properties in addition to sequence changes when characterizing the NAs of emerging influenza viruses. In addition, our results show that for NC/99 (H1N1) and WI/05 (H3N2) components of the 2006/07 vaccine, human NA-specific antibodies reacted in a similar manner as monospecific ferret antisera, suggesting the antigenic drift identified by analysis of ferret antisera is relevant to human serologic responses. Although HA is the most abundant and important antigen for inducing neutralizing antibody responses, the contribution of NA-specific antibodies to efficacy of vaccines under development that contain significant amounts of immunogenic NA, such as virus-like particles, recombinant proteins, and high-dose vaccines (22–24), should not be overlooked. Previous studies have pointed out the importance of NA-based immunity in curbing the severity or spread of emerging strains with novel HAs. Epidemiologic analyses (25, 26) support the concept that NA antibodies contributed to protection against the 1968 pandemic H3N2 virus that possessed an NA antigenically similar to that of the previously circulating H2N2 lineage (27, 28). In addition, antibodies directed to conserved regions of NA may contribute a significant level of protection against disease: the presence of such heterologous immunity has been demonstrated in H3N2 (29) and H5N1 (30) challenge studies in mice. Studies to identify the NI titer that correlates with protection in man, the vaccine dose needed to elicit a protective response, and the breadth of this protection in man, are certainly warranted to facilitate the development of NA-containing influenza vaccines.

Viruses.

The sources of viruses and plasmids are provided in SI Materials and Methods. NA genes were cloned from H1N1 and H3N2 vaccine seed strains listed in Table S1 and reassortant viruses generated by eight-plasmid reverse genetics for the NI assays, as described previously (14). Single amino acid mutations were introduced to the plasmid-encoded BR/07 NA gene using the QuikChange II site-directed mutagenesis kit (Agilent Technologies). Virus stocks were amplified in 10-d embryonated chicken eggs for use in NI and HI assays. Viruses grown in Madin-Darby canine kidney cells obtained from the National Influenza Center, Erasmus MC, Rotterdam, The Netherlands, were also used in HI testing.

Genetic Analyses.

Phylogenetic trees and genetic maps based on the coding region of NA or HA1 were constructed to analyze the genetic evolution of human H1N1 and H3N2 viruses, as described in SI Materials and Methods.

Sera and Serological Assays.

The sources of ferret and human antisera used in NI and HI analyses are provided in SI Materials and Methods. Titration of serum NI antibodies was performed by analyzing NA activity of the HA-mismatched reassortant viruses in a 96-well plate format of the conventional thiobarbituric acid assay (14), as described in SI Materials and Methods. A conventional assay was used to determine HI titers (31), as described in SI Materials and Methods.

Antigenic Cartography.

The antigenic properties of human influenza A (H1N1) and human influenza A (H3N2) viruses were characterized using NI and HI data and antigenic cartography methods as described previously for HI data of human influenza A (H3N2) viruses (12) and in SI Materials and Methods.