Haemagglutinin mutations and glycosylation changes shaped the 2012 / 13 influenza A ( H 3 N 2 ) epidemic , Houston , Texas

K M Stucker1, S A Schobel2,3, R J Olsen4, H L Hodges1,5, X Lin1, R A Halpin1, N Fedorova1, T B Stockwell2, A Tovchigrechko6, S R Das1, D E Wentworth (dwentworth@cdc.gov)1,7, J M Musser (jmmusser@houstonmethodist.org)4 1. Virology Group, J. Craig Venter Institute, Rockville, Maryland, United States 2. Bioinformatics Group, J. Craig Venter Institute, Rockville, Maryland, United States 3. Center for Bioinformatics and Computational Biology, University of Maryland, College Park, Maryland, United States 4. Department of Pathology and Genomic Medicine, Houston Methodist Hospital, and Center for Molecular and Translational Human Infectious Diseases Research, Houston Methodist Research Institute, Houston, Texas, United States 5. Current address: Department of Chemistry, University of Wisconsin-Madison, Madison, Wisconsin, United States 6. Genomic Medicine Group, J. Craig Venter Institute, Rockville, Maryland, United States 7. Current address: Virology Surveillance and Diagnosis Branch, Influenza Division, Centers for Disease Control and Prevention, Atlanta, Georgia, United States


Introduction
Influenza viruses cause significant annual morbidity and mortality in the global human population [1].Epidemics occur during the winter months, cycling roughly every six months between the northern and southern hemispheres.Two influenza A virus (IAV) subtypes, H3N2 and H1N1pdm09, and two influenza B virus (IBV) lineages, B/Yamagata and B/Victoria, have been circulating among humans since 2009.Epidemics caused by each of these subtypes/lineages vary from season to season; this is due, in part, to the selective advantage acquired by one subtype/lineage in a given season.Although the error-prone IAV RNA polymerase frequently generates nucleotide substitutions that can lead to a selective advantage in all eight genomic viral RNA segments (vRNAs), new epidemic variants are most frequently due to accumulated substitutions in the two surface glycoproteins, haemagglutinin (HA) and neuraminidase (NA).Substitutions in HA/NA can give rise to variants that escape host immunity from previous exposures or vaccinations and are selected in the non-naïve host population through a process called antigenic drift.Antigenic drift primarily occurs in epitopes recognised by antibodies that neutralise viral infectivity by blocking the interaction of HA with sialic acid receptors on host cell glycoproteins [2,3].Antigenic drift necessitates frequent updating of the strains used in the influenza vaccine [3,4].This requires global surveillance of the antigenic profile of circulating strains to inform the decisions made at biannual World Health Organization (WHO) meetings for the selection of influenza vaccine strains [4][5][6].The acquisition of as few as one to five mutations in HA can necessitate an updated vaccine strain to optimally protect the public [7,8].
The H3N2 subtype caused a severe epidemic during the 2012/13 influenza season in North America and contributed to a longer than normal season, with increased morbidity also in Europe [9,10].Human H3N2 viruses represent a very successful lineage that has circulated since the 'Hong Kong Flu' pandemic of 1968 [11,12].New H3N2 strains evolve continually, resulting in annual epidemics that are periodically severe, such as the 2003/04 season (A/Fujian/411/2002-like viruses) [3,7].The 2012/13 season in the United States (US) was notable in both greater epidemic severity and limited vaccine effectiveness, especially in the elderly [13][14][15][16][17][18].Our incomplete understanding of the factors shaping the emergence of epidemic IAV/IBV variants impairs our ability to accurately predict strain fitness and select appropriate vaccine strains.
We hypothesised that the 2012/13 epidemic was caused by the emergence of a new HA variant that rapidly displaced previously circulating strains.To gain a better understanding of the genetic and molecular mechanisms central to the intensity and severity of the 2012/13 epidemic, we sequenced 154 codingcomplete genomes from H3N2-positive nasopharyngeal swabs collected throughout the 2012/13 season from a large hospital network in Houston, Texas, US.This demographically diverse system of five hospitals provides a reasonable representation of the viruses circulating in the US during a non-pandemic seasonal influenza epidemic [19], which made it a suitable site for studying IAV genetic variation.Direct sequencing of primary swab specimens is critical because H3N2 viruses are known to rapidly acquire adaptive mutations that change their antigenicity when passaged in cell cultures or eggs [5,20].It allows accurate analysis of HA and NA, the identification of substitutions, and the detection of reassortment among vRNAs encoding internal proteins that contribute to viral fitness.

Hospital system and coding-complete viral genome sequencing
The Methodist Hospital System in Houston, Texas consists of five hospitals that serve a large population of ca 4 million people from the most ethnically diverse population in the United States (ca 32% Caucasian, 41% Hispanic/Latino, 20% African-American and 7% Asian) [21,22].The IAV-positive human nasopharyngeal swab specimens used in this study were collected between 3 November 2012 and 8 February 2013.During this period, ca 20% of the specimens were from inpatients and 80% were from outpatients.The majority of inpatient isolates were obtained from patients admitted with influenza-like illness through the emergency department, who therefore did not have nosocomially acquired infections.These viral samples were sequenced using the J. Craig Venter Institute's (JCVI's) high-throughput next-generation sequencing pipeline.Briefly, IAV vRNAs were isolated directly from the swab specimens, and the entire genome was amplified from 3 μl of RNA template using a multi-segment RT-PCR strategy (M-RTPCR) [23,24].The amplicons were sequenced using the Ion Torrent PGM (Thermo Fisher Scientific, Waltham, Massachusetts, US) and/or the Illumina MiSeq v2 (Illumina, Inc., San Diego, California, US) instruments.When sequencing data from both platforms was available, the data were merged and assembled together; the resulting consensus sequences were supported by reads from both technologies.

Sequence curation
Reference sequences for IAV HA and NA nucleotide sequences were obtained from GenBank (National Center for Biotechnology Information) and the EpiFlu database (Global Initiative on Sharing All Influenza Data; GISAID).For phylogenetic analyses, three H3N2 reference sets were used: (i) all available strains presented by the Centers for Disease Control and Prevention (CDC) at the 2012/13 US vaccine composition meeting (available at [25]), (ii) all available strains derived from a phylogenetic tree in the 2013 southern hemisphere vaccine composition report [26] and (iii) a random selection of 50 isolates from public databases collected globally between 1 September 2012 and 8 February 2013.For the genome constellation analysis, all available coding-complete human H3N2 genomes in the EpiFlu database collected between 1 October 2011 and 20 February 2013 were downloaded on 6 June 2013.

Phylogenetic analyses
HA and NA nucleotide sequences and HA amino sequences were each aligned using MAFFT v7.158b with default settings [27][28][29] and trimmed using Qiagen's CLC Genomics Workbench v6.0.Four independent maximum likelihood (ML) dendrograms were inferred using GARLI [30,31], and the trees with the best log likelihood scores were analysed further with 500 bootstrap replicates.The ML tree and bootstrap values were combined using SumTrees in DendroPy [32].For nucleotide sequences, the GTR G + I nucleotide substitution model with five rate categories was used, as determined by jModelTest v2.1.4[33,34].For protein sequences, the JTT G + I substitution model with four rate categories and empirically derived residue frequencies was used, as determined by ProtTest v2.4 [35].

Genome constellation analyses
Study and reference H3N2 genomes were parsed into gene segment-specific multi-FASTA files.MAFFT alignments for each segment were converted to a distance matrix using ANDES [36].Dendrograms were generated using the complete linkage (farthest neighbour) hierarchical clustering technique in ANDES.Dendrograms were cut at a specified per cent identity and vRNA segments were assigned to groups according to their cluster at that cut-off.The segment group assignments were combined to obtain a genome constellation that was visualised using OrionPlot [37].

Viral isolation and propagation
A subset of viruses was selected to represent each cocirculating clade.Viruses were isolated from the primary swab specimen for subsequent propagation and antigenic analysis.Selected strains were passaged twice in Madin-Darby canine kidney (MDCK) cells to avoid selecting for mutations that frequently arise when IAV is isolated in embryonated chicken eggs.Supernatants from P2 viral stocks were clarified by centrifugation at 1,800 × g for 10 min at 4 °C, aliquoted, stored at −80 °C and sequenced.

Influenza virus rescue using gene synthesis and reverse genetics
To avoid spurious substitutions and generate a uniform virus population, positive (A/Perth/16/2009(H3N2) and A/Victoria/361/2011(H3N2)) and negative (A/ Indiana/8/2011(H3N2v)) control viruses for antigenic testing were generated using gene synthesis [38] and a modified reverse genetics system [23,24] and were confirmed by sequencing.Briefly, 6:2 reassortant viruses (designated with an 'r' before the strain name, e.g.rA/Perth/16/2009) were rescued from plasmids encoding the six internal protein vRNAs from the strains indicated in Table 1 (e.g.A/Puerto Rico/8/1934) and the linear synthesised HA and NA genes of the desired H3N2 viruses.

Antisera generation/acquisition and haemagglutination inhibition
Ferret antisera were generated by inoculating ferrets (two individuals per virus) intranasally with ca 1 × 106 TCID 50 of clarified virus supernatant in 1 ml, and antisera were collected 30 days after inoculation.These ferret experiments were conducted by an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited company, BIOQUAL, Inc. (Rockville, Maryland, US).Antisera from the two ferrets inoculated with the same strain were pooled 1:1 and treated with three times their volume of receptordestroying enzyme (RDE; Hardy Diagnostics, Santa Maria, California, US) overnight at 37 °C.RDE was inactivated at 56 °C for 30 min, and the RDE-treated antisera were diluted with six times the original antisera volume using phosphate-buffered saline.Lyophilised sheep antisera raised against A/Victoria/361/2011 and A/Texas/50/2012 were provided by the Center for Biologics Evaluation and Research (CBER), Food and Drug Administration (FDA), Department of Health and Human Services, US, and were also RDE-treated.Haemagglutination inhibition (HI) assays were performed [39] to determine the ability of various antisera to inhibit binding of the IAV isolates to guinea pig red blood cells.

Modelling H3 N-linked glycosylation
All potential HA N-linked glycosylation sites among contemporary H3N2 clades in this study were deduced from sequons and compared with the HA of A/ Finland/486/2004(H3N2), for which the H3 structure has been solved [40].Comparing the asparagine (N)

Highlighted cells indicate two-way HI results
, where the specified serum is tested using the same (or a very similar) antigen to that used to generate the serum.An 'r' before the strain name indicates the virus was generated using gene synthesis and reverse genetics.Results from one experimental replicate are presented and are representative of the results obtained from at least two additional experimental replicates performed on a different day.All HA controls were between 3 and 8 HA units.
residues of predicted sequons with the three-dimensional H3 structure confirmed that they would all be on the protein surface.While crystal structures can often detect the core of N-linked glycans, the actual composition and structure of the high-mannose and complex sugars is poorly understood.Here, we use the high-mannose glycan deduced from small-angle X-ray scattering data for the N165-linked glycan of X31 (H3 structural numbering) [41] as a 'type glycan' modelled onto the A/Finland/483/2004 H3 structure (Protein Data Bank ID: 2YP3 [40]) using the AllosMod server [41] which implements an all-atom modelling approach allowing for limited protein conformational flexibility.

Phylogenetic analyses to understand H3N2 evolution
The coding-complete genomes of 154 IAV H3N2 specimens were sequenced and their genome sequences were submitted to GenBank (Table 2).
Phylogenetic analyses of HA nucleotide sequences showed that the majority of circulating H3N2 strains arose from three major H3 clades: 5, 3A, and 3C (Figure 1A; clade nomenclature was adopted from US CDC [25]).[47].These new subclades are related to, but didn't arise directly from the 2012/13 viruses circulating in Texas; their placements based on this larger analysis are labelled on the left of the HA nucleotide phylogeny (Figure 1A).
Strains from all three major clades (5, 3A, and 3C) were co-circulating in Houston throughout the early, middle, and late 2012/13 epidemic (data not shown), indicating that strong temporal segregation of viral clades was not occurring within a single IAV season (Figure 1A).This shows that the emergence of a genetically novel virus early in the season did not rapidly displace other co-circulating strains.
Analysis of HA protein sequences for the same dataset showed that there were relatively few residue changes across all HA clades (Figure 2).This is particularly true for the HA proteins of the subclade 3C.3 and subgroup 3C-2012/13, which is illustrated by the interleaving of colours based on the nucleotide clade assignments (Figure 1B).This HA protein phylogeny emphasises how subclades 3C.2 and 3C.3, as well as subgroup 3C-2012/13, are more closely related to each other than to clade 3C.1 viruses.
The NA nucleotide tree also formed a backbone characteristic of IAV that shows evidence of limited genetic drift in NA (Figure 1C).The NA phylogeny also showed three main groupings for the study samples, which roughly corresponded to the HA clade assignments 5, 3A, and 3C.However, colouring the NA tree using the HA clade assignments demonstrated that intrasubtypic reassortment between H3 and N2 occurs readily, as indicated by the interleaving of HA clade colours (Figure 1C).This was primarily found among 3C subclades and between clades 5 and 6.The most pronounced intrasubtypic reassortment was seen in A/Texas/ JMM_27/2012, which has an HA belonging to clade 3A and an NA that groups with 3C-2012/13 viruses.

Genome constellation analyses to identify reassortment
To better understand the extent of intrasubtypic reassortment occurring among seasonal H3N2 viruses, we performed genome constellation analyses using 198 genomes (154 from our study samples and 44 additional strains).A 98% nucleotide identity cut-off value provided a genome constellation resolution that matched that of the major HA clade designations (i.e. 1, 5, 6, 3A, 3B, and 3C) (Figure 3A).
This yielded 12 genome constellations and demonstrated that vRNAs coding for internal proteins are also involved in intrasubtypic reassortment.While all constellations shared highly similar PA and M segments, seven monophyletic HA 3C.2 samples had reassorted PB2 segments (constellation 12 in Figure 3A) that grouped with the PB2 of HA clade 3A (constellations 2  and 4 in Figure 3A).Similarly, while most HA clade 5 and 6 strains had unique PB2 and PB1 segments (blue boxes in the PB2 and PB1 columns of Figure 3A), three monophyletic HA clade 5 strains had retained the A/ Perth/16/2009-like PB2 and PB1 segments (orange boxes for PB2 and PB1 in constellation 7 in Figure 3A).Monophyletic HA strains sharing common intrasubtypic reassortment genome constellations demonstrated that these reassortment events are likely to occur once, creating viral strains that transmit and spread to new individuals, rather than occurring as multiple independent reassortment events.NA diversity was also observed among clade 5 and 6 viruses, which can also be appreciated by the interleaving colours when HA clade designations were mapped onto the NA phylogeny (Figure 1C).
To stringently explore the diversity among clade 3C viruses, a genome constellation analysis containing only the clade 3C genomes was performed using a 99% nucleotide identity cut-off value (Figure 3B).For the most part, the four 3C subclades/subgroups each contained their own unique genome constellations (Figure 3B).We found that most 3C-2012/13 viruses contained a unique NS segment compared with other 3C viruses (constellation 9 in Figure 3B), and all 3C-2012/13 viruses had a unique NP segment (constellations 8 and 9 in Figure 3B).In addition, seven monophyletic HA 3C.2 specimens contained unique PB2, PA, HA, and M segments at the 99% cut-off (constellation 10 in Figure 3B), and five monophyletic HA 3C.3 samples had incorporated a unique NA (constellation 6 in Figure 3B).
Overall, genome constellation analysis revealed that intrasubtypic reassortment occurs frequently in human H3N2 viruses.In this collection alone, intrasubtypic reassortment resulted in the expansion of antigenic drift variants into new genome constellations, the generation of new HA/NA combinations, and the acquisition of a unique NS vRNA.All residue changes that exist in HA among the study samples (yellow), vaccine strains (purple), and clade consensus sequences (white) are provided.All sequences that match the A/Perth/16/2009 strain (clade 1) are coloured in pink, while residues that differ from that strain are shown in blue (first residue difference) or green (second residue difference).Residues associated with the gain (+) or loss (−) of a potential glycosylation site are indicated.Antigenic site assignments are derived from several sources [60][61][62][63].P0 indicates the original clinical swab specimen, while P2 indicates the second-passage viral stock grown in MDCK cells.e Please note that each column is evaluated independently and cluster numbering is arbitrary.The number of samples observed for each genome constellation is provided, along with an HA clade assignment, for all constellations for which at least one sequence was also included in the HA phylogeny (Figure 1).Results are sorted first by HA, followed by NA, PB2, PB1, PA, NP, M, and NS.(A) Genome constellation 1 (orange) segments were defined as sharing 98% nucleotide identity with the A/Perth/16/2009 vaccine strain.Twelve unique genome constellations were identified at the 98% nucleotide identity cut-off, eight of which were observed in the study samples.
(B) Clade 3 genome constellation 1 (purple) segments were defined as sharing 99% nucleotide identity with the A/Victoria/361/2011 vaccine strain.10 unique genome constellations were identified at the 99% nucleotide identity cut-off, nine of which were observed in the study samples. A.

Genome constellation Count
Location of antigenic sites on the H3 monomer, along with key clade 3C substitutions and glycosylation sites In all panels, the peptide backbone of the HA globular head is represented as a ribbon with a translucent solid surface using the A/Finland/486/2004 HA crystal structure, 2YP3, bound to a synthetic 2,6-sialic acid ligand (cyan stick structure) [40].In panels on the right, all potential N-linked glycans in the H3 globular head were modelled using the AllosMod server [41] and rendered by PyMOL [45]  Assessment of antigenic evolution in a single epidemic season HI assays were performed [39] to determine the extent of antigenic drift among the H3 proteins during the 2012/13 epidemic.Two-way HI measurements, which paired an antigen with the antiserum raised against that antigen, provided HI titres that were typical for anti-H3N2 ferret and sheep antisera (Table 1).The specificity of the assay was shown by the negative control serum raised against a zoonotic H3N2 variant virus, (H3N2)v, which is antigenically similar to viruses that circulated in humans in the early to mid-1990s.Analysis of isolates propagated from specimens selected from the major clades showed that ferret and sheep polyclonal antisera against the recommended vaccine strains (A/ Victoria/361/2011 and A/Texas/50/2012) prevented the binding of clade 5, 3C.1, and 3C.2 isolates to the sialic acid receptors on guinea pig red blood cells (Table 1).However, the emerging 3C.3 and 3C-2012/13 viruses escaped inhibition by these vaccine sera (> 16-fold reduction) (Table 1).This illustrates that the few amino acid differences in the H3 proteins (e.g.residues 128, 142, and 145 using H3 structural numbering) between the emergent 3C.3 and 3C-2012/13 viruses vs the 3C.1 viruses have enabled escape from polyclonal antibody responses to natural infections or vaccinations.
The virus isolates tested in HI were sequenced and had the same HA amino acid sequences as the original swab specimens in all but two cases (Figure 2), while the NA vRNAs of 3C.3 and 3C-2012/13 viruses acquired a mutation (i.e.D151N/G) that has been shown to reduce NA activity [48], illustrating the difficulty in accurately analysing the true antigenicity of viruses circulating in humans using in vitro assays and the importance of directly sequencing the viral population in swab specimens rather than viruses isolated from eggs or tissue culture.Furthermore, the HA of many contemporary H3N2 viruses bind inefficiently to guinea pig RBCs and the cognate NA aids in binding to receptors on the RBCs.Addition of oseltamivir in the HA assay to reduce NA binding reduced the HA titre so much for these viruses that HI assays could not be performed in the presence of oseltamivir.
Examination of the HA and NA residue changes (Figure 2, Figure 4) demonstrated that a few simultaneously evolving amino acid substitutions, some causing the gain or loss of sequons for N-linked glycans, differentiated the HA molecules among strains in the 2012/2013 epidemic.

Discussion
The H3N2 subtype predominated in Houston during the 2012/13 season, as was the case throughout the US for this epidemic [10].While viruses from several known clades were co-circulating throughout the season, the majority of IAV H3s belonged to the 3C subclades, 3C.Our data demonstrate that intrasubtypic reassortment created some of the co-circulating viruses in the 2012/13 epidemic.This finding is consistent with a recent study [12] and demonstrates that intrasubtypic reassortment is another evolutionary mechanism that antigenic drift variants can employ to gain fitness advantages and spread among the viral population.In some cases, intrasubtypic reassortment events among H3N2 segments can become fixed in the population as part of periodic selective sweeps and incorporated into the trunk of the phylogeny [49,50].Intrasubtypic reassortment may also serve as a mechanism whereby drift variants can recover from a putative fitness loss (e.g.reduced receptor binding by HA) by acquiring epistatic changes through reassortment (e.g. in NA or NS) and/ or the subsequent increase in amino acid replacement rates that reassortment triggers [49,50].The NS gene segment encodes the NS1 protein, which is important in the evasion of the innate host immune response [51,52]; therefore, it is reasonable to hypothesize that the acquisition of a unique NS gene segment in most 3C-2012/13 viruses may offer a fitness advantage in the host when paired with the 3C-2012/13 HA segment.Experimental testing of these types of reassortants (e.g.3C-2012/13 constellations 9 and 10 in Figure 3B) is required to test this hypothesis and to better understand how various segments interact with drifted HA sequences to create successful H3N2 strains.
Importantly, our data show that the emerging 3C.3 and 3C-2012/13 subclades/subgroups represented the majority of the virus strains sequenced in this study, and the representative samples from these clades that were tested in HI assays reacted poorly with antisera raised against the vaccine strains used for the 2012/13 season (egg-and cell-grown A/Victoria/361/2011) and the 2013/14 season (A/Texas/50/2012) which are both in the 3C.1 subclade.While use of oseltamivir in the HA assay reduced the HA titre enough that HI assays could not be performed in the presence of oseltamivir, more recent data on vaccine effectiveness and vaccine strain updates have corroborated our findings that the 2012/13 H3N2 vaccine strain was generally poorly matched for the majority of circulating 2012/13 H3N2 strains [26,47,53].
Overall, our data demonstrate that a few simultaneously evolving amino acid substitutions played a central role in immune escape by the 2012/13 epidemic viruses.Our data indicate that the H3N2 viruses escape the human immune response through a combination of specific residue changes (e.g.N145S, H3 structural numbering) and the impact these changes have on N-linked glycosylation.Residue 145 is located on a loop of the H3 that is partly involved in receptor binding (Figure 4), and it is one of seven residues shown to play a central role in the antigenic drift of H3 viruses in humans [54].Viruses in 3C.2, 3C.3, and 3C-2012/13 have the S145 substitution, while the clade 3C.1 viruses have N145.This residue is part of an N-linked glycosylation sequon, and the N145S change is likely to increase glycosylation of N144 [55].Some 3C.2 viruses acquired an S124N substitution, which removes the N122 glycosylation site; they are antigenically similar to previously circulating viruses in clades 1, 5, and 3C.1 and are therefore likely to die out.However, the new 3C.2asubclade that emerged from 3C.2 (Figure 1A) and expanded in the 2014/15 season [47] has lost glycosylation at 144 (N144S) and added a new putative glycosylation site at N158.
The 3C.3 viruses acquired accessory changes in or near antigenic site A, including T128A, which removes the N126 glycosylation site, and R142G; these substitutions are likely to be responsible for these viruses' antigenic difference from 3C.1 and 3C.2 viruses.The alterations of N-linked glycosylation sequons among the various clades (Figure 2, Figure 4) are likely to contribute to more complex structural differences in the HA due to the gain or loss of glycan shields [56][57][58].In particular, it appears that glycans at HA residues N126 and N144 are the major differences among the 3C subclades/subgroups (Figure 2).
Furthermore, because 3C.3-like viruses and their descendants currently circulate globally (data not shown), vaccines based on A/Texas/50/2012 are likely to offer suboptimal protection.This hypothesis is supported by the moderate vaccine effectiveness for the 2012/13 season in the US (47% for IAV), where the H3N2 subtype predominated [13].The data presented here indicate that an ideal vaccine candidate would probably be derived from subclade 3C.3a or 3C2a.The WHO influenza vaccine selection committee recently recommended an A/Switzerland/9715293/2013-like strain, which belongs to the 3C.3a subgroup of the 3C.3 viruses (Figure 1A), for the 2015 southern hemisphere season [59].In addition to N145S, T128A and R142G substitutions, the 3C.3a viruses have acquired additional changes (A138S, F159S) that are likely to be antigenically important.

Conclusion
This study provides insights into the dynamics of the 2012/13 influenza epidemic by demonstrating that multiple H3N2 strains co-circulated during the epidemic, that different antigenic drift variants evolved concurrently and that intrasubtypic reassortment occurred frequently, suggesting that reassortment plays a role in the evolution of seasonal influenza viruses.The emergent 3C.3 and 3C-2012/13 viruses, which had substitutions impacting N-linked glycosylation at major antigenic sites, predominated, and we predict that they will come to define the new trunk of the phylogeny.
Finally, our data show that the accumulation of relatively few HA mutations can convey large selective advantages, sometimes through epistatic interactions, and support the new 2015 southern hemisphere vaccine strain recommendation.

Figure 1 Figure 2 5 )H3
Figure 1Phylogenetic relationships among the influenza A(H3N2) study samples, vaccine strains, and selected references inferred from maximum likelihood analyses of the HA nucleotide (A), HA protein (B), and NA nucleotide (C) sequences, Texas, 3 November 2012-8 February 2013 (n = 154) CDC: Centers for Disease Control and Prevention; HA: haemagglutinin; ML: maximum likelihood; NA: neuraminidase; WHO: World Health Organization.Bootstrap values for nodes with ≥ 70% support following 500 replicates are provided.Clades are classified following the CDC nomenclature [25] as closely as possible and are defined based on the HA nucleotide phylogeny (A).A new subgroup, designated 3C-2012/13, appears to have arisen during the 2012/13 influenza season.The nodes from which the new WHO-designated 3C.2a, 3C.3a, and 3C.3b subclades subsequently arise are marked on the phylogeny to place our 2012/13 study in the context of more recent seasons; placement was based on additional ML and neighbour-joining phylogenetic analyses using available H3 sequence data through 2014 (data not shown).For the HA protein (B) and NA nucleotide phylogenies (C), strain names are coloured by HA nucleotide clade definitions, showing that the HA protein and NA nucleotide clades largely match the nucleotide-based HA clade definitions.However, some interleaving of 3C subclades is observed in the HA protein tree, indicating that these subclades diverge largely through synonymous nucleotide mutations.The interleaving of colours in the NA nucleotide tree signifies intrasubtypic reassortment between the HA and NA gene segments.Scale bars indicate the average number of nucleotide changes per site.

Table 1
Antigenic properties of study samples measured by HI assays using ferret or sheep antisera raised against contemporary influenza A(H3N2) strains rA/South_Australia/3/2011, which has the same amino acid sequence as A/Victoria/361/2011 (cell-grown) except for one change in the signal peptide.
HA: haemagglutinin; NA: neuraminidase; N/A: not applicable.a The authors gratefully acknowledge the originating (WHO Collaborating Centre for Reference and Research on Influenza in North Melbourne, Victoria, Australia) and submitting (Centers for Disease Control and Prevention, Atlanta, Georgia, US) laboratories that contributed the celland egg-grown A/Victoria/361/2011 sequences to GISAID's EpiFlu database.b c These reassortant viruses are named for the strain from which the HA and NA were obtained; the six internal genes match A/Puerto Rico/8/1934(H1N1), A/New York/238/2005(H3N2) or A/New York/1682/2009(H1N1pdm09) sequences.