The evolution of influenza hemagglutinin: structural numbering and receptor binding sites of influenza viruses capable of infecting humans

Introduction

Influenza viruses are highly contagious respiratory pathogens that cause seasonal epidemics and occasional pandemics, resulting in significant global morbidity and mortality (1, 2). Influenza viruses belong to the Orthomyxoviridae family and possess a segmented, negative-sense RNA genome enclosed within a lipid envelope (3, 4). This segmented RNA genome, together with an error-prone RNA polymerase that lacks proofreading ability (5), drives frequent point mutations (antigenic drift) and enables genetic reassortment (antigenic shift) during co-infection with different strains. This exchange of genomic segments can give rise to novel, highly pathogenic variants capable of causing global pandemics (6). These evolutionary pressures are most pronounced in the major surface glycoproteins, hemagglutinin (HA) and neuraminidase (NA). These proteins can accommodate extensive amino acid substitutions without appreciable loss of viral fitness, thus allowing influenza viruses to evade existing antibody responses (7, 8).

Influenza viruses are divided into types A, B, C, and D based on antigenic differences. Of these types, only influenza A viruses (IAVs) and influenza B viruses (IBVs) infect humans. IAVs are further divided into subtypes, whereas IBVs are further divided into lineages, each of which are delineated based on differences in their HA and NA glycoproteins. There are 18 known HA subtypes (H1 to H18) and 11 known NA subtypes (N1 to N11) (9), and the specific combination of these HA and NA glycoproteins dictate the ability of a virus to infect different hosts and the severity of the illness it causes.

The HA protein is primarily responsible for viral attachment and entry into host cells, and as such, HA is the primary target of neutralizing antibodies. And in this review, HA will be the focus as we investigate the differences in receptor binding between different influenza subtypes, the evolution of vaccine strains over time, and various other nuances that contribute toward delineating influenza viruses into different subtypes.

Hemagglutinin

HA binds to sialic acid (SA) receptors on the surface of host epithelial cells, initiating virion uptake through receptor-mediated endocytosis, typically via a clathrin-dependent pathway (3, 10). The virus then traffics to the endosome where the lower pH facilitates the fusion of the viral envelope and endosomal membrane, as well as a conformational change in HA (4, 10–13). HA forms a trimer on the virion membrane of three identical monomers synthesized as a single polypeptide (HA0) of around 550 amino acids (14). During the conformational change, this HA0 precursor is subsequently cleaved into two polypeptide chains, HA1 and HA2, by a host cell protease (14).

Each monomeric HA glycoprotein, which is structurally similar between isolates of IAV and IBV, is made up of a globular head and fibrous tail (14) (Figure 1A). The head is composed of three identical domains whose sequences are entirely HA1 residues and is described as a β-structure with an eight-stranded Swiss-roll type structural framework that supports the receptor-binding domain (RBD) (3, 14). The tail, mainly referred to as the stem or stalk region, is made up of a triple-stranded coiled-coil of α-helices derived primarily from HA2 residues (3, 14). Buried in the stem region is a hydrophobic fusion peptide (14). Cleavage into HA1 and HA2 exposes the fusion peptide in a “spring-loaded mechanism” (15) for insertion into the endosomal membrane that in turn opens a pore through which viral ribonucleoproteins can be released into the host cell cytoplasm (3, 4, 10). This cleavage of HA is necessary for membrane fusion and subsequent virus infectivity (10–12, 16). For most influenza viruses, cleavage occurs at a monobasic site in airway epithelial cells. This leads to localized infection of the respiratory and/or intestinal tracts and, while severe infections can occur, usually results in asymptomatic or mild infection. Some influenza viruses, such as highly pathogenic avian influenza (HPAI) viruses, can acquire multiple basic amino acids at the HA cleavage site (a polybasic cleavage site) which allows the HA protein to be activated by a wider range of proteases found in the host’s tissues (17–19). This systemic activation allows the virus to replicate throughout the body and cause more severe disease and higher mortality rates.

HA gene sequencing numbering

The IAV HA glycoprotein is composed of between 562 and 568 amino acids, while the IBV HA has between 582–585 amino acids (Table 1). The first high resolution crystal structure of influenza HA using the 1968 pandemic H3N2 virus (20, 21) established a precedent for a structure-based numbering scheme. More recently, Burke and Smith analyzed HA sequences across the 18 IAV HA subtypes to define residues that were structurally and functionally equivalent and then aligned them to establish a recommended numbering system (22). This review section takes a similar approach with a focus on IAV HA proteins from viruses that infect humans and furthermore adds IBV HA proteins.

Table 1. Influenza A and B hemagglutinin sequence numbering.

AA’s, amino acids; Seas. H1N1, seasonal H1N1; B/VIC, B/Victoria-lineage; B/YAM, B/Yamagata-lineage.

Each HA sequence begins with a 15–18 amino acid signal peptide (Figure 1C) that is essential for targeting HA to the endoplasmic reticulum (ER) during translation (20, 21, 23). A signal peptidase enzyme in the ER cleaves the signal peptide off for HA to undergo further processing and subsequent trafficking to the cell surface. As such, mature proteins do not include this stretch of amino acids at the N-terminus of the HA reading frame (24), and thus it is most commonly omitted when numbering HA residues, as established by Wilson, Skehel, and Wiley (20, 21). Immune pressure does not act on the signal peptide since it is not part of the mature HA antigen. However, signal peptides are not necessarily sequence conserved. While there may be sequence identity between isolates, such as 100% identity for the five representative IBVs in Table 1, the sequences between different strains often differ greatly. For example, the representative strains for seasonal H1N1 (NC/99) and pdm09 H1N1 (CA/09) share ~56% identity and only 18-38% identity with other subtypes. Instead, signal peptides are functionally conserved and retain a hydrophobic core (H-region), a workable cleavage site (C-region), and a short, positively charged N-terminus (N-region) with basic amino acids (25, 26).

The HA1 region forms the globular head which is structurally flexible and modular, made up of surface-exposed loops and antigenic sites surrounding the RBD. This is the major target of antibody pressure and subsequent viral mutations to escape neutralization, and as such, there is a lot of diversity between subtypes. HA proteins are constantly mutating and acquiring amino acid substitutions and even residue insertions or deletions. Despite this, there are still conserved regions between different strains, and even between different subtypes. Using the “H3 numbering” system, influenza A viruses (Figure 2) can be aligned fairly well (22). The representative H3N2 HA1 sequences have 79.3% identity between one another, but only 33.7 – 38.5% identity with the other IAV HA1 sequences, and <23% identity to the IBV HA1 sequences (Supplementary Table 1). The representative H7 HA1 sequences are similar and share 95% identity between one another, but <39% identity with the other IAV HA1 sequences (Supplementary Table 1). The H1N1 HA1 sequences (72.5%) and H5N1 HA1 sequences (86.6%) have high sequence identity within each subtype, but there is also relatively high sequence identity between these two subtypes as well as with H2N2 HA1, ranging between 52.3 – 67.3% identity (Supplementary Table 1). IBV HA1 sequences are much more conserved (88.2 – 94.2%) between each other but share very limited (<25%) identity with IAVs (Supplementary Table 1).

The HA2 region is more functionally and structurally constrained since it contains the fusion peptide that must insert into the endosomal membrane for infectivity. Even small substitutions can abolish fusion activity or reduce infectivity (27–29). Therefore, the HA2 region is more conserved between different subtypes, especially the initial 11 residues that are fairly conserved between different IAVs (GLFGAIAGFIE) and IBVs (GFFGAIAGFLE). IAVs can be divided into two groups based on their HA stem similarity: group 1 (H1, H2, H5, H6, H8, H9, H11, H12, H13, H16, H17, H18) and group 2 (H3, H4, H7, H10, H14, H15) HAs (30–32). As such, the HA2 sequences for these representative strains (Supplementary Table 2) have higher sequence identity, for example, between H3 and H7 (~64%) than between H3 and H1 (~52%). Like the HA1 sequences, IBV HA2 sequences are highly conserved (96.9 – 99.6%) between each other and share very limited (<40%) identity with IAVs (Supplementary Table 2).

Receptor binding domain

The RBD (Figure 1B) is a shallow pocket located at the membrane-distal tip of the HA1 subunit that attaches the virus HA to host cells through recognition of terminal SA residues on glycoproteins and glycolipids along the respiratory epithelium. SA residues are attached to carbohydrate chains on different linkages, primarily types α2,3- or α2,6-linkage to a galactose residue (4, 33, 34). These different linkages form unique steric configurations that are recognized differentially by viral HA proteins. In general, human influenza viruses preferentially bind to α2,6-SA and avian influenza viruses to α2,3-SA (35–37). There is a differential expression of each of these sialic acids between humans and animals and throughout different locations within each host (35–37). In humans, α2,6-SA are normally expressed on the mucosal surface of the upper respiratory tract, whereas α2,6- and α2,3-SA are expressed in the lower respiratory tract, so viruses that are better able to bind α2,3-SA might be capable of causing more severe disease associated with lower respiratory tract infection (35, 36, 38).

Binding is driven by a network of hydrogen bonds and van der Waals contacts between amino acids lining the pocket and the SA moiety (39, 40). The RBD itself is functionally constrained and relatively conserved, but significant changes can reduce receptor affinity or impair fusion competence (41). There are however several immunodominant antigenic regions flanking the RBD that are frequent targets of antibody pressure. Antibodies directed at these sites can block attachment, destabilize HA, or sterically interfere with RBD access (42, 43). To escape such immunity, circulating influenza strains accumulate point mutations (antigenic drift) that often involve changes in influenza virus HA glycosylation patterns that shield or expose these sites (44). Glycosylation of HA (and NA) occurs through attachment of a carbohydrate chain (oligosaccharide, or glycan) to the amide nitrogen of an asparagine (Asn or N) side chain, which results in an N-linked glycosylation (N-Gly) that can be identified by the sequon N-X-S/T, where X may represent any amino acid except proline (44). The presence of this sequence motif does not guarantee the site is glycosylated, but it is a good predictor and therefore useful when analyzing the potential impact of certain mutations due to antigenic drift. N-Gly can alter viral affinity for SA-containing receptors near the RBD, it can modulate virus pathogenicity when near the HA cleavage site, and it has been shown to be important for protein folding, trafficking, and pH stability when in the stem region (44).

Mutation to the HA protein sequence, through antigenic drift, can also alter receptor binding in other ways unrelated to N-Gly. Zoonosis into humans required early influenza viruses to adapt for more efficient binding to α2,6-SA. H1N1 strains showed two vital mutations, E190D and G225D (186 and 222 in H1 numbering), that increased affinity for α2,6-SA (45). In the same manner, H2N2 strains required a single mutation (Q226L) while early H3N2 strains necessitated two (Q226L and G228S) (45). Glutamine (226Q) has a relatively long side chain that is polar and forms a narrower pocket favoring α2,3-SA binding. Replacing it with a leucine that is hydrophobic and more compact increases the flexibility of the RBD for the accommodation of α2,6-SA binding (46). In the early 1990s, a mutation at H3N2 residue 190 (E190D) was shown to be responsible for the loss of the ability to bind chicken red blood cells (47). The longer Glu190 sidechain with an additional methylene group may extend into the RBD and interact directly with bound sialic acid, whereas Asp190 may be too short to extend into the pocket and thus could have a weaker receptor interaction (48). A mutation at H3N2 residue 156 (K/Q156H) in the early 2000s was shown to shift H3N2 even more toward preferring α2,6-SA (49). A more consequential shift occurred around 2013 with changes at residue 225 (G/N225D), where introduction of a negatively charged side chain stabilized hydrogen-bonding networks within the receptor-binding pocket and strengthened receptor engagement to increase overall binding avidity (50). These combined changes to H3N2 viruses influenced experimental propagation and characterization, favoring growth in MDCK-SIAT1 cells and improving performance in HAI assays using guinea pig RBCs, both of which express more α2,6-SA relative to α2,3-SA (50).

An outbreak of an H5N1 clade 2.3.4.4b virus, A/Texas/37/2024 (TX/24), in dairy cattle (and subsequent transmission into poultry) led to a zoonotic event of at least 70 reported cases in March 2024 to May 2025 (51). The major molecular marker associated with mammalian adaptation of this TX/24 virus was in the polymerase basic 2 (PB2) gene (E627K) (51). An H5 HA Q226L mutation, as observed in H2 and H3, could exhibit a much greater degree of mammalian adaptation. This very scenario was demonstrated in a recent structural study (52), and while there are other factors to consider transmissibility (e.g., HA fusion pH and NA), it highlights the ever-present pandemic potential of H5 influenza virus. Only low pathogenic avian influenza (LPAI) H7N9 variants have been detected with the HA Q226L mutation (53, 54). Despite this, a wave of HPAI H7N9 virus was able to cause numerous infections in humans, among which 16 were fatal (53). Acquisition of a polybasic cleavage site in these H7N9 viruses (53) along with the PB2-E627K mutation (55) demonstrated enhanced disease in mammalian models. Further adaptation (e.g., increased respiratory transmission) could potentially cause an H7N9 pandemic, so continued surveillance and monitoring of genetic changes remain critical. H7N9 viruses harboring the mutations A135T (HA antigenic site A) and A160T (site B), both of which introduce N-Gly sites, along with a third substitution, L226Q (217 H7 numbering), were isolated from human and environmental samples following the last reported human case in Gansu, China (56). The acquisition of these N-Gly sites was shown to reduce binding affinity to both human and avian sialic acid receptor analogs, likely due to glycan shielding effects (56). Notably, glycosylation at residues 158–160 has also been associated with increased virulence in terrestrial poultry (57), potentially decreasing HA (H5 and H7) dependence on NA for efficient viral release (58).

IAV RBD and antigenic regions

The IAV RBD is formed by the 130-loop, the 150-loop, the 190-helix, and the 220-loop, with a set of conserved residues (Tyr98, Trp153, His183, and Tyr195) forming the base of the RBD (21, 35) (Figure 1B). These conserved residues anchor the terminal sialic acid (Neu5Ac) in the pocket, while other residues contact adjacent glycan sugars (59). The 130-loop and 220-loop are β-hairpin loops which form two sides of the RBD (59). The 130-loop lies on one flank of the pocket, while the 220-loop forms the opposite rim. Key residues at the base of these loops help determine host specificity (59). The 190-helix, a short α-helix that forms the back wall of the RBD, contains Glu/Asp190 and surrounding residues that form a hydrogen-bond with SA receptors (59). Mutation of residue 190 can alter receptor binding affinity dramatically. The 150-loop is a flexible loop that sits in front of the pocket. Some subtypes (e.g. H3, H5, H7) have a two-residue insertion in this HA loop that partially occludes the pocket and restricts HA binding (60).

Five major antigenic sites have been identified for H3, named A–E (20), with site E defined subsequent to A–D (40, 61, 62) (Figure 3A). Site A is predominantly made up of residues within the 130- and 140-loop near the membrane-distal top of the head, site B is adjacent on the HA head and overlaps the 150-loop and 190-helix, and site C overlaps with the vestigial esterase and fusion domains at the base of the head made up of residues around sites 50, 275, and 300 (Figure 4). Site D is made up of the 220-loop as well as residues within the 120-loop and 160-helix, while site E sits opposite site A and occupies the vestigial esterase subdomain (around site 60 and 76-93). Overall, sites A and B exhibit the highest mutational drift whereas C–E are more conserved, with glycan additions often shielding them. Recent evolution seems to have centered more with site B (63). Frequent site B mutations in the 150-loop (sites 155-159) and 190-helix (residues 186-196) occurred from 1968 to the early 2000s (Figure 4). More recently, these mutations are less numerous from year-to-year; however, many important sites have mutated, such as H156S, Y159N, K160I, D190N, I192F, F193S, Y195F, and A198S (Figure 4).

Five classic antigenic sites for H1 were defined shortly after, denoted as Sa, Sb, Ca1, Ca2, and Cb (64–69) (Figure 3B). Sa and Sb, “strain-specific” antigenic sites roughly equivalent to H3 sites A and B (64), sit on the distal tip of each HA1 monomer flanking the RBD and are typically most immunodominant in humans (68). Sa mostly includes the 130-loop whereas Sb includes the 190-helix region and part of the 220-loop (Figure 5) (65, 68). Virus-host attachment occurs at the RBD located between Sa, Sb, and Ca2 (68). Sa and Sb rapidly accumulate drift mutations under immune pressure, but Ca is generally more conserved (68). Ca and Cb, “cross-reactive” epitopes that are more equivalent to H3 site D (64), lie closer to the HA1–HA2 interface (Figure 3B). Ca was further divided into Ca1 (includes residues near 170, 205, and 246) and Ca2 (part of the 140-loop and 221-222) (Figure 5) as this region is formed in a cleft between two monomers (70). Cb, the equivalent to H3 site E residing within the vestigial esterase domain near the base of the head, is a small region made up of a flexible loop capable of adopting variable conformations (66). H1 does not have an obvious equivalent for H3 site C, though some may group site C and Ca together (71, 72) due to the overlapping of some residues. However, the loops corresponding to H3 site C are shorter and less protruding for H1, so antibodies have less access.

There are six HA antigenic sites for H2Nx influenza viruses (Figure 3C), with four (I-A, I-B, I-C, I-D) recognized by group I mAbs and two (II-A and II-B) recognized by group II mAbs (73). Five of these are nearly analogous to H3: sites I–A, I-B and I-C form a contiguous antigenic area in the HA head region corresponding to antigenic sites A, B and D on the H3 HA, site I-D corresponds to H3 site E in the HA1–HA2 interface, and site II-B corresponds to H3 site C in the HA stem region (73). The sixth site, II-A, is a highly conserved neutralizing epitope in the HA stem region with cross-reactivity to all H2N2 viruses, but no reactivity to fellow group 1 H1N1 viruses (73). One residue here, HA2 Phe45 (HA site 374 in Figure 2) that is present in H2 but no other subtypes, is responsible for sterically clashing with all currently characterized HA stem-binding antibodies (74, 75). A similar neutralizing epitope has not been identified in the HA stem region for either H1 or H3.

H5 (Figure 3D) and H7 (Figure 3E) HA antigenic sites are largely similar to H3 and utilize the same naming scheme of A–E (76). Likewise, sites A (residues 121-146) and B (residues 155-160, 186-198) are the major targets of immune responses, with mutations at key residues in sites A (77, 78) and B (79, 193) frequently reported in escape mutants for H5 (57, 80–83) and H7 (84, 85) viruses. HA antigenic site C does not feature RBD structural components or residues crucial for receptor specificity, so less selective pressure is found here for natural escape mutants. As such, this site is less characterized, especially for H7 viruses (Figure 2). Nevertheless, there has been increased variation in antigenic site C for North American H7 viruses compared to Eurasian H7 strains (84). Antigenic site D is located within the RBD and features key amino acids with a role in receptor specificity, such as the 220-loop which includes residues 225 and 226 that switches receptor specificity between α2,3-SA and α2,6-SA (52–54, 56). Antigenic site E is found within the vestigial esterase subdomain, and it includes various residues that can affect antibody recognition. For H5 HA, site E includes residues such as 81 and 90 which were influential in the antigenic drift of subclade 2.3.4.4 that led to the eventual TX/24 outbreak (Figure 2).

IBV RBD and antigenic regions

The RBD of IBV has a similar structure as IAV, with IBV HA binding sialylated glycans using a shallow receptor-binding pocket framed by several loop regions: 190-helix at the top, the 140-loop as the right edge, and the 240-loop as the left edge (86). Subsequent publications and additional crystal structures have further clarified the major antigenic epitopes of the IBV HA to include the 120-loop and surrounding regions, the 150-loop, the 160-loop, and the 190-helix region which includes the nearby 240-loop (87, 88), as shown in Figure 6A. Three of the conserved residues forming the base of the IBV HA RBD are similar to that of IAV HA (Trp158, His191, and Tyr202), but a phenylalanine (Phe95) is found in place of tyrosine (Tyr98) observed in IAV HA proteins (86). Of note, replacement of Phe95 with a tyrosine residue increased receptor binding levels comparable to IAV HA (89, 90). Tyr98 stabilizes the base of the RBD via hydrogen bonds to enable high binding affinities to SA, whereas Phe95 has three fewer hydrogen bonds that cause lower binding affinity to SA (86, 90, 91). Because of this, or perhaps in addition to this, the receptor binding pocket appears deeper for IBV HA proteins (92). For the hemagglutinin-inhibition (HAI) assay, the occluded RBD of the IBV HA protein is first exposed by treating the virus with diethyl ether (93–96). The treatment increases the low binding affinity and sensitivity of the assay for more reliable detection of antigen-specific serum antibody (95–97). There are some concerns regarding reduced specificity and reliability of HAI with ether-treated viruses (98), with some research groups preferring to use neutralization assays or single radial hemolysis (SRH) for the detection of anti-IBV HA antibodies (99, 100). Nevertheless, ether-treatment of IBV HA proteins remains the best option for the higher throughput HAI assay and continues to be recommended by the Influenza Division of the CDC (101).

The 120-loop (residues 111-137) and surrounding area (residues 47-60, 71-75, 182-184, and 286-288) is a major antigenic site that can accumulate immune-driven changes (87, 88). Early structural mapping identified this loop, together with the 150- and 160-loops and 190-helix, as forming a contiguous antigenic surface (91). Recent studies in field isolates confirm active drift in the 120-loop region, with single amino acid changes driving the emergence of new antigenic variants (102). These can be viewed by comparing the first structure of B/Hong Kong/8/1973 (B/HK/73) (86) HA (Figure 6B) to a more recent B/Maryland/15/2016 (B/MD/16) (103) HA (Figure 6C). Various mutations in this region have altered HA receptor binding geometry or masked antibody epitopes. For example, selective pressure in older human populations has been shown to elicit cross-reactive antibodies targeting the 120-loop region, and escape mutants (e.g. affecting residue HA1 N/K75) can arise under monoclonal antibody pressure (104, 105). B/HK/73 contains a unique N-Gly within the RBD at residue 125 that is not observed in most other IBV isolates (Figures 6, 7). Consequently, comparative modeling of the 120-loop using this strain has limitations, as some structural observations may reflect glycan-specific effects rather than conserved features of IBV HA. Regardless, the 120-loop has consistently remained a major site of antigenic pressure and immune-driven evolution.

The 150-loop (residues 139-151) lies adjacent to the RBD and is also antigenically important. Comparative antigenic analyses have shown that substitutions at positions 148–150 in this loop distinguish modern IBV lineages. Notably, amino acid changes at HA1 positions 148, 149, and 150 are principal determinants of major antigenic shifts between the Victoria and Yamagata lineages (106). These sites flank the RBD and can influence antibody recognition without drastically altering receptor contacts. The 150-loop also carries a conserved glycosylation motif at Asn145 (Figure 7). Both Victoria and Yamagata HA possess an N-Gly at residues 145-147 (at the base of the 150-loop), but this glycan is distal from the RBD and thus has only a modest direct effect on receptor binding (104). However, loss or modification of this glycan (e.g. by host adaptation) can modulate antigenicity. In summary, the 150-loop contributes to lineage-specific antigenic profiles via surface substitutions near positions 148–150 (106) and is partly shielded by a distal glycan at Asn145 (104).

The 190-helix (residues 197-210) and adjacent 240-loop (residues 233-243) form the floor and side of the HA receptor-binding pocket (91). A key feature here is an egg-adapted mutation found at residue 197 (196 in other Yamagata-lineage numbering) withing the 190-helix. Loss of this Asn197 glycan expands the HA binding capability to acquire the ability to bind both a2,6- and a2,3-linked sialosides (91). This glycan is universally conserved in Yamagata lineage HA proteins, but presence varies in Victoria-lineage HA proteins (91). In human infections, strains lacking Asn197 often bind more broadly to avian-type receptors, consistent with this glycan “shield” normally limiting receptor access. The location of the D/N197 can be visualized in Figures 6B, C. Victoria-lineage HA proteins also acquired an N-Gly through a A235T mutation (at the tip of the 240-loop), but this glycan lies some distance from the pocket and thus has only a modest impact on binding (91). Altogether, the 190-helix and 240-loop provide both carbohydrate-shielding and the structural framework for receptor binding (91).

Many residues within the 120-loop, 150-loop, 190-helix, and 240-loop lie directly at the binding site, and mutations to these regions can largely reshape the receptor pocket through changes to hydrogen bonding and other contacts (87, 88). The 160-loop, on the other hand, is a hypervariable antigenic site protruding at the tip of the HA head (87, 88). This is the only IBV HA region to undergo frequent insertions or deletions (indels) (87, 88). These indels, typically 1–3 residues around HA1 positions 162–165 (Figures 6, 7), markedly extend the HA 160-loop in modern strains (e.g. “NDK/NDN” in Victoria-like proteins vs. “DN” in Yamagata-like proteins) (87). Overall, indels and point mutations in each of these regions can alter pocket geometry or glycosylation to contribute to antigenic drift and even modulate receptor binding and host tropism (86, 104).

The evolution of seasonal influenza vaccination in humans

The influenza pandemic of 1918 was one of the deadliest pandemics, causing around 50 million deaths worldwide (107). The true origins of influenza virus prior to this are not fully known. The 1889 “Russian Flu” pandemic is believed to have been caused by an H3N8 influenza virus (107); however, limited historical data leave its origins uncertain, and recent analyses have even proposed that it may have been driven by a coronavirus instead (108).

IAV was first isolated in 1930 from pigs by Richard Shope, three years before it was first isolated from humans (109) and a decade before IBV was identified (110) (Figure 8). IBV was recognized as a distinct influenza type after ferrets inoculated with throat washings from a sick child developed typical influenza symptoms, yet their serum antibodies failed to neutralize the contemporary IAV strain, A/Puerto Rico/8/1934 (PR8) (110). Thomas Francis, Jr. subsequently proposed that this represented a newly identified influenza type, designated B/Lee/1940 after the patient from whom it was isolated (110). This led to the formulation of a bivalent influenza vaccine in 1942 composed of one strain each of IAV and IBV (111). In 1947, after only five years of using a bivalent influenza vaccine formulated with one IAV strain and one IBV strain, it was no longer effective at preventing influenza virus infections and associated disease (111). A newly circulating IAV strain led to the discovery of antigenic drift and the notion that influenza viruses could mutate and ultimately bypass the immunity provided from previous vaccinations or infections (111). The IAV strain was updated as a result (Figure 8), but then in 1957, another new strain emerged that was similar but had two surface antigens which differed greatly (111). This was the result of genomic reassortment in which the human influenza virus and an avian influenza virus combined in an intermediate host to create a new influenza variant: an antigenic shift, or major change from the viruses circulating at the time (111–113). The virus variant had three gene segments of avian virus origin, including the major surface proteins, HA and NA (113, 114). The bivalent vaccine component for IAV changed as a result, as did the naming of influenza virus strains that resulted in a nomenclature based upon the HA and NA surface antigens (111). The 1957 virus, which shared five gene segments with the PR8-like viruses, was designated as H2N2, whereas the PR8-like viruses were named H1N1. In 1968, another pandemic occurred (Figure 8) that saw the replacement of H2N2 with a H3N2 virus. A decade later, a new virus appeared, or rather reappeared as it was analogous to the H1N1 virus that seemingly disappeared when H2N2 emerged (111). Instead of replacing H3N2 strains, the H1N1 viruses began to co-circulate with H3N2 strains, and people could be infected with one subtype or the other (or both) depending on which subtype predominated in a particular influenza season (Figure 8). This led to the emergence of a trivalent vaccine that consisted of H1N1, H3N2, and IBV, a formulation that would remain for 40 years until replaced with a quadrivalent version containing representative strains for both the B/Yamagata and B/Victoria lineages (115).

Currently, seasonal influenza virus subtypes that circulate in humans are H1N1, H3N2, or IBV. Historically, other combinations have infected humans also, such as H2N2, H3N8, H5N1, and H7N9 (Figure 8). Avian influenza viruses of the H5 and H7 subtypes have infected humans, but these viruses rarely transmit between people; although lately, there has been an emphasis on H5 as more infections have been recorded (116). Seasonal influenza virus primarily circulates in the fall and winter. For the Northern Hemisphere, this period (or influenza season) typically occurs from October to April, whereas the influenza season in the Southern Hemisphere goes from April to September (117). In preparation for these seasons, influenza surveillance is conducted through various collaborative organizations, such as the World Health Organization (WHO) Global Influenza Surveillance and Response System (GISRS) (118) and the Influenza Division at the U.S. Centers for Disease Control (CDC) (119). Each publish tracking tools available to the public: FluNet through the WHO, and FluView through the CDC (120). Based upon this surveillance, recommendations for vaccine strains are made for each hemisphere. Because influenza viruses are constantly undergoing antigenic drift, vaccine formulations frequently change as new strain recommendations are made prior to each influenza season (111).

H1N1 vaccine strain changes

H1N1 caused the devastating 1918 “Spanish flu” pandemic and continued to circulate globally until 1957, when it was supplanted by the emergence of H2N2 (Figure 8). The H1N1 virus re-emerged in 1977 under circumstances suggestive of a laboratory release (121), prompting its re-inclusion in seasonal influenza vaccines. Thereafter, continued circulation and antigenic drift of H1N1 necessitated nine vaccine strain updates prior to the 2009 “swine flu” pandemic (Figure 8). During the 2009 H1N1 pandemic, protection from severe disease was observed in the elderly. This led to the hypothesis and confirmation that exposure to 1918-like viruses conferred cross-protective immune responses (122–125). The HA from the 1918 and 2009 viruses share very similar antigenic sites which were distinct from H1N1 viruses circulating prior to 2009 (126). The 2009 pandemic marked a major evolutionary transition for H1N1, after which post-2009 viruses were classified as “pandemic H1N1” (pdm09) and pre-2009 viruses were designated as “seasonal H1N1” strains. Phylogenically, the pdm09 strains are grouped close to the 1918 virus, whereas the seasonal strains are separated. Major differences between seasonal and pdm09 strains are observed in the Sa site (red), Sb site (blue), Ca1 site (magenta), Ca2 site (green), Cb site (orange), residues 240-260, and the fusion domain (residues 275-322) (Figure 5). There also seems to be a clear divide between seasonal strains prior to the mid-1980s, and those afterward. The latter seasonal strains (1986–2007) have a glycosylation motif at residues 54–56 that is not present in earlier strains; however, many earlier strains (1957-1986) have a glycosylation at residues 155–157 in antigenic region Sa that is absent in later seasonal strains and pdm09 strains (Figure 5). Most seasonal H1 strains have a glycosylation site in the 120-loop around antigenic site Sa, but pdm09 strains and the 1918–1943 strains do not. Instead, pdm09 strains have a potential glycosylation at residues 276–278 that is absent in earlier strains. And there are many Sb site differences between the earlier seasonal strains (186E, 189K, 190T, 193R, 194K) and latter seasonal strains (186G/R, 189R, 190A, 193H, 194T), each of which greatly contrast to pdm09 viruses (Figure 5).

There were 14 vaccine strain selections from the initial H1N1 vaccine component (PR8) until A/California/07/2009 (CA/09); and afterward, the H1N1 vaccine component was updated six times (Figure 8). CA/09 was the H1N1 recommendation for seven consecutive influenza seasons from 2010-2011 (127) until 2016-2017 (128). Reduced effectiveness against circulating clade 6B.1 and 6B.2 viruses prompted a change for the 2017–2018 season to a clade 6B.1 strain, A/Michigan/45/2015 (MI/15) (129). MI/15 differed from CA/09 at antigenic sites Sa (S162N, K163Q), Sb (S185T), and Ca1 (S203T), and had various other mutations (P83S, S84N, D97N, I216T, A256T, K283E, I321V, E374K, S451N, E499K) that would be clade-defining (Figure 5). Of note, S162N introduced a potential glycosylation site to the Sa region near the location of a previous glycan that was present in seasonal H1N1 strains. H1N1 infections were very mild the first season with MI/15 as the H1N1 component, but during the second season, cases were more rampant as H1N1 accounted for more than 50% of influenza positive tests reported in the United States (130). To better match the drifting H1N1 viruses, the H1N1 vaccine strain was changed in the 2019–2020 season to a clade 6B.1A virus, A/Brisbane/02/2018 (BR/18) (131). There was high sequence similarity between the 6B.1 and 6B.1A clades, but a few key mutations in antigenic site Cb (S74R), the loss of a proline (P282A), and an egg-adapted mutation (Q223R) collectively reduced the effectiveness enough to warrant a strain change (Figure 5).

As clade 5 strains began to dominate circulation, H1N1 infections continued to account for more than half of all influenza positive tests reported in the United States (130). As such, the 2020–2021 H1N1 vaccine strain included a clade 6B.1A.5a.1 (5a.1) virus, A/Guangdong-Maonan/SWL1536/2019 (GD/19) (132). The main differences between BR/18 and GD/19 were in HA antigenic site Sb (T185I, D187A, Q189E), but there were important mutations (R45G, N129D, N260D, A282P) in other areas of the globular head as well (Figure 5). During this time, however, the COVID-19 pandemic was peaking, and increased health measures such as masking and social distancing helped to reduce the spread of influenza (133). Despite a reduction in transmission, influenza drift continued. The H1N1 vaccine strain recommendation for the following 2021–2022 season (and 2022–2023 season) was a clade 6B.1A.5a.2 (5a.2) virus, A/Victoria/2570/2019 (VI/19) (134). This clade 5a.2 virus differed from clade 5a.1 through a few mutations in antigenic sites Sa (N156K, L161I) and Sb (A187D, E189Q), as well as the 130-loop (K130N) of the RBD (Figure 5). Towards the end of 2023, clade 5a.2a.1 strains made up a large portion of circulating H1N1 viruses. As such, the vaccine strain was changed for the 2023–2024 season to a clade 5a.2a.1 (subclade D) virus, A/Victoria/4897/2022 (VI/22) (135). This VI/22 strain differed from VI/19 primarily at antigenic sites Ca2 (P137S, K142R) and Sb (A186T, Q189E), but there were many additional drifted residues in other areas of the HA head (Figure 5). A large percentage of clade 5a.2a strains were still co-circulating with 5a.2a.1, but no strain changes were made for H1N1 until recently with the upcoming 2026–2027 season which features another clade 5a.2a.1 (subclade D.3.1) virus, A/Missouri/11/2025 (MO/25) (136). The only difference in the HA head is a seemingly insignificant T120A change. All other differences (I372V, I460T, V520A) are in the stem (Figure 5).

H3N2 vaccine strain changes

Multiple reassortment events between human H2N2 and avian H3 viruses led to the “Hong Kong Flu” Pandemic of 1968 (Figure 8) and the emergence of H3N2 viruses. Influenza H3N2 viruses at this time, such as A/Hong Kong/1/1968 (HK/68) and A/Aichi/2/1968 (AI/68), had HA and PB1 gene segments of avian origin while all other genes were inherited from human H2N2 viruses (114, 137, 138). Since this onset, the vaccine strain selection for H3N2 has changed 35 times (Figure 8). The antigenic difference from one change to the next can vary greatly. Antibodies from preceding strains may offer some protection or minimal-to-no protection (139). From 1968 to 2011, there were 25 different H3N2 strain selections that fell into 14 antigenic clusters (Table 2) (140–142). The last of these clusters, represented by A/Victoria/361/2011 (VI/11), became known as clade 3C. Major mutations (S45N, T48I, N312S) for this clade occurred in HA antigenic site C. Clade 3C viruses would continuously drift and go on to further expand, beginning with clade 3C.1 and the A/Texas/50/2012 (TX/12) strain in the 2014–2015 influenza season. A dominant feature of this change was a glycosylation loss in site A following a T128N mutation to TX/12 (Figure 4). Towards the end of the season, however, new viruses from clades 3C.2a and 3C.3a were predominantly circulating. This alteration from the vaccine strain selection rendered the seasonal influenza vaccine less effective (143) and dropped the seasonal flu vaccine effectiveness rate to the lowest recorded level (144).

Table 2. Antigenic clusters, clades, and subclades of H3N2 vaccine strains.

*2026 southern hemisphere recommendation, used to show the divergence into “subclade K”.

A clade 3C.3a strain, A/Switzerland/9715293/2013 (SW/13), became the vaccine recommendation for the 2015–2016 season (145). SW/13 mainly differed in antigenic site A with mutations such as T128A, R142G, and N145S (Figure 4). However, clade 3C.2a viruses exhibited a clear fitness advantage over 3C.3a viruses and became dominant toward the end of the season. This shift prompted another vaccine update for the following season (2016-2017), incorporating the clade 3C.2a strain A/Hong Kong/4801/2014 (HK/14) (128). Viruses belonging to this clade had a reversion of A128T to reestablish a potential glycosylation site. And notably, 3C.2a viruses such as HK/14 also harbored the K160T substitution, which introduced a potential glycosylation site in antigenic site B (Figure 4). Incorporation of this glycosylation site led to a change in receptor binding that resulted in an altered hemagglutination ability of turkey red blood cells (143) and poor growth in eggs (77, 146). In addition to poor growth in eggs, HK/14 virus successfully propagated in eggs had egg-adapted changes that reverted this mutation at residue 160 (77, 146). Since most vaccines are produced using egg-propagated viruses, this change in the vaccine strain led to reduced effectiveness against circulating viruses. This continued into the next H3N2 strain selection during the 2018–2019 season which saw a switch to a clade 3C.2a1 virus, A/Singapore/INFIMH-16-0019/2016 (SG/16) (78). This clade had other associated mutations such as N121K, N171K, I406V, G479E, G484E (Figure 4). During the late part of the season, clade 3C.3a1 viruses, a branch from the previous SW/13-like strains, began co-circulating. Because the diversity of these strains was so great compared to 3C.2a1, it led to a delay in the announcement of the H3N2 strain along with the 2019–2020 recommendations (147). Ultimately, the 3C.3a1 strain A/Kansas/14/2017 (KS/17) was chosen as the recommendation. This decision was controversial since 3C.2a1 viruses were still circulating at higher levels in the world outside the United States (148). Some of the major differences between KS/17 and the 3C.2a1 viruses were T128A which eliminated an N-Gly site in HA antigenic site A, A138S and S144K in HA antigenic site A, and T160K in HA antigenic site B that eliminated another N-Gly site (Figure 4). Altogether, this resulted in the lowest seasonal flu vaccine effectiveness rate since the 2014–2015 season (144).

Clade 3C.2a viruses continued to drift and dominate circulation following the KS/17 mismatch. For the 2020–2021 season, the H3N2 vaccine strain was updated to a clade 3C.2a1b.1b virus, A/Hong Kong/2671/2019 (HK/19) (145). Although this strain retained several residues found in the earlier 3C.2a1 virus (SG/16), it exhibited substantial antigenic drift, with numerous substitutions across key HA antigenic sites: site A (T128A, T135K, S137F, A138S), site B (T160I, G186V), site D (D225N), and site E (E62G, K92R) (Figure 4). Notably, three of these substitutions (T128A, T135K, and T160I) resulted in the loss of potential N-linked glycosylation sites, a change associated with enhanced immune evasion and potentially increased virulence. As mentioned previously, this 2020–2021 season marked the peak of the COVID-19 pandemic, and as a result of increased health measures, the spread of influenza was reduced (133). Nevertheless, influenza drift continued, and the following 2021–2022 season included an updated H3N2 recommendation from a clade 3C.2a1b.2a.1a strain, A/Cambodia/E0826360/2020 (CB/20) (134). Some of the key mutations associated with this clade included some reversions in HA antigenic sites (A128T, K135T, F137S, S138A, K171N, N225D), as well as new changes in site A (T131K), site B (Y195F, S198P), site E (K83E), and just outside site E (Y94N) (Figure 4). The egg-adapted virus includes S186R and T160K in antigenic site B, the latter of which is a glycosylation loss; however, the Y195F mutation counters any potential decrease in binding (149). Historically, Tyr195 is extremely conserved throughout all IAV subtypes and is considered part of the base of the RBD. The Y195F mutation results in a hydrogen bond loss which, through molecular dynamics simulations, ensures His183 interacts exclusively with Tyr98 in order to enhance the rigidity of the 130-loop directly involved in sialic acid binding (149).

The H3N2 recommendation over the next two seasons (2022-2024) was a clade 3C.2a1b.2a.2a strain represented by A/Darwin/9/2021 (DR/21) (150). The main mutations were in HA antigenic sites B (H156S, Y159N, K160I, R186N, D190N, P198S) and D (N171K, D225G), and just outside site B (L164Q) (Figure 4). The coevolved Y159N and K160I mutations were shown to weaken receptor binding, but the continuation of the Y195F mutation offset this (149). Another study showed that the D190N mutation, when paired with a G186D mutation, increases binding to α2,6-SA; however, either one of the mutations alone weakened this binding avidity (151). Therefore, the expectation would be that the DR/21 strain would have reduced human-to-human transmission, and vaccines would be more efficient. Vaccine rates were low during the first season DR/21 was used (2022-2023) but much higher in the subsequent season (144).

The 2024–2025 influenza season brought a new H3N2 recommendation, clade 3C.2a1b.2a.3a.1 (subclade J) represented by A/Thailand/8/2022 (TH/22) (152). There were various mutations to this clade in antigenic site B (N186D, I192F), site C (E50K), site D (N96S, G225D), and a reversion of residue 195 (F195Y) (Figure 4). The N96S mutation was a novel N-Gly addition at this site. This TH/22 strain retained Asn159 and Ile160 without Phe195, but the Asn190 residue together with the added N186D mutation countered any expected binding decrease (151). The H3N2 vaccine strain recommendation for the following season (2025-2026) was from the same clade, but different subclade (J.2). As such, the selection, A/Croatia/10136RV/2023 (CR/23) (153), was similar overall; however, it included some key differences such as N122D and S145N in antigenic site A, D186A in site B, Y195F, and K276E in antigenic site C (Figure 4). Of most importance, N122D represented the loss of a glycosylation site. H3N2 was rapidly drifting during this time and resulted in low vaccine efficacy (144). Another variant subclade J.2.4 strain, A/Singapore/GP20238/2024 (SG/24), became the vaccine recommendation for the 2025 southern hemisphere (154). This preceded the emergence of subclade J.2.4.1, which has become known simply as subclade K. A strain from this highly effective variant subclade K, A/Darwin/1454/2025 (DR/25), was recommended as the H3N2 vaccine component for the upcoming 2026–2027 season (136). SG/24 and DR/25 have a T135K mutation in antigenic site A, like HK/19, that removes a potential glycosylation site (Figure 4). DR/25 (but not SG/24) has a glycosylation addition via a S144N mutation in site A, a glycan that was last present with SW/13. Other mutations for DR/25 are found in antigenic site B (N158D, I160K, K189R), site C (K276E), and site D (Q173R), with SG/24 only sharing K189R and K276E (Figure 4). These mutations dramatically shifted antigenicity. As a result, post-infection antisera to earlier vaccines showed reduced reactivity to subclade K viruses (155). Despite this, and a vaccine mismatch for the 2025–2026 season, an interim vaccine-effectiveness study out of France shows statistically significant vaccine effectiveness across all age groups during a time when subclade K was prevalent (156).

IBV vaccine strain changes

IBV was first formulated into a bivalent influenza vaccine in 1942, with a B/Lee/40 virus as the IBV component (111). The IBV component changed eight times prior to the introduction of B/Victoria/2/1987 (B/VI/87) as the seasonal influenza vaccine component for the 1988–1989 influenza season (157). An antigenically drifted strain, B/Yamagata/16/1988 (B/YM/88), replaced it the following season (158). Future IBVs were subsequently delineated on whether they were more B/VI/88-like (Victoria-lineage) or B/YM/88-like (Yamagata-lineage) (Figure 8). Yamagata-lineage viruses were included in the seasonal influenza vaccine until the 2001–2002 season (158). For the next ten seasons, the selection would alternate based on which lineage was predominantly circulating, with Victoria-lineage strains selected eight times and Yamagata-lineage strains selected seven times (158). Over that decade, predictions were incorrect for half of the seasons (159). This in turn led to the approval of a quadrivalent influenza vaccine containing representative strains for both the Yamagata- and Victoria-lineage, for use beginning with the 2013–2014 season (79).

Early on it was thought that antigenicity toward the 150-loop was specific for Yamagata-lineage viruses [44] whereas the 160-loop was specific for Victoria-lineage viruses [45]. However, both are major antigenic sites. The 150-loop is a long protruding loop that is a strong neutralizing epitope for Yamagata-lineage viruses [24]. Most modern Victoria-lineage strains have only developed a minor valine to isoleucine mutation (V146I), but the newest vaccine strains have additional mutations (P144L and N150K) that could indicate increased selective pressure is beginning to be placed on this antigenic site (Figure 7). Victoria-lineage strains since 2017 have developed a series of amino acid deletions in the 160-loop. The 2018–2019 seasonal influenza vaccine included a clade V1A.1 strain, B/Colorado/06/2017 (B/CO/17) (78), with a two amino-acid deletion (2-del) of residues 162-163. In addition to this change, the HA sequence of B/CO/17 differed from the previous Victoria-lineage vaccine strain, B/Brisbane/60/2008 (B/BR/08), by just five amino acids: I117V, N129G, I180V, D197T, and R498K (Figure 7). The recommendation for the Yamagata-lineage component was B/Phuket/3073/2013 (B/PH/13), which was included in quadrivalent formulations since the 2015–2016 seasonal influenza vaccine (145). PH/13 remained the Yamagata-lineage component until the 2023–2024 season (135), following a decision to remove the Yamagata-lineage from formulations after it became eradicated as a result of COVID-19 (160–162). Prior to the peak of COVID-19, the Victoria-lineage vaccine recommendation during the 2020–2021 season was changed to a clade V1A.3 virus, B/Washington/02/2019 (B/WA/19) (132). B/WA/19 has a three amino-acid deletion (3-del) of residues 162–164 and differs from B/CO/17 by additional mutations found predominantly in the 120-loop (G129D, R133G, K136E) and 190-helix (T195S) (Figure 7).

The Victoria-lineage vaccine recommendation was changed again during the 2022–2023 season to a clade V1A.3a.2 virus, B/Austria/1359417/2021 (B/AU/21) (150). The major drifted mutations for B/AU/21 were in the 120-loop and surrounding region (A127T, R133G, G184E), the 150-loop (P144L, N150K), the 190-helix (S194D, K200R), and another residue (R276K) found near the top of the globular head (Figure 7). The same 3-del residues remained in the 160-loop and have continued with the next vaccine strain update four years later as the recommendation for the upcoming 2026–2027 influenza season includes another clade V1A.3a.2 virus, B/Tokyo/EIS13-175/2025 (B/TK/25) (136). This B/TK/25 strain differs by a mere four amino acids (E128K, A154E, D194N, T196I), but each resides in or near the receptor binding domain (Figure 7) and features a change in the side chain property. Together, these mutations suggest enhanced receptor binding through electrostatic charge changes with a net positive gain. Asn194 and Ile196 appear to be compensatory substitutions that are also observed in older IBV strains (Figure 7), but E128K and A154E are mostly novel to IBV.

Discussion

The continual evolution of influenza HA underscores its central role in host specificity, immune escape, and vaccine mismatch. Across IAV and IBV, changes in receptor-binding residues, antigenic loops, and glycosylation patterns reshape both receptor engagement and antibody recognition, while differences in HA1 and HA2 conservation help explain why some regions drift rapidly and others remain structurally constrained. The repeated need to update seasonal vaccine strains for H1N1, H3N2, and IBV reflects this ongoing antigenic drift and the limited durability of strain-specific immunity. At the center of this process is the RBD within the HA1 globular head, which experiences strong and competing selective pressures from host adaptation and population immunity. Mutations within or adjacent to the RBD can alter receptor affinity or shield epitopes from antibody access through glycosylation. Such glycan shielding has already been observed in influenza viruses and has become an increasingly important mechanism of immune escape. A prominent theory suggests that the evolution of hemagglutinin follows a temporal pattern, adding a new glycan every 5 to 7 years, up to a limit (163). After the limit is reached, no net additions occur, but instead glycans are swapped or lost at longer intervals until a pandemic eventually replaces the terminally glycosylated HA with a minimally glycosylated one from the animal reservoir to restart the cycle. This was observed with H1N1, with the shift from seasonal H1N1 to pdm09 H1N1. Currently circulating H3N2 strains contain 10–12 N-Gly motifs, with frequent glycan swapping in site A between the 120- and 130-loops. The emergence of a N-Gly motif at residues 94–96 near antigenic site E marks a new addition, but it is added while others in site A are lost. Cryogenic electron microscopy of the CR/23 strain (164) confirms the glycosylation of these 12 sites as more than just putative sequons. Further antigenic changes may revert those other residues in site A to yield more glycans, potentially pushing H3N2 beyond the proposed glycan limit. If so, this shift in site E may be a harbinger of an H3N2 pandemic in the coming years.

The COVID-19 pandemic likely accelerated the eradication of the Yamagata lineage; however, its disappearance may have been inevitable given the limited antigenic diversity among cocirculating strains in the years preceding its loss. Prior to its disappearance, all Yamagata-lineage viruses belonged to clade 3, resulting in no changes to the vaccine strain after its introduction during the 2015–2016 influenza season. Over the course of this same time, there were four different H1N1 strains, five different H3N2 strains, and three different Victoria-lineage strains. In the absence of competition from the Yamagata-lineage, the Victoria-lineage may undergo even greater antigenic evolution. This could contribute to increased infection rates among infants and young children, who are particularly susceptible to IBV infection (165, 166). Therefore, continued emphasis on seasonal influenza vaccination remains essential.

The emergence of zoonotic H5 and H7 viruses serves as a reminder that small changes in HA can have large consequences for transmissibility and pathogenicity. The HA Q226L mutation increases affinity for α2,6-SA to allow for potential mammalian adaptation. However, zoonosis or increased pathogenicity is generally aided by additional gene changes, such as the PB2 E627K mutation for H5N1 with the TX/24 outbreak, or the reassortment of multiple genes causing the 1957 H2N2 pandemic (HA, NA, PB1) and 1968 H3N2 pandemic (HA, PB1). Nevertheless, a detailed understanding of HA structure, numbering, receptor binding, and antigenic evolution remains essential for improving surveillance, interpreting vaccine strain selection, and preparing for future influenza threats.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fviro.2026.1867727/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Matías Castells, Universidad de la República, Uruguay

Reviewed by: Jiahao Zhang, South China Agricultural University, China

Valeria Caputo, Pomona Ricerca, Italy