Results

A library of influenza HAs representing the diversity of human H3N2 and H1N1 influenza in late 2025 to early 2026

Our goal was to design a library of influenza HAs that could be used to measure neutralization titres of human sera against recently circulating influenza HAs prior to the February 2026 vaccine-strain selection (Fig. 1). In November 2025, we examined all available HA sequences from human H3N2 and H1N1 strains, and chose a set of naturally occurring human seasonal HAs with the goal of covering the existing HA diversity. Specifically, we chose HAs from recently circulating strains with high frequency over the last 6 months window, strains that included mutations at previously defined antigenic or receptor binding site-adjacent sites (Caton et al. 1982, Wolf et al. 2006, Koel et al. 2013), strains that appeared to be rapidly increasing in frequency (Abousamra et al. 2024), and strains with mutations that had arisen more recurrently than expected (Bloom and Neher 2023).

Figure 1: Typical timeline for vaccine-strain selection and production for the Northern Hemisphere influenza season. A decision about which strains to include in the annual vaccine is made twice each year: typically, in early autumn for the Southern Hemisphere vaccine that will be used in the upcoming summer and in late winter for the Northern Hemisphere vaccine that will be used in the upcoming autumn and winter. Illustrated above is the typical timeline for influenza vaccine strain selection and production for the Northern Hemisphere influenza seasons. The periods for traditional virus isolation and characterization are shown alongside our timeline for the sequencing-based neutralization assay measurements reported in the current study.

Overall, our library included 53 recently circulating human H3N2 strains and 30 recently circulating human H1N1 strains (Fig. 2 and Supplementary File 1). As of early 2026, the HAs we chose continue to cover most of the diversity of the HAs of sequenced human H3N2 and H1N1 influenza (Fig. 2). Additionally, we included HAs from eight vaccine strains dating back to the 2020 vaccine for H3N2 and the 2018 vaccine for H1N1.

Figure 2: The HA sequences chosen for our sequencing-based neutralization library are representative of human seasonal influenza circulating in late 2025 and early 2026. (A) Phylogenetic trees of HA genes of H1N1 and H3N2 strains chosen for our sequencing-based neutralization assay library are shown as points, with other recently circulating strains shown as thin lines. Strains are coloured by their subclade designation. Interactive versions of these trees are at https://nextstrain.org/groups/blab/kikawa-seqneut-2025-2026-VCM/h1n1pdm and https://nextstrain.org/groups/blab/kikawa-seqneut-2025-2026-VCM/h3n2. (B) Count and fraction of all human seasonal H1N1 and H3N2 HA sequences available as of 5 February 2026 that closely match a strain in our sequencing-based neutralization assay library (within one HA1 amino-acid mutation). Counts and fractions are averaged over a sliding 10-day window.

We next generated barcoded viruses expressing each of the chosen HAs with the other viral genes from the lab-adapted A/WSN/1933 strain, using previously described approaches (Loes et al. 2024, Kikawa et al. 2026). Most HAs were associated with two or three distinct barcodes to provide internal replicates in the sequencing-based neutralization assays. After quality control, we had a total of 186 unique barcoded viral variants for the 91 different HAs.

A panel of sera collected in late 2025 from humans from multiple locations

We assembled a collection of 302 sera from individuals ranging from 0 to 103 years of age from five different locations (see Fig. 3, which defines the abbreviations we use to refer to each sera set). Most sera were collected in October 2025–November 2025, with a few sera from the HKU set collected earlier in 2025. The PENN sera were collected pre- and postvaccination with the Northern Hemisphere 2025–6 seasonal influenza vaccine (Flulaval Trivalent), which contained a J.2 clade H3N2 component and a D clade H1N1 component; note that some of these sera have recently been analysed against a few H3N2 viral strains by hemagglutination-inhibition assays (Liu et al. 2026). We included the postvaccination sera to quantify how the current vaccine affects titres against current circulating strains; however, it is thought that at a global population level, antibody titres are shaped more by infection than vaccination (Davis et al. 2020, Turner et al. 2020). Some of the HKU and NIID sera are also from individuals with information on recent infection or vaccination status (Supplementary File 2). Most of the remaining sera are from individuals with unknown or incomplete infection and vaccination histories. We prioritized sampling sera from many unique individuals from wide age ranges and geographical locations because these factors are thought to contribute to person-to-person differences in neutralizing antibody titres (Cobey and Hensley 2017).

Figure 3: Overview of the human sera used in the neutralization assays. The map above summarizes the general location, number, distribution of ages, and timeframe of collection for the five sets of human sera used in this study. All the sera were taken from unique individuals with the exception of the University of Pennsylvania sera, for which we tested matched pre- and post-vaccination sera from 59 unique individuals.

Neutralizing antibody landscapes to recent H3N2 influenza strains

We performed sequencing-based neutralization assays to measure neutralization of the 91 viral strains against all 302 sera, for a total of 27 409 titres (a small fraction of titres were dropped due to low-quality neutralization curves; see Methods). We quantify the titres as the reciprocal serum dilution that neutralizes 50% of the infectivity of a given viral strain. In this section we describe the 17 142 titres measured against the H3N2 strains.

The median titres across all sera differed by more than three-fold across the recent human H3N2 strains (Figs 4 and 5, and interactive plots linked in those figure legends). The strains with the lowest titres were in subclade K, a finding consistent with recent studies (Kirsebom et al. 2025, Dee et al. 2026, Liu et al. 2026, Separovic et al. 2026, Wang et al. 2026) showing that this rapidly growing subclade (which is now dominant among H3N2) is neutralized less well by human sera than the both the J.2:S145N strain in the current 2025–6 Northern Hemisphere vaccine and the J.2.4 strain chosen in September 2025 for the 2026 Southern Hemisphere vaccine (Fig. 4). Across subclades, titres to subclade K strains were significantly lower than all other nonsubclade K strains in the library; when compared to each other subclade individually, titres to subclade K strains were not significantly different from J.2.4 strains but were significantly lower than J.2, J.2.2, and J.2.3 strains (Supplementary Fig. 1A and B).

Figure 4: Human neutralizing antibody landscape against recent H3N2 viruses. Median titre across sera (black points) and titres for individual sera (blue lines) against the recently circulating H3N2 strains in the library. The top plot shows titres for all 302 sera, and the other plots show titres by serum set. The viral HA1 haplotype labels are coloured by their subclade. The J.2:S145N strain is the current cell-based 2025–6 Northern Hemisphere vaccine strain, and a J.2.4 strain was chosen in September 2025 for the cell-based 2026 Southern Hemisphere vaccine (these strains are labelled in black). The dynamic range of our assays (determined by the dilution series used for the sera) enabled measurements of titres between 40 and 13 619; sera with titres above or below this range are censored accordingly. See https://jbloomlab.github.io/flu-seqneut-2025to2026/human_H3N2_recent_individual_sera_vertical.html for an interactive version of this plot that can be subset on specific sera or age groups, and for which you can mouseover points and lines for details about viruses and sera. See https://jbloomlab.github.io/flu-seqneut-2025to2026/human_H3N2_recent_interquartile_range_vertical.html for a comparable plot showing the median and interquartile range. Note that this figure shows just the titres against the recent strains and the 2025–6 and 2026 vaccine strains (54 strains total); see https://jbloomlab.github.io/flu-seqneut-2025to2026/ for plots that show titres against the older vaccine strains.

Figure 5: Phylogenetic tree of H3N2 HA proteins coloured by median titre across all sera. Tree of H3N2 HA sequences in the library coloured by the median titre across all sera. The tree is built on the protein sequences and branch lengths are amino-acid mutations relative to the root. All HA1 and HA2 mutations are labelled on branches. See https://nextstrain.org/community/jbloomlab/flu-seqneut-2025to2026@main/H3N2 for an interactive version of this tree with additional colouring options, a measurements panel with the individual serum titres, and an option to show all amino-acid mutations on branches.

All sera sets also show appreciable variation in titres among different subclade K strains. The subclade K strains with the lowest titres have mutations in the defined antigenic regions (Lee and Chen 2004, Wu and Wilson 2017) D (sites 96, 207, and 223) or E (site 261) (Figs 4 and 5 and Supplementary Fig. 1C). This observation is important because the canonical subclade K strain has many mutations in antigenic regions A and B but fewer mutations in regions D and E (Liu et al. 2026); our results suggest that neutralization of subclade K is further eroded by mutations in regions D and E, suggesting strains with such mutations could spread in the future. There are also a handful of strains from J.2.4 with mutations to 135 N (which adds a potential N-linked glycosylation site) that have titres comparably low to some subclade K strains (Figs 4 and 5 and Supplementary Fig. 1D).

In addition to the aforementioned trends in the median titres across sera, there is dramatic variation across different sera (Fig. 4, especially see interactive version linked in figure legend, which makes it easier to explore individual sera). Some of this variation just represents sera that have higher or lower overall titres to all strains. But inspection of the lines in the interactive version of Fig. 4 identifies sera that are strongly impacted by specific mutations: for instance, some sera have dramatically altered titres to subclade K strains with K135E or R189K mutations (both of which are antigenic sites that have undergone recent evolution; Sabaiduc et al. 2025), although these trends are not apparent in aggregated titres since the mutations affect only a subset of sera (Supplementary Fig. 1C). We can also identify some trends that stratify by age group. For instance, sera from children (the SCH set from Seattle Children’s Hospital) have a much more pronounced trend than sera from adults (the other sera sets) towards lower titres to the more recent J.2.3, J.2.4, and K subclades versus the J.2, J.2.2, and G.1.3.1 subclades (Fig. 4).

Neutralizing antibody landscapes to recent H1N1 influenza strains

The 10 267 titres that we measured for the 302 human sera against the H1N1 strains are shown in Figs 6A and 7. The variation in median titres among H1N1 strains is about two-fold, which is somewhat lower than that among the H3N2 strains. Nearly all the H1N1 strains with the lowest median titres are in subclade D.3.1.1 (daughter clade to D.3.1 with HA1 mutations R113K, A139D, E283K, and K302E), which has become dominant among H1N1 strains over the last 6 months. Titres to all six subclade D.3.1.1 strains in our library are lower than the titres to both the subclade C.1.1 strain in the current 2025–6 Northern Hemisphere vaccine and subclade D.3.1 strains, including the strain chosen in September 2025 for the 2026 Southern Hemisphere vaccine strain (Figs 6A and 7 and Supplementary Fig. 1E).

Figure 6: Human neutralizing antibody landscape against recent H1N1 viruses. Median titre across sera (black points) and titres for individual sera (blue lines) against the recently circulating H1N1 strains in the library for (A) sera from individuals of all ages and (B) sera from individuals between 10 and 25 years of age. In both panels, the top plot shows titres for all 302 sera, and the other plots show titres by serum set. The viral HA1 haplotype labels are coloured by their subclade. The C.1.1 strain is the current cell-based 2025–6 Northern Hemisphere vaccine strain, and a D.3.1 strain was chosen in September 2025 for the cell-based 2026 Southern Hemisphere vaccine (these strains are labelled in black). The dynamic range of our assays (determined by the dilution series used for the sera) enabled measurements of titres between 40 and 13 619; sera with titres above or below this range are censored accordingly. See https://jbloomlab.github.io/flu-seqneut-2025to2026/human_H1N1_recent_individual_sera_vertical.html for an interactive version of this plot that can be subset on specific sera or age groups, and for which you can mouseover points and lines for details about viruses and sera. In particular, a slider at the bottom of this interactive plot allows you to subset on just sera from certain age ranges like is done in panel (B) of this figure. See https://jbloomlab.github.io/flu-seqneut-2025to2026/human_H1N1_recent_interquartile_range_vertical.html for a comparable plot showing the median and interquartile range. Note that this figure shows just the titres against the recent strains and the 2025–6 and 2026 vaccine strains (31 strains total); see https://jbloomlab.github.io/flu-seqneut-2025to2026/ for plots that show titres against the older vaccine strains.

Figure 7: Phylogenetic tree of H1N1 HA proteins coloured by median titre across all sera. Tree of H1N1 HA sequences in the library coloured by the median titre across all sera. The tree is built on the protein sequences and branch lengths are amino-acid mutations relative to the root. All HA1 and HA2 mutations are labelled on branches. See https://nextstrain.org/community/jbloomlab/flu-seqneut-2025to2026@main/H1N1 for an interactive version of this tree with additional colouring options, a measurements panel with the individual serum titres, and an option to show all amino-acid mutations on branches.

There are striking strain-specific patterns in H1N1 strain neutralization for subsets of sera that are not well captured by the median titres across all sera (see Fig. 6A and especially the interactive version linked in its legend). The most dramatic of these trends is that adolescents and young adults have clearly reduced titres to strains with mutations at sites 155 or 157 compared to individuals of other ages (Fig. 6B). This observation parallels prior work that has found strong age variation in neutralization titres (and likely susceptibility to infection) of different age groups to H1N1 strains with specific HA mutations (Linderman et al. 2014, Petrie et al. 2016, Arevalo et al. 2020). Similarly, some sera from Seattle Children’s Hospital (SCH) have reduced titres to the D.3.1:R113K,N130K,A139D,E283K strain, whereas most adult sera show similar or higher titres to this strain as to other D.3.1 strains. This difference is likely attributable to site 130, which transitioned from K to N between 2018 and 2022, making N130K a novel mutation for children but a reversion to an amino-acid identity seen in previous exposures for adults.

Changes in neutralization titres after vaccination with the Northern Hemisphere 2025 egg-based vaccine

Our measurements include pre- and 28-day-post-vaccination titres for 59 individuals following receipt of the 2025–6 Northern Hemisphere Flulavel trivalent egg-based vaccine (the PENN sera set, see Fig. 3). During the 2025–6 Northern Hemisphere season, the egg-based vaccine components for H3N2 and H1N1 were a J.2 strain (A/Croatia/10136RV/2023) and a D strain (A/Victoria/4897/2022), respectively.

Across recent H3N2 strains, vaccination typically induced a moderate increase in titres after 28 days (median fold changes of ~1.5–2.5), with some subclade-specific trends (Fig. 8A). Both the final post-vaccination titres and the fold increase in titres tended to be lower for strains belonging to the most recent subclades (K, J.2.4, and J.2.3) (Fig. 8A). There was appreciable variation in the fold-increase in titre across subclade K strains: the base K subclade had an approximately two-fold increase, which is on par with other clades, but other subclade K strains including those with mutations in antigenic region D (e.g. at sites 96, 207, and 223) had lower vaccination-induced titre increases of only ~1.5-fold. Note also that a few sera had much larger increases in titres postvaccination than the median increase of ~1.5–2.5-fold, as can be seen from the interactive per-sera plots linked in the legend to Fig. 8.

Figure 8: Change in human neutralizing antibody titres against recent H3N2 and H1N1 strains after receipt of the 2025–6 Northern Hemisphere influenza vaccine. The median titre (top) or fold-change in titre (bottom) across all sera in PENN set pre- and 28-day postvaccination (with a 2025–6 egg-based vaccine) for all (A) H3N2 and (B) H1N1 strains in the library. The black points are the median across sera, and the shaded blue and black regions are the interquartile range. The viral haplotype labels are coloured by their subclade or by status as a recent cell-based vaccine strain. See https://jbloomlab.github.io/flu-seqneut-2025to2026/PENN_pre_post_vax_H3N2_recent_interquartile_range_horizontal.html and https://jbloomlab.github.io/flu-seqneut-2025to2026/PENN_pre_post_vax_H1N1_recent_interquartile_range_horizontal.html for interactive versions of these plots. See https://jbloomlab.github.io/flu-seqneut-2025to2026/PENN_pre_post_vax_H3N2_recent_individual_sera_horizontal.html and https://jbloomlab.github.io/flu-seqneut-2025to2026/PENN_pre_post_vax_H1N1_recent_individual_sera_horizontal.html for comparable plots showing all individual sera.

For H1N1, the median increase in titres across sera are more modest (~1.4–2-fold) than those observed for H3N2 (Fig. 8B). While pre-vaccination titres were lowest to the D.3.1.1 clades (the key clade that emerged in 2025–6), postvaccination titre increases to D.3.1.1 viruses are similar to those for strains in the D.3.1 subclade (Fig. 8B). The weakest postvaccination responses were observed in G155E-containing strains and in D.3.1.1:K259Q.