Methods

Biosafety

All experiments were performed at biosafety level 2. The work reported here involved influenza virions expressing HA ectodomain proteins from naturally occurring recent human H3N2 or H1N1 influenza strains with non-HA genes from the lab-adapted A/WSN/1933 (H1N1) strain. Both recent human seasonal strains and the lab-adapted A/WSN/1933 strain are classified as biosafety level 2 by the CDC BMBL handbook (edition 6). This study did not involve any generation of viruses with non-natural HA amino-acid mutations; all HA ectodomains are identical to recent human seasonal viruses.

Human sera and plasma

Serum and plasma samples were taken from children and adults across ages and geographical regions through a combination of epidemiological studies, vaccination cohorts, residual blood donation samples, and residual blood draws from hospitals.

The Creative Testing Solutions (CTS) plasma samples were obtained in November 2025 from blood donations collected across the USA from individuals 19–86 years of age through a biobank maintained at CTS by a collaboration between Vitalant Research Institute and the American Red Cross.

The HKU sera were obtained through the ‘Evaluation Population Immunity in Hong Kong’ (EPI-HK) study, a community-based longitudinal observational cohort study of ~2000 individuals run by the University of Hong Kong since 2020 (Cowling et al. 2022). For this study 48 sera were tested, including (i) 35 sera from the 35 out of 42 participants between 10 and 79 years of age who were tested in a similar study prior to the September 2025 vaccine-strain selection (Kikawa et al. 2025) and remained in the study, plus 7 sera from another 7 participants as replacement to those lost to follow-up by matching on age (±10 years of birth year), sex, and vaccination history between 2020/21 and 2024/25 (winter) influenza seasons, with all 42 sera collected between July and November 2025; and (ii) 6 sera from 4 participants with influenza virus infection identified during May to August 2025 from acute respiratory illness (ARI) active surveillance and confirmed by in-house PCR, including 4 sera from 2 participants with paired pre- and postinfection sera and 2 sera from 2 participants with pre-infection sera only. However, in the figures and results here, we only report the titres for the one of these six post-infection sera collected no later than July 2025 (as we are focusing on relatively recent sera); the other titres from sera collected before July of 2025 are in the GitHub repository but not included in the results in this paper. The University of Hong Kong’s Institutional Review Board granted approval for the EPI-HK study protocol. All participants, or their legal guardians where applicable, provided written informed consent prior to enrollment.

The NIID sera were obtained from participants in a vaccine study by the National Institutes of Infectious Disease in Japan. These sera were from unique individuals 20–105 years of age at the Day 0 pre-vaccination time point in November 2025.

The PENN plasma was from adults 24–81 years of age enrolled in a vaccination cohort study by the University of Pennsylvania on the day of and 28 days postvaccination with the 2025–6 Northern Hemisphere seasonal influenza vaccine (FluLaval trivalent influenza virus vaccine from GlaxoSmithKline) between October and November 2025. This study was approved by the University of Pennsylvania Institutional Review Board under protocol number 849398. Note that these sera were recently analysed against a few H3N2 strains by hemagglutination-inhibition assays (Liu et al. 2026).

The SCH sera are deidentified pediatric sera obtained from children 0–15 years of age not known to be immunocompromised receiving routine medical care at Seattle Children’s Hospital in November 2025. These sera were obtained with a signed waiver of consent and the approval from the Seattle Children’s Hospital Institutional Review Board.

Before use in sequencing-based neutralization assays, all sera and plasma were treated with receptor-destroying enzyme II and heat-inactivated following a similar protocol as previously described (Lee et al. 2019). Briefly, receptor-destroying enzyme II (Seikan) was resuspended in 20 ml phosphate-buffered saline (PBS) and passed through a 0.22 um filter. Next, 75 μl of resuspended and filtered receptor-destroying enzyme II was incubated with 25 μl of each serum or plasma (constituting 1:4 dilution) in 96-well PCR plates (BioRad) at 37°C for 2.5 h and then 55°C for 30 min. Plasma samples were then transferred to 96-well V-bottom polypropylene plates (Corning) and spun at 2000 g for 10 min before the cleared supernatant was transferred to new plates. All sera and plasma were used immediately or stored at −80°C until use.

Choice of strains to include in the sequencing-based neutralization assays

Our goal in library design was to select HA strains representative of circulating H1N1 and H3N2 HA diversity in October–November 2025 and likely to remain representative of circulating diversity into the 2025–6 Northern Hemisphere influenza season. We used Nextstrain (Hadfield et al. 2018) pdmH1N1 and H3N2 6-month builds available in October–November 2025 to identify all HA haplotypes, and selected strains from these lists of HA haplotypes. During this process, we aimed to choose ~100 total strains that either had the relatively highest growth rates compared to other contemporaneously circulating strains (Abousamra et al. 2024) or contained HA mutations at either previously defined receptor-binding-site-adjacent and/or antigenic sites (Caton et al. 1982, Wolf et al. 2006, Koel et al. 2013) or HA mutations arising more than expected from the underlying mutation rate (Bloom and Neher 2023, Turner et al. 2026). This initial process selected 58 H3N2 and 48 H1N1 strains. The code for generating lists of HA haplotypes and metadata used to select strains is available at https://github.com/nextstrain/seasonal-flu/blob/02b5b69/notebooks/pick_library_strains.py. The documentation and code describing the design of the barcoded HA constructs is available at https://github.com/jbloomlab/flu-seqneut-2025to2026/tree/main/non-pipeline_analyses/library_design.

The library design also included the HAs from the component strains of both H3N2 and H1N1 seasonal vaccine strains. For H3N2, we included cell- and egg-based vaccine strains from the 2020 to the 2025–6 Northern Hemisphere vaccine, and the cell-based vaccine for the 2026 Southern Hemisphere vaccine. For H1N1, we included cell- and egg-passaged vaccine strains from the 2018 to the 2025–6 Northern Hemisphere vaccine, and the cell-based vaccine for the 2026 Southern Hemisphere vaccine. For H1N1 vaccine strains, most egg-passaged strains had to be dropped as they did not grow well in our system which grows viruses in mammalian cell lines.

Following library generation and quality control procedures, our final libraries comprised 186 barcodes spanning 53 recent H3N2 strains, 30 recent H1N1 strains, 4 historical H3N2 vaccine strains (egg- and cell-produced), and 4 historical H1N1 vaccine strains (egg- and cell-produced) (https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/data/viral_libraries/flu-seqneut-2025to2026-barcode-to-strain-actual.csv). The initial library design included a slightly higher number of strains and barcodes, as some were excluded during library generation and quality control processes: 235 barcodes covering 58 recent H3N2 strains, 48 recent H1N1 strains, 4 historical H3N2 vaccine strains (egg- and cell-produced), and 4 historical H1N1 vaccine strains (egg- and cell-produced) (https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/data/viral_libraries/flu-seqneut-2025to2026-barcode-to-strain-designed.csv).

Cloning of barcoded HAs

The sequencing-based neutralization assays require inserting barcodes into the HA gene so that the assay can be read out by barcode sequencing (Loes et al. 2024, Kikawa et al. 2025, Kikawa et al. 2026). We used a previously described approach to insert the barcodes into the HA gene without disrupting viral genome packaging (Loes et al. 2024, Welsh et al. 2024, Kikawa et al. 2025, Kikawa et al. 2026). The 16-nucleotide barcodes sequences were randomly generated but then checked to specifically avoid barcode sequences used in prior sequencing-based neutralization assays libraries (Loes et al. 2024, Kikawa et al. 2025, Kikawa et al. 2026) or that began with the nucleotides ‘GG’ (we have found such barcodes do not sequence well). All HA genes and linked barcodes were synthesized, cloned, and sequence-verified by Twist Biosciences. As described previously (Welsh et al. 2024, Kikawa et al. 2025, 2026), the plasmid backbone for both the H1N1 and H3N2 constructs was the derivative of the pHH21 uni-directional reverse genetics plasmid (Neumann et al. 1999); see https://github.com/dms-vep/flu_h3_hk19_dms/blob/main/library_design/plasmid_maps/2851_pHH_WSNHAflank_GFP_H3-recipient_duppac-stop.gb for a map of this plasmid backbone. Exemplar plasmid maps for a H1 and H3 strain are at https://github.com/jbloomlab/flu-seqneut-2025/blob/main/non-pipeline_analyses/library_design/plasmids/example_constructs and the full set of all plasmid maps is at https://github.com/jbloomlab/flu-seqneut-2025to2026/tree/main/non-pipeline_analyses/library_design/construct_order/plasmids. The barcodes linked to each strain in the final library are at https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/data/viral_libraries/flu-seqneut-2025to2026-barcode-to-strain-actual.csv.

Generation and titration of barcoded viral libraries

To prepare the pooled libraries of barcoded virions used in the sequencing-based neutralization assays, each strain’s barcoded plasmids (note we had multiple replicate barcodes for most HAs) were pooled and then used to separately generate the barcoded virions with that particular HA. Exactly as described previously (Kikawa et al. 2025), the barcoded viruses expressing each different library HA were generated using reverse genetics (Neumann et al. 1999, Hoffmann et al. 2000) and then passaged on MDCK-SIAT1-TMPRSS2 cells. These virions contained the HA from the barcoded HA with a recent seasonal human influenza HA and all the other genes from the lab-adapted A/WSN/1933 strain as described previously (Kikawa et al. 2025). Then, as in prior work (Kikawa et al. 2025), we made an equal-volume pool of the passaged viruses to determine the relative transcriptional titre of each virus. We used these relative transcriptional titres of each viral strain (see https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/non-pipeline_analyses/library_pooling/notebooks/260107_equal_volume_pool.ipynb) to create a repooled library where each strains’ barcodes would be present roughly equally in the pool. As described previously, we repeated this experiment to affirm strains were roughly equally balanced, as well as determine the range of virus dilutions where viral transcription tracked linearly with the amount of virus particles added to cells (see https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/non-pipeline_analyses/library_pooling/notebooks/260126_balanced_repool.ipynb). Based on this analysis, for the experiments described here we used a 1:24 dilution of the virus library infected on 1.5e5 MDCK-SIAT1 cells per well, as it was in the early part of this linear range where viral transcription is linearly correlated with viral neutralization.

Sequencing-based neutralization assay

The protocol for the sequencing-based neutralization assays was identical to that outlined previously (Kikawa et al. 2025). A step-by-step protocol is available at https://dx.doi.org/10.17504/protocols.io.kqdg3xdmpg25/v2. Exactly as described previously, sera were diluted to 1:20 (accounting for the initial 1:4 dilution from the receptor-destroying enzyme treatment described above) in 50 μl of influenza growth media (Opti-MEM supplemented with 0.1% heat-inactivated fetal bovine serum (FBS), 0.3% bovine serum albumin, 100 μg/ml calcium chloride, 100 U/ml penicillin, and 100 μg/ml streptomycin). This volume was then serially 2.3-fold diluted down 11 of the 12 columns of 96-well plate; the final column was used for the no-serum control wells required to normalize barcode counts to fraction infectivity (Loes et al. 2024). As determined in virus library titration experiments described above, virus library was then added to all wells at a 1:24 dilution in 50 μl, resulting in a range of serum dilutions from 1:40 to 1:13 619 across each 8-well column of a 96-well plate. These virus serum-mixtures were then incubated at 37°C with 5% CO2 for 1 h before 1.5e5 MDCK-SIAT1 cells were added per well. After a 16-h incubation, cells were lysed and barcodes were sequenced as described previously (Kikawa et al. 2025).

The sequencing data were analysed as described previously using the seqneut-pipeline (https://github.com/jbloomlab/seqneut-pipeline), version 6.2.0. See the analysis configuration file (https://github.com/jbloomlab/flu-seqneut-2025to2026/blob/main/config.yml) for details about the parameters used for the barcode counting parameters, curve fitting, and quality control to remove low-quality barcodes, wells, and curves. For HAs with multiple barcodes, we report the median titre across barcodes. The computer code is available at https://github.com/jbloomlab/flu-seqneut-2025to2026/. This pipeline also generates HTML renderings of notebooks performing quality control and generating neutralization curves, as well as interactive plots summarizing the data, which can be explored at https://jbloomlab.github.io/flu-seqneut-2025to2026/.

Phylogenetic analyses

The phylogenetic trees shown in Figs 5 and 7 were built on HA protein sequences and then visualized via Nextstrain Community builds. The trees are rooted on older H3N2 and H1N1 sequences, and the branch lengths represent the number of amino-acid mutations. See https://github.com/jbloomlab/nextstrain-prot-titers-tree for the computer code used to build these trees; this computer code is embedded as submodule in the main GitHub repository for the project (https://github.com/jbloomlab/flu-seqneut-2025to2026).

Note that on the trees and in the figures, strains are labelled by their subclade designation (from a new dynamic nomenclature system; Neher et al. 2026) plus any additional HA1 amino-acid mutations.