henhaobio
Sign in

eLifeEvolutionary BiologyPeer-reviewed journal article

High-throughput neutralization measurements correlate strongly with evolutionary success of human influenza strains

Caroline KikawaiD, Andrea N. LoesiD, John HuddlestoniD, Marlin D FigginsiD, Philippa L. SteinbergiD, Tachianna Griffiths, Elizabeth M. DrapeauiD, Heidi PeckiD, Ian BarriD, Janet A. EnglundiD, Scott E. HensleyiD, Trevor BedfordiD, Jesse D. BloomiD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Human influenza viruses rapidly acquire mutations in their hemagglutinin (HA) protein that erode neutralization by antibodies from prior exposures. Here, we use a sequencing-based assay to measure neutralization titers for 78 recent H3N2 HA strains against a large set of children and adult sera, measuring [~]10,000 total titers. There is substantial person-to-person heterogeneity in the titers against different viral strains, both within and across age cohorts. The growth rates of H3N2 strains in the human population in 2023 are highly correlated with the fraction of sera with low titers against each strain. Notably, strain growth rates are less correlated with neutralization titers against pools of human sera, demonstrating the importance of population heterogeneity in shaping viral evolution. Overall, these results suggest that high-throughput neutralization measurements of human sera against many different viral strains can help explain the evolution of human influenza.

In-depth summary

Claude Sonnet 4.6 · xhigh effort · 7 Nov 2026 · from the full text of version 3 (2 Dec 2025) · CC BY 4.0

A sequencing-based neutralization assay across 78 H3N2 HA strains and 150 sera shows strain growth in 2023 tracks the fraction of sera with low titers, but not pooled-serum titers.

Hemagglutination inhibition and conventional neutralization assays are low throughput, so it is hard to measure titers of many human sera against all strains circulating in a season. Here ~10,000 titers from a barcoded-virus sequencing assay are compared with strain growth rates.

Key results

  1. ShownNeutralization titers vary widely between individuals and between strains within a serum. Much of this variation is idiosyncratic and not explained by age group for 2023 strains, and the spread is larger among children than among adults.

    78 strains were tested against 150 sera (11,700 titers). The main analysis used 56 children and 39 pre-vaccination adults. A few children's sera neutralized all strains at titers >1,000, but most were about tenfold lower. Strains with mutations at site 145 had lower titers in children but not in adults. · Neutralization titers to 2023 strains are…; Discussion

  2. ShownTiters of equal-volume serum pools do not represent the individual sera. Pool titers exceeded the median of the constituent sera and varied less across strains.

    Pool titers were always above the median individual titer and only modestly correlated with it across 2023 strains. The mismatch was most marked for children, whose sera are the most heterogeneous. · Pooled sera fail to capture the heterogeneity of…

  3. SuggestedThe 2023 growth rate of a strain correlated strongly with the fraction of sera that had low titers against it. Titers against pooled sera showed no correlation with growth.

    Growth rates were estimated by multinomial logistic regression for 12 strains with at least 80 sequence counts. The correlation peaked at a titer threshold of 138 and beat all 200 permutations (P<0.005). Pooled children and adult sera showed no correlation. · Evolutionary success of viral strains correlates…; Figure legends

  4. ShownThe growth-titer correlation cannot be cleanly separated from phylogenetic structure. The count of HA1 amino-acid mutations also correlates strongly with growth, and the two predictors are collinear.

    Permutation feature importance in a multiple regression gave 68% importance for the low-titer fraction and 32% for HA1 mutation count. The variance inflation factor was 2.7. The authors conclude titers are not shown to add much beyond mutation counting. · Evolutionary success of viral strains correlates…; Modeling strain growth

  5. SuggestedVaccination raised titers to all 2023 strains, by different amounts in the two adult cohorts. Egg-passaged vaccine strains were neutralized more than cell-passaged ones of the same year.

    Increases were typically ~2-4-fold in the USA cohort (39 adults) and ~5-10-fold in the Australia cohort (8 adults), but the cohorts differ in vaccine and prior vaccination. Back-boosting to the oldest vaccine strains was small. · Vaccination of adults broadly increases…

Methods, data and reproducibility

Barcoded influenza viruses carrying 78 H3N2 HAs (3 barcodes each) were pooled and run in a neutralization assay read out by Illumina barcode sequencing. Titers were fitted with Hill curves and compared with strain growth rates from multinomial logistic regression on GISAID sequence counts.

  • Children's residual sera from Seattle Children's Hospital (ages 1-14), collected in December 2023 · human · n = 56 children (within 150 sera in total)
  • Philadelphia adult vaccine cohort, pre- and 28 days post-vaccination (ages 22-74) · human · n = 39 adults
  • Australia adult vaccine cohort, pre- and 17-21 days post-vaccination (ages 26-54) · human · n = 8 adults
  • Barcoded virus library of H3N2 HAs: 62 strains from 2023 plus 16 vaccine strains from 2014-2024 · influenza A virus (H3N2) · n = 78 strains
  • H3N2 HA sequences from GISAID EpiFlu, used for library coverage and for 2023 growth-rate estimation · influenza A virus (H3N2)
Methods & tools
seqneut-pipeline v3.1.3, neutcurve, evofr, Nextstrain, GISAID EpiFlu, baltic, Vega-Altair v5.2, UCSF ChimeraX v1.9, Illumina sequencing
Software & environment
seqneut-pipeline version 3.1.3; Vega-Altair v5.2; UCSF ChimeraX v1.9; evofr growth models assume a serial interval of 3.6 days.
Experimental conditions
Library design and construction took ~12 weeks. About 10,000 measurements took about four cumulative weeks of experimental work.
Limitations (5)
  • Sera came from few sources: children at one Seattle hospital and heavily vaccinated adults in Philadelphia and Australia. Some age groups are missing, and the sample may not represent global immunity. · stated by the authors
  • Only 12 strains had enough sequences for growth-rate estimates, and these were mostly moderate-to-high fitness, which narrowed the range of growth rates. Models give one fixed growth rate per strain for 2023. · stated by the authors
  • The strains are phylogenetically related, so the significance of the titer-growth correlation is hard to assess. Mutation count is collinear with titers, so their contributions cannot be convincingly separated. · stated by the authors
  • Only HA-directed neutralizing antibodies were measured; non-neutralizing HA antibodies and anti-neuraminidase antibodies were not. · stated by the authors
  • The growth-rate validation covers a single season (2023) and 12 strains, with no independent replication in another year. · noted in this summary

Read this if you want evidence that high-throughput neutralization titers from many individual sera can explain influenza strain success, or you want the assay's data and pipeline.

Citations
10
FWCI
32.63
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISKikawa '26Lee '18Loes '24Kikawa '25Kikawa '26Welsh '23
76% semantic matchMicrobiology
Caroline Kikawa, Andrew Butler, John Huddleston+16 more·bioRxiv·Sep 21, 2026· 1 citations