A deep mutational scanning study of every amino-acid mutation in the hemagglutinin (HA) of the recent human H3N2 influenza strain A/Massachusetts/18/2022 showed that mutations at conserved trimer-interface sites 165, 205, 220 and 229 (H3 numbering; for example N165H, S205Y, R220T and R229I) strongly escape neutralization by human sera yet destabilize HA, and have never fixed in decades of natural H3N2 evolution. How these mutations escape serum neutralization is unresolved: they may directly abrogate binding of antibodies targeting the trimer interface or epitopes spanning protomers, or their destabilizing effect may accelerate the low-pH membrane-fusion transition and shorten the window for antibody neutralization, a strategy previously reported for stalk-binding antibodies. The same study found that epistasis alleviates the pleiotropic costs of antigenic mutations on HA-mediated cell entry but not on HA acid stability, and the biophysical reason for this phenotype specificity is likewise unexplained.

Background

Influenza A virus is an enveloped, negative-sense RNA virus whose major surface glycoprotein, hemagglutinin (HA), is at once the principal target of neutralizing antibodies and the machine that mediates cell entry. Each HA spike is a trimer of three identical subunits (protomers); each protomer is cleaved into HA1 (the receptor-binding head) and HA2 (the stem). After the virus binds sialylated receptors and is endocytosed, the low pH of the endosome triggers an irreversible conformational rearrangement of HA that fuses the viral and endosomal membranes. 'Acid stability' denotes the pH threshold at which this fusogenic rearrangement triggers: a less stable HA refolds at a pH closer to neutrality. Antibody-driven antigenic drift continuously selects HA mutations that escape population immunity, but a mutation can simultaneously impair HA function or stability (pleiotropy), and the effect of a mutation can depend on the other mutations present in the same protein (epistasis). Which antigenic mutations can actually fix during evolution is therefore shaped by whether epistasis alleviates, or fails to alleviate, their pleiotropic costs.

Deep mutational scanning (DMS) is a method in which a barcoded library containing nearly all amino-acid mutations in a protein is put through a functional selection and sequencing is used to quantify the effect of every mutation at once. Yu et al. (2025) applied DMS to all 9,576 amino-acid mutations in the HA ectodomain (the extracellular portion of the protein) of A/Massachusetts/18/2022 (MA22, the H3N2 strain in the 2024–2025 seasonal vaccine), measuring three phenotypes: HA-mediated cell entry, acid stability, and neutralization by sera collected in 2023 from four children. Strikingly, mutations at sites 165, 205, 220 and 229 — located in the trimer interface, the buried surfaces where the three HA protomers contact one another — strongly escape serum neutralization and are roughly neutral for cell entry, but destabilize HA. These sites are almost invariant across decades of natural H3N2 evolution, consistent with their stability cost preventing fixation; indeed, some antigenic mutations that strongly escape human antibodies never fix in nature because they pleiotropically decrease acid stability. The same study found that the effect of a mutation on cell entry shifts across genetic backgrounds (epistasis, including entrenchment — reversions of historically fixed substitutions to earlier amino-acid states have themselves become deleterious — and enabling of formerly deleterious antigenic mutations), whereas effects on acid stability remain consistently deleterious across backgrounds.

Two bodies of prior work frame why the escape mechanism at these interface sites is puzzling. First, humans do make antibodies against the HA head interface: a prevalent, focused human antibody response to a conserved head-interface epitope (the surface region of HA recognized by an antibody) has been described (McCarthy et al. 2021), and human trimer-interface antibodies with canonical structural features have been isolated (Zost et al. 2021); but such antibodies are broadly protective rather than potently neutralizing in conventional assays, and, as the source notes, humans have not been shown to possess neutralizing anti-H3 trimer-interface antibodies — so what serum antibody activity the DMS escape mutations are evading is itself unclear. Second, HA destabilization can itself mediate escape: mutations that enhance the fusion transition allow influenza A virus to escape a broadly neutralizing stalk-binding antibody (Chai et al. 2016). Hence two mechanistic hypotheses coexist for the trimer-interface escape mutations, and, more broadly, the reason epistasis modulates cell entry but not acid stability — one hypothesis being that stability is a single additive biophysical property while entry is a higher-order phenotype depending nonlinearly on receptor binding, fusion and stability — remains untested.

Problem Statement

The problem, as posed by the source study (Yu et al. 2025), comprises two linked parts concerning the mechanistic coupling between HA stability, membrane fusion, and antigenic evolution.

(1) Mechanism of serum-antibody escape at the HA trimer interface. Determine how mutations at the conserved HA trimer-interface sites 165, 205, 220 and 229 (H3 numbering; for example N165H, S205Y, R220T and R229I in the A/Massachusetts/18/2022 H3N2 background) reduce neutralization by human serum antibodies. The source speculates about two candidate mechanisms: (a) the mutations directly abrogate the binding of serum antibodies that target the trimer interface, nearby epitopes, or epitopes spanning protomers; or (b) the destabilizing effect of the mutations accelerates the membrane-fusion transition, shortening the time window during which antibodies can neutralize the virus — a strategy previously reported for escape from stalk-binding antibodies. Resolving this part means identifying the mechanism actually responsible — whether one of these two candidates, their combination, or a different mechanism — and thereby clarifying which serum antibody specificities the mutations evade (including whether human neutralizing anti-H3 antibodies can target the trimer interface).

(2) Biophysical basis of phenotype-specific epistasis. Explain why, across H3N2 genetic backgrounds, epistasis alleviates the pleiotropic costs of antigenic mutations on HA-mediated cell entry but not on HA acid stability. The source proposes, without testing, one hypothesis: acid stability reflects a single underlying biophysical property on which mutations act largely additively, whereas cell entry is a higher-order phenotype whose nonlinear dependence on several underlying properties (receptor binding, membrane fusion, protein stability) naturally generates epistasis. This knowledge is what determines whether antibodies targeting stability-constrained regions such as the trimer interface are evolutionarily difficult to escape, and is needed to improve forecasting of HA antigenic evolution and to guide the design of therapeutics and vaccine immunogens that are more resistant to viral escape.

References

  1. Yu, T. C., Kikawa, C., Dadonaite, B., Loes, A. N., Englund, J. A. & Bloom, J. D. Pleiotropic mutational effects on function and stability constrain the antigenic evolution of influenza haemagglutinin. Nature Ecology & Evolution 10, 452–466 (2025). doi:10.1038/s41559-025-02895-1, https://doi.org/10.1038/s41559-025-02895-1 (preprint version, posted 24 May 2025 under the title 'Pleiotropic mutational effects on function and stability constrain the antigenic evolution of influenza hemagglutinin': doi:10.1101/2025.05.24.655919, https://doi.org/10.1101/2025.05.24.655919)
  2. Chai, N., Swem, L. R., Reichelt, M., Chen-Harris, H., Luis, E., Park, S., Fouts, A., Lupardus, P., Wu, T. D., Li, O., McBride, J., Lawrence, M., Xu, M. & Tan, M.-W. Two escape mechanisms of influenza A virus to a broadly neutralizing stalk-binding antibody. PLOS Pathogens 12, e1005702 (2016). doi:10.1371/journal.ppat.1005702, https://doi.org/10.1371/journal.ppat.1005702
  3. McCarthy, K. R., Lee, J., Watanabe, A., Kuraoka, M., Robinson-McCarthy, L. R., Georgiou, G., Kelsoe, G. & Harrison, S. C. A prevalent focused human antibody response to the influenza virus hemagglutinin head interface. mBio 12, e01144-21 (2021). doi:10.1128/mbio.01144-21, https://doi.org/10.1128/mbio.01144-21
  4. Zost, S. J., Dong, J., Gilchuk, I. M., Gilchuk, P., Thornburg, N. J., Bangaru, S., Kose, N., Finn, J. A., Bombardi, R., Soto, C., Chen, E. C., Nargi, R. S., Sutton, R. E., Irving, R. P., Suryadevara, N., Westover, J. B., Carnahan, R. H., Turner, H. L., Li, S., Ward, A. B. & Crowe, J. E. Canonical features of human antibodies recognizing the influenza hemagglutinin trimer interface. Journal of Clinical Investigation 131, e146791 (2021). doi:10.1172/jci146791, https://doi.org/10.1172/jci146791
  5. Graber, A. L., Simmons, H. C. & McCarthy, K. R. Panzootic H5 influenza viruses acquired resistance to human head interface antibodies. PLOS Pathogens 22, e1014005 (2026). doi:10.1371/journal.ppat.1014005, https://doi.org/10.1371/journal.ppat.1014005
  6. McCarthy, K. R. Some exposure for influenza hemagglutinin interface antibodies. Immunological Reviews 342, e70167 (2026). doi:10.1111/imr.70167, https://doi.org/10.1111/imr.70167
  7. Yang, G., Chen, P.-L., Rovito, S. W., Minari, K., Writt, H. N., DeBeauchamp, J., Crumpton, J. C., Kercher, L., DuBois, R. M., Webby, R. J. & Russell, C. J. Resistance mutations to broadly neutralizing antibodies destabilize hemagglutinin and attenuate H1N1 influenza virus. Viruses 18, 32 (2025). doi:10.3390/v18010032, https://doi.org/10.3390/v18010032
  8. Ahn, J. J., Yu, T. C., Dadonaite, B., Radford, C. E. & Bloom, J. D. Influenza hemagglutinin subtypes have different sequence constraints despite sharing extremely similar structures. Virus Evolution 12, veag018 (2026). doi:10.1093/ve/veag018, https://doi.org/10.1093/ve/veag018
  9. Larsen, B. B., Harari, S., Gen, R., Stewart, C., Veesler, D. & Bloom, J. D. Functional and antigenic constraints on the Nipah virus fusion protein. Proceedings of the National Academy of Sciences 123, e2529505123 (2026). doi:10.1073/pnas.2529505123, https://doi.org/10.1073/pnas.2529505123
  10. Huddleston, J. & Bedford, T. Timely vaccine strain selection and genomic surveillance improve evolutionary forecast accuracy of seasonal influenza A/H3N2. eLife 14, RP104282 (2025). doi:10.7554/eLife.104282, https://doi.org/10.7554/eLife.104282
  11. Liang, R., Peccati, F., Ponse, N. L. D., Uslu, E., de Rooij, A. J. H., Han, A. X., Boons, G.-J., Unione, L. & de Vries, R. P. Epistasis in the receptor-binding domain of contemporary H3N2 viruses that reverted to bind sialylated di-LacNAc repeats. Cell Reports 44, 116007 (2025). doi:10.1016/j.celrep.2025.116007, https://doi.org/10.1016/j.celrep.2025.116007
  12. Zost, S. J. Learning a viral protein's vocabulary. Nature Ecology & Evolution 10, 399–400 (2025). doi:10.1038/s41559-025-02928-9, https://doi.org/10.1038/s41559-025-02928-9