Tim Yu

Data, code, and interactive visualizations for comparing amino-acid preferences across H3, H5, and H7 available at: jbloomlab.github.io/ha-preferenc...

Thanks to @jahn0.bsky.social for leading this with me, and also @bdadonaite.bsky.social, Caelan Radford, and @jbloomlab.bsky.social!

This also highlights limitation of using experimental measurements derived from a single genetic background for viral surveillance and vaccine immunogen design. Deep mutational scanning can be useful for predicting mutation effects in closely related variants, but less so across divergent homologs.

Overall, these results consistent with evolutionary contingency. Mutations can modify constraints at other sites, which snowballs over time.

One example is site 176. H5/H7 tolerate similar amino acids, but both are sharply diverged from H3 which only tolerates positively charged K. Structure shows how contacting sites form constrained hydrogen bond network in H3, but same sites have been rewired into hydrophobic environment in H5/H7.

What explains sites with divergent amino-acid preferences? We find that they tend to be buried in the protein and have biochemically distinct wildtype amino acids in the subtypes.

~50% of sites display significant divergence in amino-acid preferences between HAs. HA2 domain of H3/H7 is noticeably less divergent, consistent with higher amino-acid conservation.

We then compared the H7 measurements to previously generated data for H5 (journals.plos.org/plosbiology/...) and H3 (www.nature.com/articles/s41...).

High divergence in amino-acid preferences = HA subtypes tolerate distinct amino acids (ex. site 86).

These data helpful for H7 vaccine immunogen design and viral surveillance. Explore the data interactively at: dms-vep.org/Flu_H7_Anhui...

We first used pseudovirus deep mutational scanning to measure how all mutations to a recent H7 HA affect cell entry. This approach uses virions that can only undergo one round of cell entry and are therefore not capable of causing disease.

Epistatic drift causes gradual decay of predictability in protein evolution The effect and fate of most mutations gradually become unpredictable as proteins evolve.

During protein evolution, mutation effects become less correlated as homologs diverge (www.science.org/doi/10.1126/...).

For HA, we wondered how sequence divergence on a nearly fixed structural backbone affects tolerance to further mutations.

As background, there are at least 19 influenza A virus HA subtypes. Many subtypes are highly diverged at the sequence level (~40% amino-acid identity), but protein structure and cell entry function are highly conserved.

In new work by @jahn0.bsky.social and I in @jbloomlab.bsky.social, we investigate how sequence constraints differ across influenza HA subtypes.

We find ~50% of sites in HA display substantially different amino-acid preferences across H3, H5, and H7.

doi.org/10.64898/202...

Data, code, and interactive visualizations for comparing amino-acid preferences across H3, H5, and H7 available at: jbloomlab.github.io/ha-preferenc...

Thanks to @jahn0.bsky.social for leading this with me, and also @bdadonaite.bsky.social, Caelan Radford, and @jbloomlab.bsky.social!

This also highlights limitation of using experimental measurements derived from a single genetic background for viral surveillance and vaccine immunogen design. Deep mutational scanning can be useful for predicting mutation effects in closely related variants, but less so across divergent homologs.

Overall, these results consistent with evolutionary contingency. Mutations can modify constraints at other sites, which snowballs over time.

One example is site 176. H5/H7 tolerate similar amino acids, but both are sharply diverged from H3 which only tolerates positively charged K. Structure shows how contacting sites form constrained hydrogen bond network in H3, but same sites have been rewired into hydrophobic environment in H5/H7.

What explains sites with divergent amino-acid preferences? We find that they tend to be buried in the protein and have biochemically distinct wildtype amino acids in the subtypes.

~50% of sites display significant divergence in amino-acid preferences between HAs. HA2 domain of H3/H7 is noticeably less divergent, consistent with higher amino-acid conservation.

We then compared the H7 measurements to previously generated data for H5 (journals.plos.org/plosbiology/...) and H3 (www.nature.com/articles/s41...).

High divergence in amino-acid preferences = HA subtypes tolerate distinct amino acids (ex. site 86).

These data helpful for H7 vaccine immunogen design and viral surveillance. Explore the data interactively at: dms-vep.org/Flu_H7_Anhui...

We first used pseudovirus deep mutational scanning to measure how all mutations to a recent H7 HA affect cell entry. This approach uses virions that can only undergo one round of cell entry and are therefore not capable of causing disease.

Epistatic drift causes gradual decay of predictability in protein evolution The effect and fate of most mutations gradually become unpredictable as proteins evolve.

During protein evolution, mutation effects become less correlated as homologs diverge (www.science.org/doi/10.1126/...).

For HA, we wondered how sequence divergence on a nearly fixed structural backbone affects tolerance to further mutations.

As background, there are at least 19 influenza A virus HA subtypes. Many subtypes are highly diverged at the sequence level (~40% amino-acid identity), but protein structure and cell entry function are highly conserved.

Pleiotropic mutational effects on function and stability constrain the antigenic evolution of influenza hemagglutinin The evolution of human influenza virus hemagglutinin (HA) involves simultaneous selection to acquire antigenic mutations that escape population immunity while preserving protein function and stability...

In new study led by @timyu.bsky.social, we measure how mutations to H3 flu HA affect cell entry, stability & antibody escape

We find pleiotropic effects of mutations on these phenotypes shape evolution: epistasis alleviates cell-entry but not stability constraints

www.biorxiv.org/content/10.1...

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

In study led by @ckikawa.bsky.social & Andrea Loes, we use new assay to measure ~10,000 neutralization titers to recent influenza strains & show titers correlate w evolutionary success of viral strains

Similar data could help forecast evolution for vaccine selection

www.biorxiv.org/content/10.1...