H5N1 Variants From a Severe Human Case Show Weakened Grip on Host Receptors

In November 2024, an adolescent girl in British Columbia was hospitalized with respiratory failure caused by a novel H5N1 influenza virus, a strain now designated BC24 within clade 2.3.4.4b. Intact virus could not be isolated from the patient, but sequencing of material from a tracheal aspirate collected eight days after symptom onset revealed a mixed viral population. Alongside the well-known E627K mutation in polymerase basic 2, the most frequent changes appeared in hemagglutinin, the surface glycoprotein that influenza uses to attach to and enter host cells. Two substitutions in the receptor-binding site, E190D and Q226H by H3 numbering, each accounted for roughly a third of sequence reads, evidence that viruses carrying them were fit enough to persist in a severely ill host. A new study published in Nature Communications has now dissected, at atomic resolution, what those mutations do to the receptor-binding machinery of the H5 hemagglutinin, and the answer is unexpected.

Hemagglutinin is the linchpin of influenza A infectivity. It recognizes sialosides, sugar chains terminating in sialic acid, on the surface of host cells, and after the virus has been taken up by endocytosis it triggers the fusion of viral and endosomal membranes. Avian influenza viruses typically favor sialic acids linked through an alpha-2,3 configuration, which are found deeper in the human respiratory tract, whereas human-adapted viruses prefer alpha-2,6-linked sialosides enriched in the upper airway. Both linkage types exist in human airways, and their distribution is thought to shape where a given virus can replicate. Whether E190D or Q226H might switch H5 receptor preference has been a pressing question, because position 190 has been implicated in receptor switching in H1 viruses and leucine at position 226 in H5 hemagglutinin has been shown to confer alpha-2,6 binding, although histidine at that position had not shown the same effect.

To find out, a team led by Sriram Subramaniam at the University of British Columbia used cryogenic electron microscopy to solve the structures of three H5 hemagglutinin ectodomains expressed and purified under identical conditions: the BC24 protein carrying both mutations of interest, the clade 1 protein from the fatal 2004 Vietnam case A/Vietnam/1203/2004 (VN04), and the clade 2.3.4.4b protein from A/Michigan/90/2024 (MI24), a recent strain that caused only mild human disease and is closely related to BC24. Globally, the three proteins are structurally very similar, with root mean square deviations of 0.441 angstroms between BC24 and VN04 and 1.095 angstroms between BC24 and MI24. The mutations do not substantially alter the fold of the protein near or far from the receptor-binding site. The dramatic difference lay elsewhere: in the electron density maps.

Previous structural work on H5 hemagglutinins had revealed that an N-linked glycan attached to asparagine 169 folds back on itself and occupies the virus’s own receptor-binding pocket, acting as an auto-glycan that mimics an alpha-2,3-linked sialoside. In the new structures, clear density for this auto-glycan is visible in the receptor-binding sites of both VN04 and MI24 hemagglutinin, but it is entirely absent from BC24 hemagglutinin. The researchers ruled out the possibility that the glycan simply was not attached in their preparation, since the same expression system, comparable glycosylation at other sites, and the close sequence relationship between MI24 and BC24 all argue that N169 glycosylation should occur similarly in both. The most plausible explanation is that the BC24 receptor-binding site has lost the affinity needed to hold the glycan in place, which in turn implies greatly reduced affinity for alpha-2,3-linked sialosides on host cells.

The structural basis for this loss is geometrically precise. In the VN04 and MI24 structures, the glutamate at position 190 positions its carboxylate for hydrogen bonding 2.9 and 2.8 angstroms, respectively, from the C9 hydroxyl of the terminal sialic acid, while the glutamine at 226 places its amide 3.3 and 3.2 angstroms from the C4 hydroxyl of the second sugar, galactose. In BC24, the aspartate introduced by the E190D mutation sits one methylene shorter and ends up 4.9 angstroms from where the bound sialic acid would align, while the histidine of Q226H pivots away from the galactose hydroxyl. The residues that are ideally positioned to grip an alpha-2,3-sialylated glycan in other strains are precisely the ones mutated in the patient’s virus, offering a clean structural explanation for the empty binding pocket.

Biochemical assays confirmed and extended the structural finding. On N-linked and O-linked glycan microarrays, VN04 and MI24 hemagglutinin bound exclusively to alpha-2,3-linked sialosides, with MI24 showing the broader breadth toward fucosylated glycans such as sialyl Lewis X that has been reported for clade 2.3.4.4 proteins. BC24 hemagglutinin, by contrast, showed no detectable binding to any glycan on the arrays, neither alpha-2,3 nor alpha-2,6 linked. Dynamic light scattering confirmed that all protein preparations were comparably stable, indicating that the differences reflect genuine receptor selectivity rather than artifacts of sample quality. Enzyme-linked immunosorbent assays probing four representative biantennary sialosides told the same story: neither BC24 nor any mutant carrying aspartate 190 or histidine 226, on either the MI24 or BC24 backbone, bound significantly to either linkage type at concentrations up to 50 micrograms per milliliter, while restoring glutamate 190 and glutamine 226 on the BC24 background reverted binding to alpha-2,3-linked sialosides.

Crucially, the mutagenesis work addressed the mixed viral population observed in the patient. Because E190D and Q226H each appeared in roughly a third of sequence reads, and it is unknown how often both mutations reside on the same hemagglutinin allele, the team generated single and double mutants in both forward and reverse directions. Every combination carrying either single mutation reproduced the severely reduced binding phenotype, meaning that multiple distinct minority variants circulating in the patient independently shared the weak receptor-binding trait. The findings also rule out the hypothesis that these mutations switched receptor specificity from alpha-2,3 to alpha-2,6 sialosides; instead, they simply crippled binding to both.

Yet the virus was clearly functional. In cell-based fusion assays using A549 human alveolar carcinoma cells, which display both sialoside linkage types, full-length BC24 hemagglutinin drove membrane fusion and syncytium formation, albeit at substantially reduced levels, yielding about 0.32-fold the luminescent signal of VN04 hemagglutinin while MI24 reached 1.3-fold. The authors suggest that multivalent avidity effects, in which many weak interactions collectively support attachment, or binding to alternative receptors not represented on the arrays, such as sulfated glycans, may compensate for the low biochemical affinity. Notably, BC24 hemagglutinin mediated fusion at levels statistically significantly above those of the severely impaired Y98F mutant, a variant known to permit viral proliferation depending on cell type, suggesting the weakly binding variants remain biologically competent. The broadly neutralizing antibody MEDI8852, which targets the conserved stem region, potently inhibited BC24 hemagglutinin-driven fusion with estimated IC50 values comparable to VN04 and MI24, and the patient’s neuraminidase carried no known resistance mutations, leaving conventional antiviral options intact.

The immunological data add a further layer. Serum from the patient, collected at four time points, showed strong reactivity against both BC24 and MI24 hemagglutinin and cross-reacted robustly with H1, H2, and H3 hemagglutinins, indicating a vigorous humoral response that nonetheless failed to prevent severe disease. Healthy pediatric controls displayed modest cross-reactivity against the clade 2.3.4.4b proteins, likely from seasonal vaccination or prior influenza exposure, supporting growing evidence that the population may not be entirely naive to H5N1. The authors propose that weakened receptor binding may have helped these variants evade upper-airway receptors and enrich in the distal airways, where the patient’s viral loads were highest, a hypothesis consistent with the historical association between lower respiratory tract infection and H5N1 severity. No human-to-human transmission followed the case, consistent with the absence of alpha-2,6 binding, but the study demonstrates that severe human disease can occur without classical human-receptor adaptation, and that weakly binding hemagglutinin variants can be fit in a severe infection context, adding an important dimension to surveillance and pandemic preparedness efforts.

Subject of Research: Structural and functional analysis of H5N1 hemagglutinin receptor-binding mutations from a severe human influenza case

Article Title: Diminished sialoside binding in novel H5N1 influenza hemagglutinin variants identified in a human patient

Article References: Diminished sialoside binding in novel H5N1 influenza hemagglutinin variants identified in a human patient. (n.d.). https://doi.org/10.1038/s41467-026-77829-x

Image Credits: AI Generated

DOI: 10.1038/s41467-026-77829-x

Keywords: H5N1, influenza, hemagglutinin, sialoside binding, cryo-EM, receptor-binding site, E190D, Q226H, avian influenza, clade 2.3.4.4b, glycan microarray, zoonosis