Egg-adaptive mutations of human influenza H3N2 virus are contingent on natural evolution

Abstract

Egg-adaptive mutations in influenza hemagglutinin (HA) often emerge during the production of egg-based seasonal influenza vaccines, which contribute to the largest share in the global influenza vaccine market. While some egg-adaptive mutations have minimal impact on the HA antigenicity (e.g. G186V), others can alter it (e.g. L194P). Here, we show that the preference of egg-adaptive mutation in human H3N2 HA is strain-dependent. In particular, Thr160 and Asn190, which are found in many recent H3N2 strains, restrict the emergence of L194P but not G186V. Our results further suggest that natural amino acid variants at other HA residues also play a role in determining the preference of egg-adaptive mutation. Consistently, recent human H3N2 strains from different clades acquire different mutations during egg passaging. Overall, these results demonstrate that natural mutations in human H3N2 HA can influence the preference of egg-adaptation mutation, which has important implications in seed strain selection for egg-based influenza vaccine.

Fig 1. Differential preference of egg-adaptive mutations in human H3N2 vaccine strains.

(A) Frequencies of mutations among different egg-passaged seasonal H3N2 vaccine strains are shown as sequence logos. Amino acid variants from residues 183 to 196 are shown (H3 numbering). The number of egg-passaged strains included in this analysis is indicated in the parenthesis. The relative size of each amino acid letter represents its frequency in the sequences. Grey letters represent the amino acid variants that are observed in the corresponding unpassaged parental strains. (B) Key egg-adaptive mutations in this study are shown in orange (PDB 6BKT) [17]. Sialylated glycan receptor is in yellow sticks representation. Strains in parentheses are associated with the corresponding egg-adaptive mutations. Only three strains of interest (Sing16, Switz17, and Kansas17) are included here. (C) Replication fitness of different mutants of human H3N2 vaccine strains was examined in a virus rescue experiment in mammalian cells (293T/hMDCK). Viral titers were measured by TCID₅₀. (D) The viral titer of each human H3N2 vaccine strain with either V186/L194 or G186/P194 was measured during egg-passaging. (C-D) The means of three independent experiments are shown with SD indicated by the error bars. The dashed line represents the lower detection limit. Amino acid variant representing an egg-adaptive mutation is underlined. (E) Frequencies of mutations in the receptor-binding subdomain (residues 117–265) [21] that emerged during serial passaging in eggs. The strain of inoculum for each passaging experiment is indicated above each plot, with those representing egg-adaptive mutations underlined. Frequencies are shown as means of three biological replicates. Only those mutations that reached a minimum average frequency of 10% after the fifth passage are plotted.

Fig 2. Epistatic interactions that involve egg-adaptive mutations.

(A) Amino acid variants from residues 150 to 220 among the indicated strains are shown. The rest of the sequences from residues 150 to 220 were completely conserved among these strains. Major egg-adaptive mutations G186V and L194P are in yellow. Amino acid variants that are unique to Kansas17 strains are in red while variants unique to Switz17 and Sing16 are in blue. Of note, Kansas17 X-327, Switz17 NIB-112, and Sing16 NIB-104 are egg-adapted strains, whereas Kansas17 X-327 V186G/L194P (red) was a mutant generated in this study. (B) Replication fitness of Kansas17 X-327 with indicated mutations was examined in a virus rescue experiment in mammalian cells (293T/hMDCK). (C) The replication fitness of Kansas17 X-327 and Kansas17 X-327 with triple mutations (V186G/L194P/N190D) in eggs were examined during serial egg-passaging. (D) Frequencies of mutations in the receptor-binding subdomain (residues 117–265) [21] that emerged during egg-passaging of the indicated strains. Results are shown as means of three independent biological replicates. Only those mutations that reached a minimum average frequency of 10% after the fifth passage are plotted. (E) Replication fitness of Sing16 and Switz17 L194P virus with or without K160T and D190N mutations was assessed in a virus rescue experiment in mammalian cells (293T/hMDCK). (B, C, E) All viral titers were measured by TCID₅₀. The means of three independent experiments are shown with SD indicated by the error bars. The dashed line represents the lower detection limit.

Fig 3. Incompatibility between N190 and P194.

(A) Frequencies of amino acid variants at residues 160, 186 and 190 in human H3N2 HA over time are shown. Only those variants that reached a maximum annual frequency of 35% are plotted. (B) Two wild type (WT) strains in clade 3C.2a1b.1a (Vic20 and Italy20) with or without L194P mutation were examined in a virus rescue experiment in mammalian cells (293T/hMDCK). Of note, most unpassaged strains in clade 3C.2a1b.1a, including both Vic20 and Italy20, naturally contain N190. (C) The replication fitness of Italy20 with indicated mutations was examined in a virus rescue experiment in mammalian cells (293T/hMDCK) as well as in an egg-passaging experiment, in which 10⁴ TCID₅₀ of the rescued viruses were propagated in eggs for 48 hours. (D) The replication fitness of Italy20 with indicated mutations was measured in a virus rescue experiment in mammalian cells (293T/hMDCK). (B-D) Viral titers were measured by TCID₅₀. The means of three independent experiments are shown with SD indicated by the error bars. The dashed line represents the lower detection limit. (E) The locations of mutations between Sing16 and Italy20 in the receptor-binding subdomain (residues 117–265) [21] are shown in blue (PDB 6BKT). The locations of T160 and Y98 are shown in grey [17]. L194P is in orange. Sialylated glycan receptor is in yellow sticks representation. (F) The structural impact of each mutation was modeled by Rosetta [48]. The distance between the phenolic oxygen of Tyr98 (OH₉₈) and the Cα of residue 190 was computed as the height of receptor-binding site (RBS). Three replicates, each with 100 simulations, were performed. Each data point represents the lowest scoring pose in a replicate. Error bars represent the SD. P-value was computed by two-tailed t-test. Only the difference between WT and G186V is statistically significant (P < 0.05).

Fig 4. Egg-adaptive mutations in recent human H3N2 HA.

A rooted phylogenetic tree was built using maximum likelihood (ML) method with 100 bootstraps on the HA sequences of 61 human H3N2 strains without egg-passaging (unpassaged strains). Different clades are highlighted in different colors. WHO-recommended vaccine strains are indicated by a red triangle symbol. These unpassaged human H3N2 strains correspond to the parental strains of 72 egg-passaged strains. Since multiple egg-passaged strains with different HA sequences could be generated from a single unpassaged strain, the number of unpassaged strains is less than that of egg-passaged strains. Frequencies of amino acids at HA residues 160, 186, 190, 194, 219 and 225 among egg-passaged strains in different clades are shown as sequence logos. The relative size of each amino acid letter represents its frequency in the sequences. For each clade, amino acid variants that are also observed in the unpassaged strains are in grey and listed below each sequence logo.