Panzootic H5 influenza viruses acquired resistance to human head interface antibodies

TL;DR

H5 influenza viruses evolved to resist human antibodies, enabling global spread and mammal-to-mammal transmission.

Contribution

The study identifies a single mutation (P221S) that enables H5 influenza to evade human antibodies and traces its origin.

Findings

Head interface antibodies bind many H5 isolates but not most from the GsGd lineage.

The P221S substitution is key to antibody resistance and emerged in Chinese avian reservoirs by 2000.

Viruses with P221S have caused a global panzootic and transmit between mammals.

Abstract

Antibodies to the influenza hemagglutinin protein (HA) confer the strongest protection against infection. Immunity elicited by endemic, seasonal, human viruses is correlated with diminished disease severity and death caused by antigenically novel viruses. Antibodies to the HA head interface are broadly protective and abundant in human serologic and memory repertoires. Notably, few head interface antibodies from H5 naive donors are reported to bind H5 HAs. We find head interface antibodies engage a wide range of H5 isolates but fail to engage most isolates from the goose Guangdong (GsGd) lineage. We identify a single substitution, P221S, largely dictates antibody binding. Phylogenetic analysis indicates that P221S arose in a Chinese avian reservoir by the year 2000. Descendants of these viruses have caused the current global panzootic and have achieved sustained mammal-mammal transmission in farmed and wild mammals. Our findings demonstrate that viral evolution in non-mammalian species can, by chance, produce viruses that resist broadly protective human antibody responses.

Figures

Head interface antibody binding to HAs.(A) Left: HA trimer from A/Goose/Guangdong/1/1996 (H5N1) (PDB: 4MHI); protomers are colored white, light gray, and light yellow. Middle: A/Goose/Guangdong/1/1996 (H5N1) (P221) HA head aligned with A/Vietnam/1203/2004 (H5N1) (PDB: 3FKU) (S221) HA head colored dark gray. Right: Head interface antibodies S5V2-29 (PDB:6E4X) colored in blues and S8V2-18 (PDB: 6XQ0) colored in reds modeled on the A/Goose/Guangdong/1/1996 (H5N1) HA head. (B) A zoomed in view of panel A rotated 90° about the x-axis. (C) Dissociation constants of antibodies to rHA were measured by ELISA. Stem-directed antibody FI6v3 [18] was utilized as a positive control and a SARS-CoV antibody, CR3022 [19], was used to assess nonspecific binding. Complete isolate information is available in S1 Table. (D) Dissociation constants of antibodies against A/Hong Kong/156/1997 P221/P221S, and A/Viet Nam/1203/2004 S221/S221P.

Phylogenetic analysis of H5NX isolates from 1995-2000 and HAs used in this study.Sequences containing P221 are colored black. Sequences containing S221 are colored red. Stars indicate HAs used in ELISAs: black star indicates a sequence between 1995-2000, and open star indicates a sequence outside 1995-2000. Red and black circle marks the last common ancestor of P221 and S221 Goose Guangdong viruses. Open gray circle marks node from which post-2000 Goose Guangdong lineage members descend. A labeled tree is in S4 Fig.

Full text

Introduction

Antibodies to the influenza virus hemagglutinin protein (HA) are most strongly correlated with protection against infection [1,2]. Pandemics can occur when humans are infected by antigenically novel viruses which can readily transmit between humans. Descendants of pandemic viruses sustain circulation in humans by acquiring mutations that enable them to evade protective immunity elicited by previous infections. Immunity elicited by endemic human influenza viruses is correlated with protection against severe and lethal outcomes during subsequent infections with antigenically novel viruses [3–7]. Influenza exposed humans harbor antibodies that engage conserved sites on HA [8]. These antibodies are broadly binding and can confer protection against divergent subtypes in animal models. These include antibodies to the head interface/trimer interface [9–12], an occluded, conserved surface (S1 Fig) that is a point of contact between heads in the pre-fusion HA trimer (Fig 1 A and B). Antibodies to this site are abundant in memory B cell [9] and serologic repertoires [13] of some adults and are likely present in most humans exposed to influenza. Few head interface antibodies from H5-naive donors have been found to bind representative H5 HAs, despite binding many other subtypes with high affinity [9,10,13].

Head interface antibody binding to HAs.(A) Left: HA trimer from A/Goose/Guangdong/1/1996 (H5N1) (PDB: 4MHI); protomers are colored white, light gray, and light yellow. Middle: A/Goose/Guangdong/1/1996 (H5N1) (P221) HA head aligned with A/Vietnam/1203/2004 (H5N1) (PDB: 3FKU) (S221) HA head colored dark gray. Right: Head interface antibodies S5V2-29 (PDB:6E4X) colored in blues and S8V2-18 (PDB: 6XQ0) colored in reds modeled on the A/Goose/Guangdong/1/1996 (H5N1) HA head. (B) A zoomed in view of panel A rotated 90° about the x-axis. (C) Dissociation constants of antibodies to rHA were measured by ELISA. Stem-directed antibody FI6v3 [18] was utilized as a positive control and a SARS-CoV antibody, CR3022 [19], was used to assess nonspecific binding. Complete isolate information is available in S1 Table. (D) Dissociation constants of antibodies against A/Hong Kong/156/1997 P221/P221S, and A/Viet Nam/1203/2004 S221/S221P.

We re-examined head interface antibody binding to H5 HAs using panels of divergent H5 HAs and monoclonal head interface antibodies isolated from five H5-naive donors [9,10,14]. These antibodies were isolated from different repertoires (serologic, memory, and plasma cell) and were chosen for their genetic diversity. We find most are capable of binding H5 HAs but fail to bind most members of a specific Eurasian H5 lineage of pandemic concern named for an archetypic isolate, A/goose/Guangdong/01/1996(H5N1) (GsGd) [15–17]. We show that this lineage acquired a mutation at an otherwise conserved site, P221S (H3N2 numbering), in Chinese Aves by the year 2000 that is sufficient to disrupt head interface antibody binding. Comparisons between reported structures with P221 and S221 suggest the mutation causes a slight adjustment of the 220 loop, a major point of contact for head interface antibodies (Fig 1A, 1B, and S1 Fig) [9–12]. Descendants of H5 S221 viruses are responsible for a global panzootic, have achieved sustained mammal-mammal transmission in farmed mustelids, dairy cattle, and wild pinnipeds, and have caused sporadic human infections [15–17]. Viral evolution in non-human reservoirs can impact the binding of broadly protective antibodies elicited by endemic human viruses.

Results

Head interface antibodies bind H5 hemagglutinins

We produced a recombinantly expressed panel of divergent soluble HA ectodomains (rHA) that samples H5 spatial, temporal, and genetic diversity (n = 14) alongside human H1 and H3 rHA controls. We also produced a panel of 12 recombinantly expressed monoclonal human head interface antibodies. These include four from the germline biased, broadly binding, IGkV1-3901 class [9–12] that were isolated from three donors: subjects 1 and 5 (S1 & S5) [9,10] and Donor 1 (D1) [13]. We also included eight antibodies of diverse gene utilization from subjects 1, 5, and 8 (S1, S5, and S8) [9,10], and donor EI-13 [14]. We evaluated binding by enzyme-linked immunosorbent assay (ELISA) and included broadly binding stem antibody FI6v3 [18] and SARS-CoV-1 antibody CR3022 [19] as positive and negative binding controls, respectively (Fig 1C and S2A Fig). Ten of twelve bound at least one H5 rHA, with eight binding to the majority of rHAs. Antibodies from the IGkV1-3901 class bound the most H5 rHAs. The pattern of H1 and H3 rHA binding matched previous reports [9,10]. The two antibodies that did not bind to an H5 rHA were reported to have limited-to-no breadth of binding to group 1 HAs [9,10].

Most head interface antibodies failed to bind a common subset of rHAs. These are descendants of the GsGd lineage that were isolated after 1997. All of these rHAs are from human infections, except A/dairy cow/Texas/24-008749-001/2024(H5N1) (similar H5N1 viruses from dairy cattle have infected humans). These HAs share a mutation at an otherwise conserved position, P221S. Among the tested antibodies, only S8V2-18 [9,10] bound A/Viet Nam/1203/2004(H5N1) rHA, while S1V2-37 weakly bound A/Egypt/N03072/2010(H5N1) rHA, and S1V2-58 [9,10] weakly bound A/dairy cow/Texas/24-008749-001/2024(H5N1) rHA. Notably, most of the head interface antibodies bound A/Cambodia/i0125001G/2024(H5N1) rHA, which has a proline at position 221.

The P221S mutation disrupts head interface antibody binding

We introduced S221 into A/Hong Kong/156/1997(H5N1) (H5-HK-97) and P221 into A/Viet Nam/1203/2004(H5N1) (H5-VN-04) rHAs to determine the impact of P221S on head interface antibody binding. Most antibodies in our panel bound H5-HK-97 rHA (except S8V2-37 [9,10], S1V2-83 [9,10] and S1V2-51 9,10) and only S8V2-18 [9,10] bound H5-VN-04 rHA (Fig 1D and S2B Fig). The H5-HK-97-P221S mutant was not bound by any head interface antibody, while six of the 12 head interface antibodies bound the H5-VN-04-S221P mutant. This mutation did not disrupt S8V2-18 [9,10] binding. For these H5 rHAs, the presence of a serine or proline at position 221 is a strong determinant of head interface antibody binding. Introduction of the P221S mutation in A/California/04/2009(H1N1) HA also diminished or abrogated the binding of many of these antibodies (S3 Fig).

Origin of H5 S221 viruses

To determine the origins of the S221 H5s, we surveyed all reported full-length H5NX HA sequences in GISAID [20]. We established that S221 is prevalent in the GsGd lineage (98.2% of sequenced isolates) (S4 Fig). In contrast, P221 is found in 99.4% of American and 98.9% of non-GsGd Eurasian lineage viruses. Phylogenetics and sequence metadata indicate that GsGd viruses acquired the P221S mutation by the year 2000 in an avian reservoir (Figs 2, S5 Fig, and S1 Table). In the year 2000, these viruses were widely distributed among ducks and geese in Eastern and Southeastern China (Hong Kong, Guangdong, Shanghai, Huadong, and Zhejiang) (Fig 2). Descendants of these viruses have largely retained S221 and are responsible for the current global panzootic.

Phylogenetic analysis of H5NX isolates from 1995-2000 and HAs used in this study.Sequences containing P221 are colored black. Sequences containing S221 are colored red. Stars indicate HAs used in ELISAs: black star indicates a sequence between 1995-2000, and open star indicates a sequence outside 1995-2000. Red and black circle marks the last common ancestor of P221 and S221 Goose Guangdong viruses. Open gray circle marks node from which post-2000 Goose Guangdong lineage members descend. A labeled tree is in S4 Fig.

Discussion

Influenza viruses evolve in wild avian reservoirs and infect farmed and captive animals. Highly pathogenic avian influenza viruses (HPAI), like those from the GsGd lineage, have transmitted to humans, caused severe disease and death, and are considered to have pandemic potential. We establish that the GsGd lineage, which has caused a global panzootic and achieved sustained mammal-mammal transmission in wild and farmed animals [15–17], has acquired resistance to an abundant, broadly binding, broadly protective class of human antibodies. Resistance is conferred by a single mutation, P221S, which emerged in Chinese Aves by the year 2000. HPAI viruses may also have additional resistance to head interface antibodies because their HA trimers are stabilized by efficient furin-mediated processing of their polybasic sequences [12]. Viral evolution in nature can produce variant viruses that, by chance, evade broadly protective human antibodies elicited by past influenza exposures.

Head interface antibodies have been isolated from donors that were immunized with “pre-pandemic” GsGd S221 encoding H5 vaccine candidates [11,12]. These antibodies bind representative GsGd rHAs. Vaccination may be sufficient to elicit or update these specific head interface antibodies. Panzootic GsGd lineage 2.3.4.4b (S221) members are currently circulating on six continents, are responsible for the deaths of tens of thousands of mammals, have achieved sustained transmission in farmed mustelids and dairy cattle, and have caused severe illness and death in humans [15–17]. Given the scale of this event and the species involved, the extent to which head interface antibodies confer protection in humans and whether vaccination can elicit 2.3.4.4b-binding head interface antibodies remain pressing questions.

Methods

Cells

293F cells were maintained at 37°C with 8% CO_2_ in FreeStyle 293 Expression Medium (ThermoFisher) supplemented with penicillin/streptomycin (ThermoFisher).

Recombinant HA and IgG Expression and Purification

Sequences encoding HA ectodomains (FLsE) were subcloned into a modified pVRC8400 vector encoding a T4 fibritin (foldon) trimerization tag and a 6xHis tag [21,22]. The A/dairy cow/Texas/24-008749-001/2024(H5N1) rHA was further modified with stabilizing mutations (HA0: H355W, K380I and E432I) [23]. When present, polybasic sites were not modified. Antibody heavy and light chain variable domains were cloned into modified pVRC8400 expression vectors to produce full IgG1 heavy chains and light chains [21,22]. HA and antibodies were expressed by polyethylenimine (PEI) facilitated, transient transfection of 293F cells. Transfection complexes were prepared in Opti-MEM (Gibco) and added to cells. 5 days post transfection, cells were pelleted, supernatants were harvested. rHA supernatants were incubated overnight with Talon Metal Affinity Resin (Takara) at 4°C. Resin was collected in a chromatography column, washed with 10mM Tris, 150mM NaCl pH 7.5, eluted using 350mM imidazole in 10mM Tris-HCl, 150 mM NaCl pH 7.5, concentrated using a 10kDa centrifugal filter (Millipore), then purified via size exclusion chromatography on a Superdex 200 (GE Healthcare). IgG1 supernatants were incubated overnight with Protein A Agarose Resin (GoldBio) at 4°C. Resin was collected in a chromatography column and washed with 10mM Tris, 150 mM NaCl pH 7.5. mAbs were eluted in 0.1M glycine pH 2.5 which was immediately neutralized by 1M Tris (pH 8.5), concentrated in 10kDa centrifugal filter (Millipore), and dialyzed in PBS pH 7.4.

Recombinant HA ELISA

ELISAs were performed as previously described [21,22]. 500 ng of rHA FLsE were adhered to wells of a high binding 96-well plate (Corning) in PBS pH 7.4 overnight at 4°C. Plates were washed with PBS with 0.05% TWEEN-20 (PBS-T) and blocked at room temperature for 1 hour with PBS-T with 2% BSA. Blocking solution was removed, and serially 5-fold diluted mAbs in PBS-T 2% BSA were added to plates in triplicate. A positive binding control, mAb FI6v3 [18], was present on each plate. Plates were incubated for 1 hour at room temperature, then washed 3 times with PBS-T. Anti-human HRP IgG (abcam) diluted 1:10,000 in PBST 2% BSA was added to wells and incubated for 30 minutes at room temperature. Plates were washed 3 times with PBS-T, developed with 1-Step ABTS (ThermoFisher), and stopped by 1% SDS solution. Absorbances at 405nm were measured on a Molecular Devices SpectraMax 340PC384 Microplate Reader. Background subtracted data were normalized to the FI6v3 [18] standard curve present on each ELISA plate per rHA. Equilibrium dissociation constant K_d_ values were calculated using a nonlinear fit (One site binding, hyperbola) and graphed using GraphPad PRISM v10.

Phylogenetics

H5 phylogenetic tree was generated by first downloading all available, full length H5 HA sequences from GISAID [20]. Sequences were aligned using MAFFT [24]. Downselected alignments were produced for samples collected from January 1^st^ 1995 through December 31^st^ 2000 or using CD-HIT [25] (clustered at 95%, retaining one representative sequence per cluster). HA sequences for isolates characterized in this study were added to these alignments. Phylogenetic trees were produced using MrBayes [26] and images made in FigTree [27].

All genome sequences and associated metadata in this dataset are published in GISAID’s EpiFlu database. To view the contributors of each sequence with details such as accession number, Virus name, Collection date, Originating Lab, Submitting Lab and the list of Authors, visit EPI_SET_250807vm DOI: https://doi.org/10.55876/gis8.250807vm.

Supporting information

S1 TableStrain name and accession numbers for HA isolates.(PDF)

S1 FigHA ectodomain sequence alignment.Sequence alignment of HA the ectodomains for HAs used in this study. Differences from A/Goose/Guangdong/1/1996 (H5N1) are indicated. Numbering is based on the H3 numbering. Key points of contact for head interface antibodies are bolded, the polybasic site is underlined, and serines at position 221 are colored red.(TIFF)

S2 FigELISA titration curves of recombinant HAs.The pan-influenza A virus-binding stem-directed antibody FI6v3 [18] was used as a positive control and a SARS-CoV antibody, CR3022 [19], was used to assess nonspecific binding. A. Binding curves for H1, H3, and H5 HA ELISA panel. B. Binding curves for wild type and HA 221 mutant HAs. Data points represent the average of three technical replicates. Error bars are the standard error of the mean.(TIF)

S3 FigHead interface antibody binding to a wild type and P221S H1 HA.(A) Equilibrium dissociation constants for antibodies binding to A/California/04/2009 P221 and A/California/04/2009 P221S HAs. The stem-directed antibody FI6v3 [18] was utilized as a positive binding control and a SARS-CoV-1 antibody, CR3022 [19], was used as a no binding control. Wild type is abbreviated WT. (B) Percentage of H1N1 sequences containing proline or serine at position 221. (C) ELISA titration curves.(TIF)

S4 FigH5 phylogenetic tree.North American Non-Goose Guangdong strains are colored orange. Eurasian Non-Goose Guangdong strains are colored Green. Goose Guangdong strains are colored Blue. Black stars indicate strains tested for antibody binding in ELISAs. The frequencies of P221 or S221 isolates in each lineage are reported in the inset.(TIF)

S5 FigPhylogenetic tree of H5NX isolates from 1995-2000 and HAs used in this study with labels.Sequences containing P221 are colored black. Sequences containing S221 are colored red. Stars indicate strains tested by ELISA.(TIF)