Broadly neutralizing antibodies target a haemagglutinin anchor epitope

Abstract

Broadly neutralizing antibodies that target epitopes of haemagglutinin on the influenza virus have the potential to provide near universal protection against influenza virus infection¹. However, viral mutants that escape broadly neutralizing antibodies have been reported^2,3. The identification of broadly neutralizing antibody classes that can neutralize viral escape mutants is critical for universal influenza virus vaccine design. Here we report a distinct class of broadly neutralizing antibodies that target a discrete membrane-proximal anchor epitope of the haemagglutinin stalk domain. Anchor epitope-targeting antibodies are broadly neutralizing across H1 viruses and can cross-react with H2 and H5 viruses that are a pandemic threat. Antibodies that target this anchor epitope utilize a highly restricted repertoire, which encodes two public binding motifs that make extensive contacts with conserved residues in the fusion peptide. Moreover, anchor epitope-targeting B cells are common in the human memory B cell repertoire and were recalled in humans by an oil-in-water adjuvanted chimeric haemagglutinin vaccine^4,5, which is a potential universal influenza virus vaccine. To maximize protection against seasonal and pandemic influenza viruses, vaccines should aim to boost this previously untapped source of broadly neutralizing antibodies that are widespread in the human memory B cell pool.

Fig. 1. The anchor epitope is a common target of stalk-binding antibodies.

a, Negative-stain EM of representative 2D class averages and 3D reconstructions of Fabs binding to A/California/7/2009 (E376K) HA. 047-09 4F04 was imaged at ×52,000 normal magnification and 222-1C06 and 241 IgA 2F04 were imaged at ×62,000 normal magnification. b, Juxtaposed 3D reconstructions of Fabs binding to A/California/7/2009 (E376K) HA. c, Binding footprints of anchor-binding Fabs relative to mAbs targeting the CS epitope (CR9114 and FI6v3). d, Competition of stalk-binding mAbs with CR9114 or 047-09 4F04 (bold mAbs). e, Neutralization potency of anchor-binding (n = 50) and CS-binding (n = 37) mAbs to A/California/7/2009. Data are represented as mean ± s.d. and were analysed by a two-tailed unpaired non-parametric Mann–Whitney test. IC₅₀, half-maximal inhibitory concentration. f, Proportion of anchor-targeting (n = 50) and CS-targeting (n = 37) mAbs binding to other group 1 influenza virus A subtypes. Data were analysed by Fisher’s exact tests. g, Representative 2D class averages (×62,000 normal magnification), 3D reconstructions and footprints of 222-1C06 binding to H2 and the relative footprint on H1. See also Extended Data Figs. 1–3. Source data

Fig. 2. Anchor-targeting mAbs bind to the HA fusion peptide via public binding motifs.

a, b, VH (a) and VK (b) gene usage by anchor-binding mAbs. The number in the centre of the pie graphs indicates the number of mAbs analysed. c, Cryo-EM structure of anchor-targeting 222-1C06 (blue) and lateral patch-targeting 045-09 2B05 (dark grey; see Methods) binding to A/California/7/2009 (E376K) HA (light grey). d, Heavy chain and light chain footprint of 222-1C06 binding to HA based on the cryo-EM structure. e, HA epitope contact residues (maroon) and heavy chain (green) and light chain (yellow) antibody contact residues of the 222-1C06 paratope. Peach-highlighted amino acids represent the fusion peptide of HA2. f, K-CDR3 NWP and H-CDR2 Y58 motifs of 222-1C06. Bold residues are HA residues. g, Major contacts of 222-1C06 K-CDR1 and H-CDR3 (normal typeface) binding to HA (bold residues). h, i, Weblogo plot and germline sequence of Y58 following the H-CDR2 motif (h) and the K-CDR3 NWP motif (i). j, k, Amino acid conservation of s contact residues across human, swine and avian H1 viruses (j) and group 1 influenza A viruses (k). Bold residues are contacts conserved with A/California/7/2009 H1N1 (k). The strain information used for conservation models in j and k are in Supplementary Tables 4, 5, respectively. See also Extended Data Fig. 4. Source data

Fig. 3. MBCs and serum antibodies commonly target the anchor epitope.

a, Proportion of all cH5/1⁺ B cells with repertoire features of anchor-binding mAbs, VH1-69/κ (CS epitope) or with other repertoire features. The number in the centre of the pie graph is the number of B cells analysed. b, Proportion of B cells with anchor-binding mAb features or that use VH1-69/κ-chain by participant (n = 20 donors). Lines connect the same participant. Data were analysed using a two-tailed paired non-parametric Wilcoxon matched-pairs signed rank test. c, Electron microscopy polyclonal epitope mapping (EMPEM) summary of polyclonal antibodies (pAbs) binding to A/Michigan/45/2015 HA in the serum of participants 236 and 241 collected at day 7 and day 14 following 2014 quadrivalent inactivated influenza vaccine. Fabs shown as graphics with dotted lines represent predicted placements due to limited particle representation. 2D class averages were imaged at ×62,000 normal magnification d, Overlap of 241 IgA 2F04 Fab and pAb binding the anchor epitope from participant 241. See also Extended Data Fig. 5. Source data

Fig. 4. cHA vaccination in humans robustly recalls MBCs targeting the anchor epitope.

a, cHA vaccine trial design including vaccine groups (right; group 1 n = 10 participants; group 2 n = 7 participants; group 4 n = 7 participants; group 3 and 5 n = 6). i.m., intramuscular; i.n., intranasal; LAIV, live-attenuated influenza vaccine; PBS, phosphate buffered saline. Bottle images created with BioRender.com. b, c, Fold change by participant of serum antibodies competing for binding to the anchor and CS epitopes after the prime (d29/d1; b) and the boost (d113/85; c). d, e, Proportion of participants who seroconverted (d) and half-maximal effective concentration titres (EC₅₀; e) to the anchor and CS epitopes. Individuals in the cHA (IIV + AS03) cohort were those who received the cHA vaccine with adjuvant (cH8/1 IIV + AS03 prime and cH5/1 IIV + AS03 boost; n = 17 donors) and the IIV cohort were those who received licensed IIV vaccines (2009 monovalent inactivated influenza vaccine, 2010 trivalent inactivated influenza vaccine and 2014 quadrivalent inactivated influenza vaccine; n = 11 donors). Data in b, c and e are mean ± s.d. The dotted line represents the limit of detection. Data in b and c were analysed by two-tailed non-parametric Kruskal–Wallis tests. Data in d were analysed by Fisher’s exact test. Data in e were analysed by two-tailed unpaired non-parametric Mann–Whitney test. See also Extended Data Fig. 6. Source data

Extended Data Fig. 1. Binding and neutralization features of anchor epitope-binding mAbs. Related to Fig. 1.

a, Proportion of HA⁺ mAbs binding to distinct HA domains (left) and proportion of stalk-binding mAbs binding the CS domain (right). Number in the center of the pie graphs represent the number of mAbs tested. b, Proportion of mAbs per cohort that bind the HA stalk domain. c, Negative stain 2D class averages of 047-09 4F04, 241 IgA 2F04, and 222-1C06 binding to H1 (A/California/7/2009 E376K HA). Imaging of 047-09 4F04 was performed at ×52,000 normal magnification and of 222-1C06 and 241 IgA 2F04 at ×62,000 normal magnification. d, Overlay of 047-09 4F04, 241 IgA 2F04, 222-1C06, and FISW84 (PDB:6HJQ) Fabs binding the anchor epitope of A/California/4/2009 HA. e, Overlay of CR8020 (PDB:3SDY), CR8043 (PDB:4NM8), and FISW84 (PDB:6HJQ) modeled on A/California/7/2009 E376G (PDB:4M4Y). f, Footprints of anchor mAb 222-1C06 on H1 (top; PDB: 4M4Y) and CR8020 and CR8043 on H3 (bottom; PDB:4WE4). g, Heatmap of apparent affinity (K_d; M) of anchor-targeting mAbs binding to historical and recent H1N1 viruses. h, Neutralization potency of anchor-binding mAbs (n = 15) against H1-expressing viruses. i, Representative microneutralization curves of anchor- (n = 42) and CS-binding (n = 29) mAbs against A/California/7/2009. j, IC₈₀ of anchor- and CS-binding mAbs against A/California/7/2009. k, ADCC activity of mAbs targeting the CS and anchor epitopes. Dashed line represents the limit of detection (L.O.D). l, ADCC potency of mAbs targeting the anchor (n = 18 mAbs) and CS (n = 8 mAbs) epitopes. Data in h, j, and l are represented as mean ± S.D. Data in j and l were analyzed using a two-tailed unpaired non-parametric Mann-Whitney test. Source data

Extended Data Fig. 2. Anchor-targeting mAb binding to influenza subtype, viral mutants, and polyreactivity antigen panel. Related to Fig. 1.

a, Proportion of anchor- (n = 50 mAbs) and CS-targeting mAbs (n = 37) binding influenza B viruses and H3N2 viruses. b, Negative stain 2D class averages (×62,000 normal magnification) of 222-1C06 binding to H2 (A/Singapore/1/1957), and H5 (A/Indonesia/5/2005). c, H2N2 neutralizing data of anchor- (n = 11 mAbs) and CS-binding mAbs (n = 4) represented as minimum neutralizing concentration. The limit of detection (L.O.D.) is 30 mg/ml. d, Proportion of mAbs targeting the anchor (n = 50 mAbs) or CS (n = 50 mAbs) epitope that are polyreactive. e, LPS binding strength, represented as area under the curve (AUC), of polyreactive mAbs targeting the anchor (n = 30 mAbs) and central stalk (n = 43 mAbs) epitopes. Data are mean ± S.D. f, g, Anchor- and CS epitope-binding mAbs were tested for binding to A/California/7/2009 HA with naturally occurring and experimentally determined mutations induced by 045-09 2B06, a CS-binding mAb. f, Location of mutations modeled on A/California/4/2009 HA (PDB: 4JTV). Residues in blue are located on HA1 and residues in red are located on HA2. Outlines represent binding footprints of 047-09 4F04 (sky blue) and CR9114 (green). g, Heatmap of mAb binding to WT and mutant HAs shown as the proportion of signal relative to mAb binding to the WT HA. Data in a and d were analyzed using Fisher’s Exact tests. Data in e were analyzed using a two-tailed unpaired non-parametric Mann-Whitney test. Source data

Extended Data Fig. 3. Anchor epitope-targeting mAbs are potently protective in vivo and lack ADCC activity. Related to Fig. 1.

a, b, Mice were prophylactically (2 h prior to infection; a) or therapeutically (48 h after infection; b) administered i.p. a cocktail of mAbs (n = 5 mAbs/cocktail) against the anchor- or CS-, or an anthrax-specific antibody. Mice were infected with 10 LD₅₀ of A/Netherlands/602/2009 H1N1. Weight loss (top) and survival (bottom) of mice in each treatment group. c, Lung viral titers of mice in each prophylactic treatment group infected with 1 LD₅₀ of A/Netherlands/602/2009. dpi, days post infection. d, Mice were prophylactically (2 h prior to infection) administered i.p. a cocktail of mAbs (n = 5 mAbs/cocktail) against the anchor- or CS-, or an anthrax-specific antibody. Mice were infected with 10 LD₅₀ of A/Fort Monmouth/1/1947 H1N1. Weight loss (top) and survival (bottom) of mice in each treatment group. For a, b, and d, 10 mice per treatment group were used and data are pooled from two independent experiments. For c, 5 mice per treatment group and timepoint were used except for anchor cocktail group at dpi 3 only 4 mice were used. Data in a, b, and d are represented as mean ± S.D and data in c are represented by mean ± S.E.M. Kaplan-Meier curves in a, b, d were analyzed using a Log-rank Mantel-Cox test, and data in c were analyzed using multiple two-tailed unpaired non-parametric Kruskal-Wallis tests. Source data

Extended Data Fig. 4. Additional repertoire and structural features of mAbs binding the anchor epitope. Related to Fig. 2.

a, VH locus usage by anchor- (n = 52 mAbs) and CS-binding mAbs (n = 37 mAbs). b, VH1 gene usage of mAbs targeting the CS epitope. c, VK locus usage by anchor- (n = 52 mAbs) and CS-binding mAbs (n = 37 mAbs). d, JK gene usage by anchor epitope-binding mAbs. e, Clonal expansions of anchor epitope-targeting mAbs. Numbers indicate heavy and light chain parings, which are described in Extended Data Table 2. f, Heavy and light chain sequences of the public clone. g, h, Mutations (g) and CDR3 amino acid (AA) lengths (h) of heavy and light chains of mAbs binding the anchor (n = 52 mabs) or CS (n = 37 mAbs) epitopes. Data are mean ± S.D. i, Cryo-EM map of 222-1C06 binding to A/California/7/2009 E376K HA. j, k, Local resolution (j) and Fourier Shell Correlation (k) of 222-1C06 binding to HA. l, Aromatic pockets of 222-1C06 binding A/California/7/2009 E376K and FISW84 binding to A/duck/Alberta/35/1976 (PDB:6HJQ; top) and overlay of epitope:paratope interaction (bottom). m, MD simulations demonstrating the K-CDR3 NWP and H-CDR2 Y58 motifs of 222-1C06, FISW84, 241 IgA 2F04, and SFV009 3G01 binding to HA A/California/7/2009 HA. For left-hand panels in l and all panels in m, HA epitope contact residues (maroon) and heavy chain (green) and light chain (yellow) antibody contact residues of anchor mAb paratopes. Peach highlighted amino acids represent the fusion peptide of HA2. n, Fab-Fab interactions of the aromatic pocket of 222-1C06. o, MD simulation of the paratope flexibility of 222-1C06, highlighting the p-stacking of H-CDR2 and K-CDR3. p, Conservation of side-chain contacts of 222-1C06 across seasonal human H1N1 viruses circulating between 1918-2019. q, Deep mutational scanning of the side-chain contacts of 222-1C06. Data in a and c were analyzed using a Chi-square test, and data in g, h were analyzed by two-tailed unpaired non-parametric Mann-Whitney tests. Source data

Extended Data Fig. 5. Features of anchor-targeting MBCs and EMPEM 2D classes. Related to Fig. 3.

a, 33 mAbs with anchor epitope-binding mAb repertoire features were generated and tested for competing for binding with 047-09 4F04. b, c, Number of heavy chain mutations (b) and isotype usage (c) of B cells with repertoire features of anchor-binding mAbs (n = 119 cells) or utilize VH1-69/kappa (n = 365 cells). d, e, 2D class averages of pAbs from donors 236 (d) and 241 (e) at days 7 and 14 post immunization binding to A/Michigan/45/2015 HA (×62,000 normal magnification). The last row of 2D classes in d is HA monomer complexes processed independently from trimer complexes. Data in b are represented as mean ± S.D. Data in b were analyzed using a two-tailed paired non-parametric Wilcoxon matched-pairs signed rank test. Source data

Extended Data Fig. 6. Serum antibody kinetics of anchor- and CS-epitope binding antibodies after cHA vaccination and mAb binding to recombinant HAs. Related to Fig. 4.

a, b, EC50s of serum antibodies competing for binding with 047-09 4F04 for binding to the anchor epitope (a) and CR9114 for binding to the CS epitope (b). a, b, Kinetics of serum antibody responses against the anchor (a) and CS (b) epitopes. Data are mean + S.D. c, d, Proportion of stalk⁺ mAbs per donor (c) or proportion of donors with an isolated anchor mAb (d) upon first exposure to the pH1N1 virus (2009 MIV cohort) relative to donors who have repeatedly been exposed to pH1N1 (2010 TIV and 2014 QIV). Data in c are mean ± S.D. Data in c includes only donors with an isolated anti-stalk mAb, whereas d includes all donors. e, f, Antibody titers (EC₅₀) of serum antibodies collected on day 113 and day 420 against the anchor epitope (e) and the CS epitope (f). Lines connect titers from the same donor and each pair of symbols represents one donor. g, Proportion of anchor epitope-binding mAbs binding to cHA (cH6/1) or mini-HA (n = 50). h, Representative flow cytometry plots of mAbs binding to A/California/7/2009 Cal09 HA and mini-HA (left) and geometric mean fluorescence intensity (gMFI) of mAbs binding to Cal09 and mini-HA (right). Data represent the mean ± S.D. and each symbol represents an individual mAb. i, Proportion of anchor epitope-targeting mAbs binding to A/California/7/2009 recombinant HA with a GCN4 or fibritin trimerization domain (n = 50). For data in a, b, e, and f, Group 1 n = 10 participants, group 2 n = 7 participants, group 4 n = 7 participants, group 3&5 n = 6. For data in c, first exposure n = 7 donors and repeated exposure n = 4 donors. For data in d, first exposure n = 10 donors and repeated exposure n = 13 donors. Data in a, b were analyzed using a two-tailed two-way ANOVA testing for simple effects within rows, data in c and h were analyzed using a two-tailed unpaired non-parametric Mann-Whitney test, data in d, g, and i were analyzed using Fisher’s Exact test, and data in c, d were analyzed using a two-tailed paired non-parametric Wilcoxon matched-pairs signed rank test. See also Supplementary Fig. 1 for gating strategy for panel h. Source data