The Pre-Existing Human Antibody Repertoire to Computationally Optimized Influenza H1 Hemagglutinin Vaccines

Influenza viruses are a major cause of morbidity and mortality worldwide each year (1). In particular, influenza A viruses (IAVs) and influenza B viruses cause annual epidemics in humans, and IAVs have caused multiple pandemics during the past century (2). Currently, H1N1 and H3N2 IAVs cause epidemic disease (3, 4). Long-term protection to influenza viruses remains a challenge due to high mutation rates caused by a low-fidelity RNA polymerase, which leads to antigenic drift, as well as reassortment events of hemagglutinin (HA) and neuraminidase with avian influenza viruses, which is termed antigenic shift (5). Current seasonal influenza vaccines provide protection against matched circulating viral strains. However, vaccine efficacy varies year to year due to mismatches between circulating strains and vaccine strains, as well as differences in HA protein glycosylation patterns between vaccine and circulating strains (6–8). This variability in vaccine efficacy highlights the importance of developing an improved influenza vaccine, which would elicit an immune response to most circulating influenza A and/or B viruses (9). Current vaccines typically elicit strain-specific Abs, and only a minority show cross-reactivity to other viral subtypes. The Ab response to influenza virus infection and vaccination focuses predominantly on HA. Within HA-targeting Abs, those targeting the variable globular head domain dominate the response, whereas those that bind the more conserved stem domain are elicited less frequently (10).

H1N1 IAVs have caused two known pandemics, including the Spanish influenza pandemic of 1918–1919, which caused an estimated 40–50 million deaths, and the 2009 swine influenza pandemic, which caused an estimated 575,000 deaths (11). Circulating 2009 pandemic pH1N1/09-like viruses have replaced pre-2009 seasonal H1N1 influenza viruses in the human population (11). Antigenic sites defined on the H1 subtype HA have been characterized through mutagenesis studies in the presence of neutralizing Abs (12). These highly variable sites are present on the immunodominant head domain, and include the Sa, Sb, Ca1, Ca2, and Cb sites (12). More recently discovered Ab epitopes include the receptor-binding site (RBS), the lateral patch, and the intratrimeric epitope, which exhibit broader reactivities (9, 13–15).

Computationally optimized broadly reactive Ag (COBRA) HA immunogens aim to elicit a broader Ab response compared with current seasonal vaccines (16, 17). In this approach, multiple-layered consensus building alignments of HA sequences are used to generate an immunogen encompassing multiple epitopes for a single subtype (16). The resulting constellation of consensus epitopes, focused primarily in the antigenic sites of the head domain, represent diverse sequences that elicit broadly reactive Abs in several animal models, including in mice and ferrets (16, 17). Structural analysis of COBRA HA immunogens has shown that these Ags resemble wild-type HA proteins (18). The primary mechanism of COBRA HA-induced Abs are through hemagglutination inhibition (HAI) and neutralization via HA head domain-binding Abs (16). In contrast, stem-directed Abs do not appear to be a major component of COBRA HA vaccine-induced immunity (19, 20). Some H1 subtype–based COBRA HAs have been previously described that incorporate both seasonal (pre-2009) and pandemic-like (post-2009) influenza virus HA sequences. These include P1, which incorporates human sequences from 1933 to 1957 and 2009 to 2011 as well as swine sequences from 1931 to 1998, and X6, which incorporates human sequences from 1999 to 2012 (21). The Y2 COBRA HA, encompassing sequences from 2014 to 2016, represents the most recent set of H1 subtype viruses that are antigenically similar to the 2009 pandemic H1N1 virus. Importantly, the Y2 COBRA HA has been shown previously to elicit broadly HAI-active Abs against recent 2018 and 2019 H1N1 pandemic-like viruses in a mouse model, whereas the P1 and X6 COBRA HAs elicit a less broad response against these isolates (22).

An individual’s immune history to influenza also plays a major role in the Ab response to vaccination. For example, the idea of original antigenic sin describes the dominant nature of the Ab response to the first influenza virus strain compared with exposures to subsequent strains (23). Although COBRA HAs have been shown to be efficacious in naive as well as preimmune mouse and ferret models of influenza infection, pre-existing immunity to COBRA HAs in humans has not been investigated. This is important to understand as these Ags move toward clinical trials. In this study, we identify epitope and repertoire characteristics of the pre-existing Ab response from previous infection and vaccination that is recognized by H1 COBRA HA Ags. We show that human Abs and B cells that cross-react with COBRA HAs, predominantly with the Y2 COBRA HA and a minority with the P1 and X6 COBRA HAs, are present in individuals vaccinated with the 2017–2018 and 2019–2020 quadrivalent influenza vaccine (QIV). A panel of mAbs was isolated, and these mAbs bind five distinct epitopes on the A/California/04/2009 HA protein, including an epitope near the viral membrane, termed the anchor epitope. Moreover, a subset of these Abs bind both pre- and post-2009 pandemic strains with demonstrable HAI and neutralization activity. Overall, our data identify the major epitopes and repertoire characteristics of pre-existing human Abs that recognize COBRA HA Ags.

All human studies were approved by the University of Georgia Institutional Review Board. mAb isolation was conducted from subjects aged 22–51 y vaccinated with the 2017–2018 seasonal influenza vaccine (Fluzone) from PBMCs isolated from blood draws 21–28 d following vaccination. Of these subjects, 80% (4/5) showed HAI titers >40 to at least one pre-2009 H1 virus before vaccination. Repertoire sequencing was completed from a single human subject vaccinated with the 2019–2020 influenza vaccine (Fluzone) from blood obtained 28 d following vaccination.

PBMCs were plated at a density of 25,000 cells/well in a 96-well plate on a layer of gamma-irradiated NIH 3T3 cells (20,000 cells/well) expressing human (h)CD40L, hIL-21, and hBAFF in the presence of CpG and cyclosporine A as previously described (24, 25). B cell supernatants were screened by ELISA at 7 d postplating of PBMCs.

Trimeric wild-type HA or COBRA HA ectodomains were expressed and purified in Expi293F cells following the manufacturer’s protocol and as previously described (26). Collected supernatants containing the HA Ags were purified on a HisTrap Excel column following the manufacturer’s recommended protocol. Eluted fractions were pooled and purified proteins were verified for integrity by probing with an anti-HIS tag Ab (BioLegend) as well as with subtype-specific mAbs via SDS-PAGE and Western blot.

Untreated 384-well plates (VWR) were coated with recombinant HA proteins diluted to 2 μg/ml in PBS at 4°C overnight. Plates were washed once with water, then blocked with 2% blocking buffer (PBS + 2% nonfat dry milk [Bio-Rad] + 2% goat serum + 0.05% Tween 20) for 1 h at room temperature. Plates were washed three times with water, and 25 µl of B cell supernatants, hybridoma supernatants, mAbs, or recombinant Abs (rAbs) was added. mAbs and rAbs were serially diluted 3-fold in PBS from 20 μg/ml prior to addition for 12 total dilutions. Plates were incubated at 37°C for 1 h, then washed three times with water. Goat anti-human IgG Fc–alkaline phosphatase secondary Ab (SouthernBiotech), diluted 1:4000 in 1% blocking buffer (1:1 dilution of PBS and 2% blocking buffer), was added and plates were incubated at room temperature for 1 h. Plates were then washed five times with PBST (PBS + 0.05% Tween 20). p-Nitrophenyl phosphate substrate, diluted in substrate buffer (1.0 M Tris + 0.5 mM MgCl₂, pH 9.8) to 1 mg/ml, was added and plates were incubated for 1 h and read at 405 nm on a BioTek plate reader. To quantify HA-reactive IgG from each subject, plates were coated overnight with eight 2-fold serial dilutions of human plasma IgG standard (Athens Biotechnology) starting at 10 μg/ml. All steps were followed as for Ag, except PBS was used in the primary Ab step. GraphPad Prism was used to interpolate Ag-reactive IgGs from the human plasma IgG standard curve. The EC₅₀ value for each mAb was determined by using the four-parameter logistic curve fitting function in GraphPad Prism software.

Eight days following plating of PBMCs, wells identified to contain positive B cells by ELISA were selected for electrofusion to generate hybridomas as previously described (24, 25). Hybridomas were plated in 384-well plates for HAT selection, and grown for 14 d at 37°C, 5% CO₂. Following screening by ELISA, hybridomas were single-cell sorted using a MoFlo Astrios cell sorter using live/dead staining by propidium iodide. The sorted hybridomas were cultured in 25% medium E (STEMCELL Technologies) + 75% medium A (STEMCELL Technologies) for 2 wk, then subjected to another round of screening by ELISA. Hybridomas with the highest signal were grown in 250 ml of serum-free medium (Life Technologies) for ∼ 1 mo. Secreted mAbs were purified using a protein G column (GE Healthcare) and concentrated for use in downstream assays.

Hybridoma cell lines encoding each mAb were sequenced utilizing the primers described by Guthmiller et al. (27). Briefly, RNA was extracted from each hybridoma and cDNA was generated using the SuperScript IV first-strand cDNA synthesis kit (Invitrogen). A nested PCR protocol was used to generate sequencing products. In the first nested PCR step, a primer mix specific to the H, κ-, or λ-chain V gene and the C region was used to amplify the V region using the cDNA as template. In the second PCR step, the first PCR product was used as a template with a nested primer mix to improve product specificity and yield. The second nested PCR products were sequenced using the C region primer and the V, D, and J alleles were identified by IMGT/V-QUEST (28). Percent identity of mAb variable regions to germline was calculated as the similarity to the germline allele at the nucleotide level using IMGT/V-QUEST (28).

The HAI titer for each mAb was determined as previously described (20). Influenza viruses were titered to 8 hemagglutination units). Fifty microliters of mAbs or rAbs diluted to 20 μg/ml in PBS was added to the first well of a 96-well U-bottom plate (VWR) and diluted 2-fold in PBS for 25 μl of mAb total per dilution. Eight hemagglutination units of virus was added in a 1:1 ratio to each mAb dilution, and each well was mixed and incubated for 20 min at room temperature. Following this, 50 μl of 1.0% turkey RBCs (Lampire Biological Laboratories) was added per well. Plates were read 45 min after the addition of 1.0% turkey RBCs.

Focal reduction assays were completed for each mAb as previously described (20). MDCK (Madin–Darby canine kidney) cells were plated in 96-well plates overnight to achieve >95% confluency the next day. Cells were washed twice with PBS, and 50 µl of virus growth medium (VGM; DMEM + 2 μg/ml TPCK-trypsin + 7.5% BSA) was added and the plates were returned to the incubator at 37°C, 5% CO₂. mAbs at 20, 8, or 1 μg/ml were serially diluted 2-fold in VGM, and virus was diluted to a concentration of 1.2 × 10⁴ focus-forming units/ml in VGM. MDCK cells were washed with PBS and 25 μl of serially diluted mAbs was added, followed by 25 μl of 1.2 × 10⁴ focus-forming units/ml virus. Plates were incubated at 37°C, 5% CO₂ for 2 h, and then 100 μl/well of overlay medium (1.2% Avicel + modified Eagle’s medium) as added and incubated overnight. The overlay was removed and wells were washed twice with PBS. Ice-cold fixative (20% formaldehyde + 80% methanol) was added and plates were incubated at 4°C for 30 min. Plates were washed twice with PBS and permeabilization buffer (PBS + 0.15% glycine + 0.5% Triton X 100) was added, followed by a 30-min incubation. Plates were washed three times with PBST and primary IAV anti-nucleoprotein mouse Ab (International Reagent Resource), diluted 1:2000 in ELISA buffer (PBS + 10% goat serum + 0.1% Tween 20), was added. Plates were incubated at room temperature for 1 h. Plates were then washed three times with PBST, and secondary goat anti-mouse IgG-HRP Ab (SouthernBiotech), diluted 1:4000 in ELISA buffer, was added. Plates were incubated at room temperature for 1 h and then washed with PBST. Kirkegaard & Perry Laboratories TrueBlue peroxidase substrate was added per well and plates were incubated for 10–20 min. Plates were washed and dried, and foci were enumerated using an ImmunoSpot S6 Ultimate reader with ImmunoSpot 7.0.28.5 software (Cellular Technology). Neutralizing IC₅₀s were calculated using the GraphPad Prism four-parameter logistic curve fitting function.

The panel of mAbs isolated from human subjects was competed for binding using the A/California/04/2009 HA protein on the OctetRED384 system as previously described (24). Anti–penta-HIS biosensors (Sartorius) were immersed in kinetics buffer (PBS + 0.5% BSA + 0.05% Tween 20) for 60 s to obtain a baseline reading. Biosensors were then loaded with 100 μg/ml A/California/04/2009 HA protein diluted in kinetics buffer for 60 s. Biosensors were returned to kinetics buffer for a baseline of 60 s. Following this, biosensors were immersed in the first mAb (100 μg/ml in kinetics buffer) for 300 s for the association step. The biosensors were then immersed in the competing, second mAb (100 μg/ml in kinetics buffer) for 300 s. The biosensors were then regenerated in 0.1 M glycine (pH 2.7) and PBS alternately for three cycles before proceeding to the next mAb competition set. The extent of competition was calculated as the percentage of the signal from the second mAb in the second association step in the presence of the first mAb to that of the second mAb alone in the first association step for all biosensors. A ratio of ≤33% was considered complete competition, >33 and ≤67% moderate competition, and >67% no competition.

To measure Ab-dependent phagocytic activity, 2 × 10⁹ 1-µm NeutrAvidin-coated yellow-green FluoSpheres (Invitrogen, F8776) were resuspended in 1 ml of 0.1% PBS. The FluoSpheres were then centrifuged at 5000 rpm for 15 min, 900 μl of supernatant was removed, and the FluoSpheres were resuspended with 900 µl of 0.1% PBS. This process was repeated for a second wash, and then the FluoSpheres were resuspended with 20 µg of biotinylated Y2 protein. The FluoSpheres were then incubated overnight at 4°C, protected from light, with end-to-end rocking. Next, HA-specific Abs were diluted in complete RPMI medium (RPMI 1640 + 10% FBS) to a final concentration of 1 µg/ml in a U-bottom 96-well plate. Then, 20 µl of Ab dilution was transferred into a clean F-bottom 96-well plate, and 10 µl of FluoSpheres was added with the Ab followed by a 2-h incubation at 37°C for opsonization. After 1.5 h, THP-1 cells were centrifuged at 200 × g for 5 min, washed once with PBS, and then resuspended in culture medium (RPMI 1640 + 10% FBS) at a concentration of 5 × 10⁵ cells/ml. Then, 200 µl of cells was added to each well and incubated at 37°C with 5% CO₂ while shaking for 6 h. Once the incubation finished, the plate was then centrifuged at 2000 rpm for 5 min. Then, 100 µl was pipetted out of each well and replaced with 100 µl of cold 4% paraformaldehyde to fix the cells. The plate was then left at room temperature for 20 min, protected from light. The plate was then stored at 4°C in the dark. Cells were then analyzed with a NovoCyte Quanteon flow cytometer.

For electron microscopy studies, Y2 HA COBRA was cloned using Gibson assembly into a derivative of pcDNA3.1+ (29). Plasmids for the P1-05 H and L chain were synthesized (GenScript) and cloned into pcDNA3.1+. Cells and media were purchased from Thermo Fisher Scientific/Life Technologies unless stated otherwise. Y2 HA protein expression was initiated by transfection of endotoxin-free DNA into Chinese hamster ovary (CHO)-S cells using flow electroporation technology (MaxCyte). Transfected cells were suspended in CD OptiCHO supplemented with 2 mM GlutaMAX, HT, 0.1% pluronic acid, and incubated at 37°C, 8% CO₂, 85% humidity in an orbital shaker (Kuhner). After 24 h, cultures were supplemented with 1 mM sodium butyrate, and the culture temperature was dropped to 32°C. Cultures were supplemented daily with MaxCyte CHO A feed (0.5% yeastolate, 2.5% CHO-CD efficient feed A, 2 g/l glucose, 0.25 mM GlutaMAX). The media were harvested 8–12 d posttransfection and filtered. For purification of Y2, media were diluted with an equal volume of buffer A (500 mM NaCl, 20 mM sodium NaH₂PO₄, 20 mM imidazole) and loaded onto a 1-ml HisTrap column (GE Healthcare). The column was washed with buffer A and the protein was eluted with a gradient to buffer B (500 mM NaCl, 20 mM NaH₂PO₄, 500 mM imidazole) on an ÄKTA pure chromatography system (GE Healthcare). Fractions containing the protein were pooled, concentrated, and further purified and buffer exchanged on a Superdex 200 10/300 column (GE Healthcare) equilibrated in PBS (Sigma-Aldrich). Fractions were pooled and concentrated, then flash-frozen in liquid N₂ and stored at −80°C until use. For P1-05, the mAb was purified using a 1-ml HiTrap protein A high performance column (GE Healthcare). The media were diluted with an equal volume of protein A IgG binding buffer (Thermo Scientific) and loaded onto the column. The column was washed with binding buffer, then eluted with a gradient to protein A IgG elution buffer (Thermo Scientific). To adjust the pH, 55 μl of 1.89 M Tris (pH 8) was added per 1-ml fraction. Fab was generated and purified using the Pierce Fab preparation kit according to the manufacturer’s instructions (Thermo Scientific). The Fab product in PBS was flash frozen in liquid N₂ and stored at −80°C until use.

Y2 COBRA was cloned into the pBacPAK8 vector in frame with an N-terminal gp67 signal sequence and C-terminal thrombin cleavage site, T4 fibritin domain, and hexahistidine/Strep-tag II tags. The construct design results in predicted vector-supplied sequences of AATNA and LVPRGSPGSGYIPEAPRDGQAYVRKDGEWVLLSTFLGHHHHHHGGSWSHPQFEK at the N and C termini, respectively. Baculovirus was generated using the flashBac kit according to the manufacturer’s instructions (Mirus Bio). The protein was expressed in 2 l of Sf9 cells at 2 × 10⁶ cells/ml maintained in ESF921 medium (Expression Systems) by adding 25 ml of virus per liter of culture. The medium was harvested after 3 d, pH adjusted with NaCl and Tris (pH 8), and stored at −20°C. Prior to purification, the thawed medium was filtered and concentrated to 150–200 ml by tangential flow with a Vivaflow 200 (Sartorius). The resulting sample was diluted with an equal volume of buffer A, filtered, and loaded onto a 5-ml HisTrap column (GE Healthcare). The column was washed with buffer A, and the protein was eluted with a gradient to buffer B. The protein was pooled, concentrated, and supplemented with 5% glycerol prior to flash freezing and storage at −80°C.

Biolayer interferometry kinetic assays were performed in triplicate on the Octet Red384 system (Sartorius) with a buffer containing PBS, 1% BSA, and 0.05% Tween 20. Anti–penta-HIS biosensors were immersed in buffer for 120 s, then loaded with 10 μg/ml Y2 for 300 s. The biosensors were then dipped into buffer for 120 s to obtain a baseline and dipped into buffer containing P1-05 Fab in a dilution series ranging from 54 to 0.67 nM for 300 s, and buffer for 600 s to measure dissociation. The data were processed in the Octet data analysis HT software v7 (Sartorius). Each curve was reference subtracted, aligned to the baseline, and aligned for interstep correction through the dissociation step for each curve. Each replicate was fit globally for well-resolved curves in the dilution series using a 1:1 binding model. Parameters were optimized based on the R², χ², and individual K_D error values to maximize the goodness of fit. The final reported K_D value (98.5 ± 32.3 pM) represents the mean ± SD of three independent experiments.

The protein samples were thawed on ice. To generate the immune complex, P1-05 Fab and Y2 COBRA produced in CHO cells were combined in a 3:1 Fab/HA trimer ratio and incubated at room temperature for 1 h. For negative stain analysis, the immune complex was deposited at 15 μg/ml onto carbon-coated, glow-discharged, 400-mesh copper grids (Electron Microscopy Sciences) and stained with 2% w/v uranyl formate. The sample was imaged on an Arctica Talos 200C electron microscope (FEI) operating at ×73,000 nominal magnification with a Falcon II direct electron detector and a Ceta 4k camera (FEI). Micrographs were collected with Leginon, and particles were picked using a difference of Gaussian particle picker and processed with Appion (30–32). Particles were classified in two and three dimensions in Relion 3.0 and cryoSPARC2 and reconstructed in three dimensions in cryoSPARC2 (33, 34). Figures were made in UCSF Chimera (35).

PBMCs were stained with the following Abs and proteins for flow sorting: anti–CD19-allophycocyanin (1:10 dilution, clone HIB19, catalog no. 982406, BioLegend), anti–IgD-FITC (1:20 dilution, clone IA6-2, catalog no. 348206, BioLegend), anti–IgM-FITC (1:20 dilution, clone MHM-88, catalog no. 314506, BioLegend), Ghost Dye red (1:1000), Y2-PE (1:20 dilution), and Y2-BV605 (1:20). AviTagged Y2 COBRA HA proteins containing the Y98F mutation to reduce sialic acid binding were biotinylated using the BirA biotin-protein ligase in the BirA500 kit (Avidity) and complexed to streptavidin (SA) fluorophores SA-PE (1:500 dilution, catalog no. S866, Thermo Fisher Scientific) and SA-BV605 (1:250 dilution, catalog no. 405229, BioLegend). CD19⁺IgM/IgD⁻PE⁺BV605⁺ double-positive, Ag-specific B cells were flow sorted on the MoFlo Astrios and resuspended in PBS + 0.04% BSA. These cells were then used to generate Single Cell 5′ v2 dual index V(D)J libraries using the 10x Chromium Next GEM Single Cell 5′ reagent kit v2 (10x Genomics). Libraries were then sequenced using a NextSeq 550 sequencer (Illumina). Single-cell V(D)J FASTQ files were generated and demultiplexed using Cell Ranger v4.0.0, and data were visualized using the Loupe VDJ v3.0.0 browser. Only B cells with complete, ungapped variable regions and singly paired H and L chains were considered for downstream analysis. Six BCR sequences were expressed as recombinant IgGs in 293 cells and used for ELISAs and HAIs.

The three-dimensional reconstruction of Y2 + P1-05 was deposited to the Electron Microscopy Data Bank under deposition ID EMD-26586 (https://www.ebi.ac.uk/emdb/EMD-26586).

To determine the size of the H1 COBRA HA–reactive B cell population within seasonally vaccinated individuals, total B cells from four vaccinated subjects (2017–2018 cohort) were stimulated on an irradiated feeder layer as previously described (24). B cell supernatants were assayed for activity against A/California/04/2009 HA, P1 COBRA HA, and X6 COBRA HA recombinant proteins by ELISA. As expected, HA- and COBRA HA–reactive IgG titers were higher in day 21 postvaccination samples compared with those obtained prevaccination (Fig. 1). Comparisons of A/California/04/2009 HA–reactive IgG titers to those of P1 HA– and X6 HA–reactive IgGs indicated that binding to A/California/04/2009 HA protein was consistently higher. Most subjects demonstrated significant P1 HA–reactive IgG titers that, although lower than A/California/04/2009 HA–reactive IgG titers, were higher than or equivalent to X6 HA–reactive IgG titers in three of four subjects. The disparity in Ab titers between wild-type A/California/04/2009 HA– and COBRA HA–reactive proteins may be attributed to the relatively high abundance of pandemic strain–specific Abs as well as the absence of these potential binding epitopes on the P1 and X6 H1 COBRA HA proteins. Moreover, the degree of similarity of each COBRA HA to the A/California/04/2009 HA appeared to be reflected in the degree to which reactive IgG titers were elicited. Namely, the P1 COBRA HA, representing pandemic-like human and swine H1 HA sequences, demonstrates 84.63% identity to A/California/04/2009 HA, whereas the X6 HA, representing seasonal-like H1 HA sequences, demonstrated a lower 80.53% identity to A/California/04/2009 HA.

To further probe the pre-existing B cell response to COBRA HA Ags, we isolated 26 mAbs from five additional human subjects vaccinated with the 2017–2018 quadrivalent influenza vaccine within the same cohort, using A/California/04/2009 HA and P1 COBRA HA as screening Ags. The Ab-encoding genes were sequenced, and the results indicated the usage of several different Ig V genes across the entire panel (Fig. 2, Supplemental Table I). When comparing usage of H chain genes, V_H1, V_H3, V_H4, and V_H5 gene families were represented (Fig. 2A). ∼50% of all mAbs used a gene from the V_H3 family, and approximately another 50% used a gene from the V_H4 family. In the L chain, for those mAbs utilizing the κ-chain, many used genes V_K3-11 and V_K3-15. The remainder used V genes V_K3-20 or those from V_K1 or V_K2 families. mAbs utilizing the λ-chain used predominantly V_L3-21 and V_L2-14. Paired H and L chain V genes showed variation across the Ab panel, with the V_H3-7:V_K3-15 and V_H4-39:V_L2-14 pairings being the most abundant for κ- and λ-chain–utilizing mAbs, respectively (Fig. 2B). The CDR3 regions ranged in length from 10 to 22 aa for the H chain, 8 to 10 aa for the κ-chain, and 10 12 aa for the λ-chain (Fig. 2C). The percent identities of the variable genes to the germline sequence had averages of 93% for both the H and κ-chains, and 96% for the λ-chain at the nucleotide level (Fig. 2C).

P1 and X6 COBRA HA proteins incorporate historical epitopes from both pre- and post-2009 viruses, whereas Y2 COBRA HA incorporates more recent 2009 pandemic-like sequences from 2014 to 2016, representing the HA epitopes from more recent H1 isolates (22) (Fig. 3). Most isolated mAbs demonstrated high binding to A/California/04/2009 HA protein by ELISA, with an average EC₅₀ of 30 ng/ml (Fig. 4A, Supplemental Fig. 1A). Of these A/California/04/2009 HA protein-reactive mAbs, only a subset demonstrated binding to the divergent P1 and X6 COBRA HA proteins. mAbs P1-02, P1-05, and 163-20 showed reactivity against the P1 COBRA protein, and mAbs CA09-26, CA09-30, CA09-45, P1-02, and P1-05 demonstrated binding to the X6 COBRA protein. The limited mAb binding to P1 and X6 COBRA proteins correlated with the lower reactive B cell frequencies to these respective HAs in (Fig. 1. All mAbs had similar EC₅₀ values and reactivities to the Y2 COBRA HA compared with A/California/04/2009 HA (Fig. 4A, Supplemental Fig. 1A). No binding was observed for any mAb to the H3 subtype HA HK14 or to an irrelevant Ag control (Supplemental Fig. 1A). We also determined whether mAbs were broadly reactive by utilizing a chimeric HA protein bearing a H6 HA head and a H1 HA stem (cH6/1) (Fig. 4A). mAb P1-05 bound to the chimeric protein with high avidity, suggesting that this mAb may bind the stem region of the H1 HA protein, or some other conserved epitope present in both H1 and H6 HAs. These results indicate that 2017–2018 QIV-vaccinated subjects possessed mAbs with potent binding to the 2009 pandemic-like Y2 COBRA HA protein. Moreover, mAbs from different subjects with reactivity against the divergent P1 and/or X6 COBRA HAs were found, although these represented a small part of the total Ab response. Based on these data and the B cell screening data, COBRA HA–reactive B cells constitute part of the human B cell response to influenza vaccination, and COBRA HA Ags can likely recall such B cells targeting both the head and stem regions.

To characterize the functional activities of the isolated mAbs, HAI and neutralizing activities were assessed (Fig. 4B, 4C, Supplemental Fig. 2). Most mAbs showed HAI activity against the pandemic-like A/Michigan/45/2015 virus (Fig. 4B). These data are consistent with the fact that most mAbs bound the head domain of A/California/04/2009 HA (Fig. 4A). Of those mAbs with the highest HAI activity of the panel against the recent pandemic-like A/Michigan/45/2015 virus, CA09-26 and CA09-45 were tested for HAI against two prepandemic H1 viruses, as these mAbs bind the X6 HA COBRA, which incorporates prepandemic sequences, and both target the HA head domain. CA09-26 had HAI activity against A/New Caledonia/20/1999 and A/Brisbane/59/2007 viruses, while CA09-45 had HAI activity against A/New Caledonia/20/1999, and no activity against A/Brisbane/59/2007 (Fig. 4B). We next assessed neutralizing activity against A/California/07/2009 (Fig. 4C, Supplemental Fig. 2). ∼60% of mAbs (16/26 mAbs) neutralized the pandemic A/California/07/2009 virus (A/CA/09). Notably, mAbs CA09-26 and CA09-45 were among the most potent mAbs in the panel with IC₅₀s of 0.013 and 0.032 μg/ml, respectively. These two mAbs were also tested for neutralizing activity against A/New Caledonia/20/1999 and A/Brisbane/59/2007 (Fig. 4C). They had IC₅₀ values of 0.081 and 0.286 μg/ml, respectively, against A/New Caledonia/20/1999, indicating potent neutralization activity (Supplemental Fig. 2). However, these two mAbs did not demonstrate neutralizing activity against the A/Brisbane/59/2007 virus, in accordance with the observation that little to no HAI activity was observed for the same strain.

Neutralization-independent, Fc-dependent activities are an important aspect of anti-influenza Abs that bind both the head and stem domains (36–38). Stem-binding Abs elicited by P1 HA vaccination also demonstrate Fc activity by inducing cellular cytotoxicity (20). To determine the extent of one such Fc effector function, Ab-dependent phagocytosis (ADP) activity was measured by assessing the capacity for the monocytic THP-1 cell line to phagocytose Y2 COBRA HA–coated beads through mAb binding (Fig. 4D, (Fig. S3A, S3B). The entire mAb panel demonstrated ADP activity relative to a negative mAb control. These included both neutralization/HAI-positive mAbs as well as mAbs that did not demonstrate significant HAI or neutralization activity.

To determine the epitopes bound by the panel of 26 mAbs isolated from these vaccinated subjects, biolayer interferometry-based epitope binning was performed as previously described (24, 25). Biosensors were loaded with A/California/04/2009 HA, associated with one mAb, and then exposed to a second mAb to determine mAb competition (Fig. 5). Control mAbs Ab6649, 5J8, and CR6261 were used to determine the relative locations of each epitope. Ab6649 binds the lateral patch, proximal to the Sa antigenic site; 5J8 binds the RBS, comprising antigenic sites Sb and Ca2; and CR6261 binds a conserved portion of the stem region found for all group 1 viruses (Fig. 5A). Five distinct epitopes on A/California/04/2009 HA protein were distinguished (Fig. 5B). Of the epitopes on the globular head domain, two known major epitopes, termed epitope 1 and epitope 3, corresponding to those of mAbs Ab6649 and 5J8, respectively, were identified. mAbs competing with Ab6649 were comprised in part of those using the H chain gene V_H3-23, including mAbs CA09-19 and CA09-29, similar to those described recently that bind the lateral patch, although these mAbs did not contain the characteristic YXR motif within the H chain CDR3 (39). The position of one predominantly bound epitope, epitope 2, could not be identified by epitope binning with the control mAbs used. Two other epitopes, characterized only by the competition of a single mAb to itself, were epitopes 4 and 5, which correspond to mAbs CA09-38 and P1-05, respectively. No mAbs competed with CR6261, indicating that although mAb P1-05 may target the stem, as evidenced by binding to the cH6/1 protein (Fig. 3, Supplemental Fig. 1A), this mAb might target a different epitope on the stem of the H1 HA protein. Overall, these data suggest that the epitopes bound by mAbs from vaccinated subjects are likely comprised, in part, of regions within or around conserved sites on the head domain, such as those involving the RBS and the lateral patch, in addition to portions of the stem.

Epitope 5, characterized by binding by mAb P1-05, was likely located on the stem of the H1 HA protein yet did not overlap with other Ab epitopes (Fig. 5). To determine the epitope of mAb P1-05, we generated a complex of Y2 HA bound to P1-05 Fab fragments and evaluated its structure by negative-stain electron microscopy (Fig. 6A, 6B). The two-dimensional class averages revealed that P1-05 binds to the base of the HA stem in an upward angle (Fig. 6B). We also observed that insertion of residues between the Y2 C terminus and the foldon trimerization domain disrupted mAb P1-05 binding, potentially due to trimer splaying and disruption of this membrane-proximal epitope (Fig. 6C). Recently, a similar class of mAbs targeting this region on HA, termed the “anchor” epitope, was discovered, and such mAbs protect against H1N1 infection in mice (40, 41). Anchor mAbs do not compete with known stem mAbs and use V_K3-11 or V_K3-15 κ V genes that encode a germline-encoded NWP motif in the CDR3 region (41). The restricted L chains can pair with V_H3-23, V_H3-30/V_H3-30-3, or V_H3-48 V genes. mAb P1-05 utilizes V_K3-11 paired with V_H3-23 and also possesses the NWP motif. Furthermore, it was also recently reported that binding of anchor mAbs is disrupted by the use of a GCN4 trimerization domain (41), which has different spacing than the foldon domain, which matches our data with the disruption of binding observed in (Fig. 6C. These data affirmed that P1-05 binds to the anchor epitope in the stem, consistent with data from (Fig. 3A, rather than another conserved HA epitope in the head domain. These observations are critical for subunit HA protein vaccine development as they indicate the importance of Ag design, stability, and the incorporation of mAb binding avidity studies to ensure that important epitopes are properly displayed on candidate vaccine Ags.

To further probe the repertoire of pre-existing COBRA HA-specific B cells in a more recent vaccine season, we conducted a single-cell RNA sequencing experiment using B cells from a single subject vaccinated with the 2019–2020 seasonal influenza vaccine. ∼3000 CD19⁺IgM⁻IgD⁻ B cells positive for the Y2 COBRA HA were sorted and subjected to 10X barcoding (Fig. 7, Supplemental Fig. 3C). Prior to loading onto the 10x controller, sorted Y2-specific cells were supplemented with the CA09-26 hybridoma clone as a loading control. One hundred nineteen unique paired H and L chains were obtained following data demultiplexing and analysis compared with the human genome database. Similar to the mAb sequencing, the V_H1 and V_H4 gene families were highly prevalent in the B cell repertoire. In particular, V_H4-39 and V_H4-59 were prevalent in both mAb sequencing and B cell sequencing results. We also identified several additional mAbs utilizing the V_H3-23 gene, with one in particular (clone 70) having an NWP motif in a paired V_K3-15 L chain, although this clone used a J_K2 gene, rather than the J_K4 or J_K5 gene, which were previously used to identify anchor epitope-specific mAbs (41) (Supplemental Table I). A fraction of clones using the V_H1-69 gene were also identified, which is used by mAbs targeting the stalk epitope. Hence, pre-existing mAbs binding Y2 use a relatively diverse repertoire.

To identify whether these B cell clones were reactive against the baiting Y2 Ag, we recombinantly expressed six selected clones, 32, 58, 60, 70, 73, and 86, as rAbs. We then evaluated binding to the Y2 and cH6/1 HAs, as well as HAI activity against the more recent 2019 H1N1 strain A/Guandong-Maonan-SWL-1536/2019 (A/GM/19) (Fig. 7E, 7F, Supplemental Fig. 1B). All rAbs bound the Y2 COBRA HA with high affinity (Fig. 7E) but had no significant activity against the H1 HA stem as assessed by cH6/1 HA binding, including clone 70, which possessed the NWP motif but used the J_K2 gene. To evaluate functionality, we performed HAI assays against the recent A/GM/19 virus. All but one rAb, 70, possessed HAI activity against this strain (Fig. 7F). These data suggest that Y2-reactive Abs isolated from PBMCs from a 2019–2020 seasonally vaccinated subject generally possess functionality against recent 2009 pandemic-like H1N1 strains, including a drifted 2019 virus. Importantly, this virus represents a strain not included in the original design period of the Y2 COBRA HA (2014–2016), yet rAbs derived from Y2-reactive B cells possessed HAI activity, indicating functionality against a future H1N1 strain with the COBRA HA platform.

H1 COBRA HA Ags have been successful at broadening the Ab response compared with wild-type HA sequences in naive and preimmune mouse and ferret models of influenza infection (21, 22). However, pre-existing immunity to influenza in humans remains a major challenge to overcome due to repeated previous exposure to the influenza HA protein during infection and vaccination events. In this study, we sought to determine the extent of the H1 subtype COBRA HA–reactive pre-existing B cell repertoire in human subjects to predict recall responses as COBRA HA Ags move toward clinical trials. At the oligoclonal B cell level, pre-existing B cell responses were observed for P1 and X6 COBRA Ags in individuals vaccinated with the 2017–2018 seasonal influenza vaccine, which incorporated the pandemic-like A/Michigan/45/2015 vaccine strain. COBRA HA–reactive B cell responses were lower than those observed for A/California/04/2009 HA protein, likely due to the loss of strain-specific variable head epitopes and incorporation of seasonal prepandemic and swine HA sequences in the X6 and P1 Ags, respectively. Although it is likely that seasonal vaccination induced these COBRA HA–reactive B cells, it is possible that prior exposures to H1 viruses before vaccination may also have induced these HA-reactive B cell subsets. Moreover, we examined Ab responses to COBRA HAs in a small cohort of five vaccinated subjects, which may limit our conclusions on the extent of pre-existing Abs to H1 COBRA HAs in the general population. Only a small subset of mAbs isolated against A/California/04/2009 HA reacted with P1 and X6 COBRA Ags. In contrast, the mAb binding profile to the recently described Y2 COBRA HA, which incorporates more recent 2009 pandemic-like H1 sequences from 2014–2016, was similar to that observed for the A/California/04/2009 HA protein. The presence of such Y2-reactive mAbs may indicate specificities against common epitopes present only in more recent 2009 pandemic-like strains but not in historical pre-2009 viruses, which could be more relevant for immunity against future H1 virus exposures. Most of the mAbs had HAI activity and neutralizing activity against A/Michigan/45/2015 and A/California/07/2009, and two head-binding mAbs that bind the X6 protein, CA09-26 and CA09-45, had HAI activity and neutralizing activity against the prepandemic strain A/New Caledonia/20/1999. These data suggest that X6 HA–reactive mAbs are mainly endowed with functional activity against both prepandemic and pandemic-like H1 viruses. Overall, the amino acid similarity of COBRA HA Ags to A/California/04/2009 HA correlated with high B cell and mAb reactivity. In addition to binding, neutralization, and HAI activity, we also assessed whether COBRA HA–reactive mAbs had Fc-mediated functions, namely ADP, and all mAbs were able to induce THP-1 phagocytosis of Y2-coated beads.

Several epitopes on the H1 HA protein have been previously defined (9), and we determined mAb epitopes on the A/California/04/2009 HA protein using biolayer interferometry. Most of the mAbs targeted three head-binding epitopes on the Sa and Sb/Ca2 sites, and an undefined epitope, epitope 2, which is currently under structural characterization. One limitation of the BLI-based binning assay is that it does not definitively characterize Ab breadth and functionality, only providing data on the relative positions of epitopes. For instance, we observed a significant number of mAbs competing with the widely reactive 5J8 and Ab6649 Abs, yet most isolated mAbs competing with 5J8 and Ab6649 did not bind as widely, with most lacking reactivity to the divergent P1 and X6 COBRA HAs. mAb CA09-38 did not exhibit HAI or neutralizing activity, nor did it bind the cH6/1 HA, suggesting that this mAb may target an undefined, non-neutralizing epitope on the head. We discovered that P1-05 targets a unique epitope on the H1 HA stem region, and this epitope is similar to the recently described anchor epitope (41). Based on these data, although COBRA Ags were primarily designed to induce broadly reactive Abs to the head domain, these Ags will likely also recall broadly reactive anchor mAbs in humans in addition to head-based recall and de novo Ab responses. Further repertoire analysis in a subject vaccinated with the 2019–2020 seasonal influenza vaccine identified similar sequences to our mAbs targeting the head domain from the 2017–2018 season, which possessed HAI activity against a drifted 2019 H1N1 virus, A/Guangdong-Maonan/SWL-1536/2019, indicating that the COBRA-reactive B cell population may be similar across subjects and influenza vaccine seasons.

These data provide evidence that a preimmune population with exposure to the seasonal influenza virus vaccine exhibits B cell reactivity toward conserved epitopes present on COBRA HA Ags. As the COBRA HA platform enters clinical trials, it is likely that head-specific and some stem-specific Abs will be elicited as part of a recall response. Moreover, the Ab epitopes identified in this work overlap in part with those previously identified on the head domain near the RBS and the lateral patch, in addition to those on the stem. These epitopes are the focus of future structural studies, particularly for those mAbs that cross-react with the X6 COBRA HA as well as with the HA stem domain. Our data also exemplify the importance of structural analysis of protein epitopes to ensure epitopes that elicit broadly neutralizing Abs, such as the anchor epitope, remain intact following design optimization for subunit HA vaccines.

The authors have no financial conflicts of interest.