Structural basis for the broad antigenicity of the computationally optimized influenza hemagglutinin X6

Summary

Influenza causes significant morbidity and mortality. As an alternative approach to current seasonal vaccines, the computationally optimized broadly reactive antigen (COBRA) platform has been previously applied to hemagglutinin (HA). This approach integrates wild-type HA sequences into a single immunogen to expand the breadth of accessible antibody epitopes. Adding to previous studies of H1, H3, and H5 COBRA HAs, we define the structural features of another H1 subtype COBRA, X6, that incorporates HA sequences from before and after the 2009 H1N1 influenza pandemic. We determined structures of this antigen alone and in complex with COBRA-specific as well as broadly reactive and functional antibodies, analyzing its antigenicity. We found that X6 possesses features representing both historic and recent H1 HA strains, enabling binding to both head- and stem-reactive antibodies. Overall, these data confirm the integrity of broadly reactive antibody epitopes of X6, and contribute to vaccine design efforts for a next-generation vaccine.

Graphical Abstract

eTOC

Current influenza vaccines confer narrow, relatively strain-specific protection. Through analyses of a next-generation vaccine, X6, Nagashima et al. describe the characteristics responsible for enhanced antibody-dependent breadth. Glycans and antibody-binding sites of the major surface protein, hemagglutinin (HA), were conserved from circulating strains from before and after the 2009 H1N1 pandemic.

Introduction

Influenza virus poses a major health concern worldwide. It is estimated that within the United States, between 140,000 and 810,000 hospitalizations due to the virus occur each year¹. Seasonal influenza vaccines provide only limited effectiveness against circulating strains due to antigenic drift within the hemagglutinin (HA) glycoprotein, the major viral surface antigen³. Moreover, the protective antibody response elicited by the vaccine is short-lived. Therefore, the seasonal vaccine is sensitive to viral immune evasion and must be reformulated every year based on predictions of circulating viruses. Timelines for manufacturing the current vaccine can take up to six months⁵ and accurate strain prediction remains a challenge. Delaying strain selection may enable more accurate strain prediction, but usually does not occur, as this reduces vaccine yield and availability.

Several next-generation influenza vaccines have been investigated to elicit enhanced breadth against a wider range of virus strains⁷. One approach, termed COBRA (computationally optimized broadly reactive antigen), employs a layered consensus building strategy to combine wild-type viral sequences into a single antigen⁹. This methodology has been applied to the HA protein for influenza A and B viruses^{9,11,13,14,17}. The resulting COBRA HA possesses enhanced effectiveness as a vaccine antigen, whereby studies in mouse and ferret animal models have shown increased functional antibody breadth following COBRA HA vaccination^9,19. The primary mechanism by which these COBRA HAs elicit protection is through inducing broadly HAI-active, HA head domain-targeting antibodies that block receptor binding^9,21,23. While it was hypothesized that stem domain-binding antibodies did not contribute significantly to the COBRA HA response²⁴, we have previously shown that the COBRA HA stem is intact for an H1 HA COBRA, Y2, which both retains binding to anchor epitope-specific antibodies and induces central stem-binding antibodies^25,26. Similar stem antibodies have been isolated from H1 COBRA-vaccinated mice²⁷. Furthermore, seasonally vaccinated subjects possess antibodies that can bind an H3 COBRA HA²⁸. This suggests that COBRA HAs possess the potential to elicit antibodies that bind both the HA head and stem domains.

The 2009 H1N1 influenza pandemic marked a significant antigenic shift in circulating H1N1 viruses. As a reassortant virus resulting from swine, avian, and human strains, the 2009 pandemic virus HA possessed a novel head domain and a more conserved stem relative to circulating pre-2009 seasonal strains²⁹. Infection with this 2009 virus elicited stem-reactive antibodies that are thought to have led to the disappearance of pre-2009 seasonal H1N1 viruses and the subsequent dominance of the pandemic-like strains³⁰. The X6 COBRA HA has been previously characterized in animal models to elicit broadly reactive serum antibodies against viruses of the H1 subtype³¹. This antigen incorporates sequence elements from both pre-2009 pandemic and post-2009 pandemic viruses from 1999 to 2012³¹. Immunization with X6 has been shown to elicit polyclonal serum antibodies with broad reactivity against several H1 viruses¹¹. Seasonal vaccination can elicit monoclonal antibodies (mAbs) in human subjects that recognize the X6 COBRA HA, indicating the conservation of epitopes between the seasonal vaccine H1 HA and the X6 COBRA HA²⁵. One such mAb, CA09-26, has neutralization activity against both pre-2009 and 2009 pandemic H1 viruses through binding to the receptor-binding site (RBS)²⁵. In this study, we demonstrate that other human and murine mAbs, #58 and BE1, show similarly broad H1 reactivity and HAI activity across multiple H1 HAs as a consequence of binding the RBS³² and lateral patch epitopes^33,34. In addition, we show that conserved H1 HA epitopes are also present within the X6 stem domain by solving the structure of the Fab fragment of a group 1 broadly reactive central stem-binding mAb, CR6261³⁵, in complex with X6. Altogether, these data indicate that the COBRA X6 vaccine incorporates conserved and intact antibody epitopes on the HA protein. These structural analyses provide insights into the mechanism by which COBRA HAs induce enhanced antibody breadth.

Results

Sequence and structural features of the X6 COBRA HA

The design of COBRA X6 includes human-tropic H1 viruses from 1999 to 2012, spanning both pre-2009 and post-2009 HAs in its design. Sequence alignments to wild-type HAs revealed that X6 is more similar to pre-2009 influenza HA sequences compared to post-2009 sequences (Figure 1A). X6 possesses ~97% similarity to pre-2009 HAs, and ~80% identity to post-2009 pandemic-like HAs. Moreover, its antigenic sites were generally more similar to pre-2009 HAs (Figure S1). We mapped differences between the A/New Caledonia/20/1999 (NC99) and A/Michigan/45/2015 (MI15) HAs, finding most differences to be in the head domain and a minority in the stem (Figure 1B). We also structurally mapped individual sequence differences between X6 and NC99, as well as between X6 and MI15 (Figure 1C). We found that the higher divergence of X6 from MI15 was attributable to differences localized predominantly within the head domain, whereas minimal differences in the head or stem domain were found when structurally aligning X6 to NC99. This comparison affirmed that X6 possesses greater sequence similarity to pre-2009 H1 HAs compared to post-2009 H1 HAs.

Figure 1. Sequence and structural comparisons of pre-2009 and 2009 pandemic-like HAs to X6.

(A) Heat map of overall percent identity of pre-2009 and post-2009 HAs to X6. (B-C) Structural comparisons between HAs, with mutated residues colored from blue to red based on low to high amino acid similarity as measured by the BLOSUM90 matrix. (B) Structural comparison of the A/New Caledonia/20/1999 (NC99) and A/Michigan/45/2015 (MI15) HAs, mapped to the NC99 HA (PDB 7MFG). (C) Structural differences between X6 and NC99 (left) and X6 and MI15 (right) mapped to X6. See also Figure S1.

To characterize the structural features of X6, we determined the crystal structure of X6 to a resolution of 3.25 Å. As expected, the X6 HA formed a trimer similar to wild-type HAs (Figure 2A). Structural alignments of X6 to wild-type HAs revealed that it was also similar to natural HAs, with RMSDs of 0.914 Å for X6 to NC99, and 0.679 Å for X6 to MI15. X6 also retained glycosylation sites similar to those found on the pre-2009 NC99 and the post-2009 MI15 HAs (Figure 2B, 2C, Figure S2). Seven N-linked glycosylation sites in X6 were predicted in the NetNGlyc server³⁶. Within the crystal structure, all of these glycosylation sites were observed at residues N11, N23, N54, N87, N125, N159, and N286. Of these sites, four were found on both NC99 and MI15, found primarily on the stem domain and the side of the HA head. Two glycosylation sites, at residue N159 on the top of the globular head domain and at N54 by the bottom of the head domain, were shared only with NC99 but not with MI15. While glycans were present at N480 in NC99 and MI15, no glycan was found at this position for X6. This difference may allow for more stem-directed responses following vaccination with X6. It may be possible that X6 glycans could allow for redirection of the antibody response towards more conserved epitopes found on both pre-2009 and post-2009 pandemic-like HAs. Overall, these data suggest that the glycosylation profile of X6 represents HAs within its design period and is skewed towards pre-2009 HAs.

Figure 2. Crystal structure and glycosylation features of the X6 COBRA HA.

(A) The X6 structure is shown as a trimer, with glycans indicated as light orange spheres. (B) Individual HA monomers for X6, NC99 (PDB 7MFG), and MI15 (PDB 6XGC) are shown. Glycans for NC99 are shown in green, and those for MI15 are shown in pink. (C) Overlay of the X6, NC99, and MI15 HA structures. See also Figure S2 and Table S1.

Structures and characteristics of X6-binding broadly reactive head binding antibodies

X6 contains epitopes of monoclonal antibodies (mAbs) isolated from humans, including from seasonally vaccinated populations³⁷. These include epitopes near the RBS, which are often associated with broad hemagglutination inhibition (HAI) and neutralizing activities. These overlap with some of the classically defined antigenic sites, including Sa, Sb, and Ca2, within the HA head domain. To characterize the head-dependent antigenicity of X6, we determined 2D cryo-EM class averages of X6 with the Fab of human mAb CA09-26, and cryo-EM structures of X6 bound to the Fab fragment of human mAb #58 to 3.57 Å resolution, and bound to the mouse BE1 Fab fragment at 3.45 Å resolution (Figure 3, Figure 4, Figure S4A,B).

Figure 3. Cryo-EM analyses of HA RBS-binding antibody Fabs with X6.

(A) 2D class averages of the CA09-26 Fab:X6 complex. Particle orientations are shown in labeled boxes for each class average. (B) The electron potential map of the #58 Fab with X6. Left: side view, right: top view. (C) Overlay of the #58 epitope to the RBS on the X6 COBRA HA. (D) The interface between the #58 Fab and X6. (E) Residue comparisons in the epitope of #58 are shown for the NC99 (top) and MI15 (bottom) HAs. Mutated residues relative to X6 are shown in red. See also Figures S3, S4, and Tables S2, S3.

Figure 4. Cryo-EM structure of the HA lateral patch-binding antibody BE1 Fab with X6.

(A) The binding angle of the BE1 Fab to X6, as well as the cryo-EM map. (B) The interface between the BE1 Fab and X6 with interacting residues shown. (C-D) Residue comparisons in the BE1 epitope for (C) NC99 and (D) MI15, respectively. Mutated residues relative to X6 are shown in red. (E-G) Residue comparisons and steric clashes at position 165, shown in white, for (E) X6, (F) NC99, and (G) MI15, respectively. Clashes are shown in red with more significant clashes shown as larger discs. (H) Structural comparisons of BE1 to other lateral patch antibodies, Ab6649 and 045-09 2B05. See also Figures S4, S5 and Table S4.

CA09-26 was isolated previously from a seasonally vaccinated subject who received the 2017–2018 vaccine³⁷. It possesses both HAI and neutralizing activities against recent 2009 H1N1-like strains such as A/Michigan/45/2015 (A/MI/15) and A/California/07/2009 (A/CA/09), in addition to a seasonal pre-2009 H1N1 strain, A/New Caledonia/20/1999 (A/NC/99). Although insufficient side orientations of the CA09-26 Fab:X6 complex precluded 3D reconstruction, 2D class averages revealed binding to the X6 head domain proximal to the RBS epitope (Figure 3A). This corroborated previous biolayer interferometry-based epitope binning observations suggesting competition between 5J8, a known RBS-binding antibody³², and CA09-26³⁷.

mAb #58 was also isolated previously from a human subject, D160, who was vaccinated in 2013 and 2014 with the Fluzone seasonal vaccine. Serum from this subject possessed binding and HAI activity to H1 HAs and H1N1 viruses (Figure S3A,B). #58 was isolated following the 2014 vaccination, and possesses binding activity against some pre-2009 and post-2009 HAs, significant HAI activities for A/NC/99 and A/MI/15, and neutralizes the A/CA/09 2009 pandemic strain (Figure S3C,D,E). Similar to CA09-26, mAb #58 also bound to the X6 head domain at the RBS (Figure 3B). The contacts made by this mAb overlapped with the periphery of the RBS, near the top at the 190-helix and the bottom in the 130- and 220-loops (Figure 3C). mAb #58 also bound the X6 head domain and used both heavy and light chain residues to contact the RBS (Figure 3D, S4A). Specifically, R188 of X6 may interact through polar interactions with the side chain of Y38 in HCDR1. V111.3 of HCDR3 may also participate in electrostatic interactions with X6, making contacts with the main chain of V131. In addition, D112.3 of HCDR3 likely makes hydrogen bonding contacts with the side chain of Q222 in X6. Moreover, H109 of the mAb #58 LCDR3 loop participates in hydrogen bonding with the side chain of D186.

Three of four residues in the epitope of mAb #58 were conserved in both the pre-2009 NC99 and post-2009 MI15 HAs, permitting binding (Figure 3E). In NC99, the residue at position 186 is an asparagine, mutated from D186 in X6. This D186N mutation may still permit hydrogen bonding interactions with the light chain H109 as both are similarly polar residues. In contrast, NC99 retains both R188 and V131, likely permitting electrostatic interactions with Y38 and V111.3 of #58, respectively. MI15 possesses a glutamine at position 188, mutated from R188, which may still permit polar interactions with Y38 of the heavy chain. However, both D186 and V131 are conserved as well in the #58 epitope of this more recent 2009 pandemic-like strain. Both strains possess a conserved Q222 that likely permits binding to D112.3 in the #58 HCDR3. Overall, despite some sequence flexibility in the #58 epitope between pre-2009 and post-2009 viruses, the overall set of interactions with this antibody is conserved.

We isolated mAbs from X6-vaccinated mice and characterized their binding and HAI activity to H1 HAs and H1 viruses (Figure S5). We found mAbs possessing binding to pre-2009 seasonal H1 HAs, but not to post-2009 pandemic-like HAs. These mAbs, 1G8 and 5CA9, did not bind the chimeric H6/1 (cH6/1) HA, which possesses the H6N1 (isolate A/mallard/Sweden/81/2002) head and the HA stem of 2009 pandemic H1N1 (isolate CA/09), suggesting binding to a head epitope. They also lacked HAI activity except for A/Phil/13 for mAb 5CA9, suggesting binding to a conserved, non-RBS head domain epitope. Another mAb, BE1, was also isolated, which similarly possessed binding activity against HAs from before the 2009 H1N1 pandemic and did not bind the cH6/1 HA (Figure S5A). Moreover, BE1 did not show significant HAI activity for most H1N1 viruses except for low activity against A/SI/06, suggesting that it bound a non-RBS epitope (Figure S5B). Therefore, to elucidate the epitope of this class of antibodies, we determined the structure of the BE1 Fab fragment in complex with X6, resulting in an HA trimer bound to two Fabs. From the cryo-EM map, we found that it bound nearly horizontally to a non-RBS epitope on the side of the head domain distal to the RBS (Figure 4A). BE1 uses both heavy and light chain residues to interact with its epitope (Figure 4B). Heavy chain interactions with the epitope involved Y111 and Y112.1 in HCDR3. Y111 participates in polar interactions with the side chain of N166 in X6, as well as with the main chain carbonyl oxygen of K170. Y112.1 also participates in a polar interaction through its side chain with the main chain of E115. Light chain interactions were more extensive than heavy chain interactions for this antibody. The side chain of S65 is involved in hydrogen bonding with that of S121. K66 participates in a salt bridge with the side chain of E115, and Y56 is involved in hydrogen bonding with the main chain of I116. Y38 also participates in a hydrogen bond with the side chain of N167.

We also compared the BE1 epitope in X6 with those found in pre-2009 and post-2009 H1 HAs (Figure 4C,D). We found that all participating residues were conserved in the NC99 HA, consistent with the binding activity observed with pre-2009 H1 HAs (Figures S5A, 4C). In that of the post-pandemic MI15 HA, most residues were conserved relative to that of X6 (Figure 4D). The only mutation found in this epitope relative to X6 was a N167D mutation that may reduce the effectiveness of binding. Other structural features may be responsible for the abrogation of BE1 binding to post-2009 H1 HAs. For instance, we found that in MI15 and other post-2009-like HAs, a clash with BE1 was predicted at position 165 in antigenic site Ca1, close to other residues in the BE1 epitope, which was not found for the NC99 pre-2009 HA (Figure 4E–G). Specifically, MI15 possesses a bulky isoleucine residue that might clash mainly with Y36 and slightly with Y108 of the light chain, interfering with binding. In X6, the residue at this position is an alanine, and corresponds to a valine for NC99, which both possess sufficiently small side chains to accommodate BE1. This epitope overlapped with the previously described lateral patch, which is conserved across both pre-2009 pandemic and the 2009 pandemic H1 HAs^33,34. We also compared the binding orientation of the BE1 Fab to those of the Fab fragment of Ab6649, isolated from a subject who received a monovalent A/California/07/2009 vaccine³³, and 045-09 2B05, isolated from another subject who received a 2009 monovalent influenza vaccine³⁴ (Figure 4H). Structural alignment of the BE1 Fab to those of mAbs Ab6649 ³³ and 045-09 2B05³⁴ indeed confirmed some overlap with the lateral patch, binding at an angle between these Ab6649 and 045-09 2B05 Fabs to the head domain. Overall, these data suggest that X6 contains the RBS epitope in addition to the non-RBS lateral patch epitope, contributing to its broad reactivity.

To experimentally validate the effects of residue 167 in X6 as well as that of residue 165 in the MI15 HA on BE1 mAb binding, we generated X6 N167D and MI15 D167N mutants, as well as a X6 N167A mutant (possessing a small side chain at this position) and a X6 N167R mutant (possessing a bulky, flexible side chain). We further substituted I165 in MI15 with an alanine (as in X6) or to a valine (as in NC99), and A165 in X6 to an isoleucine (as in MI15) or to a valine (as in NC99). We then used ELISA to assess the extent of binding of BE1 to its lateral patch epitope to these point mutants (Figure 5). Whereas mutating I165 in MI15 to either an alanine or a valine did not restore binding of BE1, there was a minor decrease in binding of BE1 to X6 A165I relative to X6. Furthermore, substituting MI15 D167 with an asparagine was still insufficient to restore BE1 binding, and the N167A or the N167R mutations in X6 either did not significantly alter binding or somewhat decreased binding, respectively, relative to wild-type X6. Interestingly, the X6 N167D mutation completely abolished BE1 binding, potentially due to charge repulsion with the acquisition of a formal negative charge at this residue. We also finally assessed binding of BE1 to the pre-2009 HA NC99 and the recently described Y2 COBRA HA. We found that BE1 bound to pre-2009 HA NC99, but not to the 2009 pandemic-like COBRA HA Y2, consistent with the lack of binding to recent, post-pandemic like H1 HAs (Figure S5A).

Figure 5. Binding of BE1 to HA point mutants.

ELISAs of BE1 to X6, X6 point mutants, MI15, MI15 point mutants, and to NC99, Y2 COBRA, and Ply (negative control) antigens are shown. The data shown are the mean±SD of an experiment performed in quadruplicate. Four-parameter nonlinear fits were performed with Prism 9.

Structure and characteristics of X6-binding broadly reactive stem-binding antibodies

Stem-binding antibodies generally possess greater breadth than head-binding antibodies, which could contribute to long-lasting protection³⁸. To determine the integrity of potential X6 stem domain epitopes, we obtained a cryo-EM structure of X6 bound to the Fab fragment of CR6261, a known group 1-reactive antibody³⁵. The CR6261 Fab:X6 structure was solved to 2.64 Å resolution (Figure 6, S4C). The cryo-EM structure revealed that the CR6261 Fab bound X6 at the expected central stem epitope near the middle of the stem similar to previously characterized HAs ³⁵ (Figure 6A). We also investigated the structural features of the CR6261 Fab:X6 interaction in comparison to that of previously determined structures of the CR6261 Fab with historic, seasonal pre-2009, and recent swine-origin post-2009 H1 HAs (Figure 6B). The CR6261 Fab interacted with X6 using only its heavy chain. Significant contacts included polar interactions of the side chain of N379 in X6 with the backbone of F30 on CR6261. Q368 of X6 also interacts through polar interactions with the main chain atoms of CR6261 S36 and Y110. The side chain of Y110 also undergoes hydrogen bonding with the backbone oxygen of X6 D345. We further assessed whether CR6261 binding to X6 used similar contacts to historic, seasonal, and recent H1 HAs. When comparing the CR6261 epitope of X6 to those of the historic A/South Carolina/1/1918 (SC18), the seasonal pre-2009 A/Bayern/07/1995 (BA95), and the recent H1N1 variant (H1N1v) swine-origin A/Ohio/09/2015 (OH15)³⁹ HAs, most of these interactions were conserved. The hydrogen bond made from T375 of X6 with S36 of CR6261 HCDR1 was a novel contact not found in wild-type HAs. Additionally, the electrostatic interaction between Q111 of CR6261 to Q364 of SC18 was not found in other CR6261 epitopes analyzed here. In the SC18, BA95, and OH15 structures, S36 interacted through its side chain to the main chain of D372 in the SC18 and OH15 HAs or to the main chain of N372 in BA95 using hydrogen bonding. Nonetheless, all other contacts were found in the SC18, BA96, and OH15 HAs, indicating that overall, the CR6261:X6 interaction occurs through similar residue interactions as for historic, seasonal, and recent HAs.

Figure 6. Cryo-EM structure of the Fab of broadly reactive stem-binding antibody CR6261 with X6.

(A) The CR6261 Fab:X6 cryo-EM electron potential map for side (left) and top (right) views. (B) Interacting residues between CR6261 and X6, the A/South Carolina/1/1918 HA (SC18, PDB 3GBN), the A/Bayern/07/1995 HA (BA95, PDB 8DIU), and the A/Ohio/09/2015 HA (OH15, PDB 6UYN). Conserved residues between X6 and wild-type HAs involved in polar interactions with CR6261 are underlined. See also Figure S4 and Table S5.

From a mouse immunized with X6, we also isolated a murine mAb, B3, that bound the cH6/1 HA and all wild-type H1 HAs. This mAb further lacked HAI activity against pre-2009 or post-2009 H1 viruses (Figure S5). These data implied that B3 bound the H1 stem domain, further corroborating the antigenic integrity of the stem in a mouse model. Overall, these data suggest that although X6, and COBRA HAs in general, were previously thought to elicit broadened immunity through head-binding antibodies, they may still elicit broadly reactive stem-binding antibodies.

Most major antibody epitopes are intact on wild-type and COBRA HAs

The X6 COBRA HA captures both pre-2009 and post-2009 HAs in its sequence. Similar to other previously characterized HAs such as NC99, MI15, and COBRA HAs P1 and Y2, these HAs retain both variable and conserved regions on the head and stem domains. To probe whether conserved epitopes are intact on the head and stem of such HAs, we employed both ELISA and BLI to assess the EC₅₀s and general binding kinetics of RBS-, lateral patch-, central stem- and anchor epitope-binding mAbs (Figure 7A, B). We found that the stem-binding mAbs tested, P1-05 and CR6261, bound to all HAs. In contrast, head domain epitopes were more variable in their conservation across wild-type and COBRA HAs. For instance, Ab6649 bound neither to the Y2 COBRA HA nor to the MI15 HA (Figure 7A,B). This is consistent with previous studies showing that Ab6649 could not bind nor neutralize the MI15 HA and the A/MI/15 virus, respectively³⁴, likely due to selective pressure against this epitope in more recent H1N1 strains. Furthermore, the MI15 HA possesses a K166Q mutation that is also conserved on the Y2 COBRA HA that significantly reduces Ab6649 binding^33,34. In addition, by ELISA, 5J8 binding to the NC99 HA was reduced relative to other HAs as has also been described previously³². A previously determined structure of another H1 COBRA HA spanning across pre-2009 and post-2009 viruses in its design, P1, showed conserved structural features particularly in the head domain near the RBS^11,40. To compare structural features between P1 and X6, we first performed a structural alignment between these two COBRA HAs (Figure 7C). We found an RMSD of 0.942 Å between the two HAs, confirming that these two COBRA antigens are highly similar. We further compared the glycosylation profiles between these two COBRAs, finding nearly identical glycans between these HAs. One notable difference, however, was the absence of the N127 head domain glycan in the X6 COBRA that was previously found in P1 to abrogate binding of a head-binding antibody. To determine the similarity of head and stem domain epitopes between these COBRAs, we further compared the antibody epitopes between these HAs (Figure 7D). We first analyzed the antigenic sites Sa, Sb, Ca1, Ca2, and Cb. Antigenic site Sa was identical between X6 and P1, with the exception of a N159K mutation (using X6 numbering) relative to X6, showing the acquisition of a positive charge in this site in P1 that is more similar to 2009 pandemic-like HAs. In Sb, several residue differences were observed between the two COBRAs, such as a R188Q mutation relative to X6, that suggested that P1 is more similar to post-pandemic HAs at this epitope. In Ca1, X6 possesses a glutamate at position 169, whereas this is mutated to a neutral glycine in P1 but most other residues are conserved. Similarly, Ca2 is minimally mutated between these HAs, where N138 in X6 corresponds to an alanine in P1, and D221 in X6 to a glycine in P1. Cb was more plastic between these HAs, with a notable charge inversion at position 74, from a glutamate in X6 to an arginine in P1. We confirmed the conservation of recently described broadly reactive epitopes between COBRA HAs as well. The lateral patch was identical, as was the anchor epitope. The central stem was nearly completely identical with the exception of a minor V344I mutation. Overall, these data also corroborated the similar binding profiles between P1 and X6 to the tested mAbs (Figure 7A,B). In conclusion, broadly reactive antibodies directed against the HA stem were more tolerant of antigenic changes than those directed against the head domain.

Figure 7. ELISA and BLI of broadly reactive head- and stem-binding mAbs to H1 COBRA and wild-type HAs.

COBRA Y2 represents human-tropic H1 sequences from 2014 to 2016, COBRA P1 represents human-tropic H1 HAs from 1933 to 1957 and from 2009 to 2011 and swine sequences from 1931 to 1998, and COBRA X6 represents human-tropic sequences from 1999 to 2012. (A) ELISA binding curves of head-binding mAbs 5J8 and Ab6649 against the RBS and the lateral patch, respectively, and stem-binding mAbs CR6261 and P1-05, targeting the central stem and anchor epitopes, respectively, for HAs Y2, P1, X6, NC99, and MI15. Ply is a negative control. The data shown are the mean±SD of an experiment performed in quadruplicate. Four-parameter nonlinear fits were performed with Prism 9. (B) BLI association and dissociation data of the indicated mAbs to the denoted HAs loaded onto HIS1K biosensors. The association step is shown up to the dotted line and the dissociation step after this line. (C) Comparison of glycans between X6 (protein labeled blue and glycans peach) and P1 (protein labeled cyan and glycans orange). (D) Residue comparisons between X6 and P1 COBRAs in head and stem epitopes. Numbering is based on X6. The P1 structure used for glycan and epitope comparisons in (C) and (D) has PDB ID 7UYI.

Discussion

Here we show that the X6 COBRA HA possesses conserved antibody epitopes of HAs from both pre-2009 and post-2009 strains. This is similar to previously characterized COBRA HAs of the H3 and H5 subtypes^28,41. This observation implies that it could stimulate broadly reactive B cell clones in human populations. From sequence alignments to wild-type HAs, X6 possesses greater identity to pre-2009 HA strains than post-2009 HAs both overall and at the variable antigenic sites. Structurally, these sequence differences were expectedly found to be in the more variable head domain, rather than in the conserved stem domain (Figure 1C). The observation that it can bind several antibodies reactive to more recent virus strains, however, suggests that other features, such as glycosylation sites, could additionally be responsible for this enhanced breadth.

X6 is structurally similar to wild-type HAs NC99 and MI15, forming trimers with glycosylation sites that recapitulate both seasonal and pandemic-like HAs. As the majority of these sites are shared across both NC99 and MI15, one potential mechanism of eliciting broadly reactive antibodies could be the redirection of antibody responses to conserved epitopes found across both pre-2009 and post-2009 HAs and away from strain-specific sites. One glycosylation site found on the top of the X6 head domain, at N159, was shared only with NC99 but not with MI15. It is possible that this glycan could block antibodies that bind nearby antigenic sites, such as Sa, in which this glycan is located. This has been seen for other COBRA HAs, such as P1, where removal of a glycosylation site enhanced antibody binding⁴⁰.

COBRA HAs are thought primarily to elicit head-binding antibodies⁷. To investigate whether X6 possessed broadly reactive head epitopes, we determined the cryo-EM fits and structures of three mAbs, CA09-26, #58, and BE1, with X6. We found that mAbs CA09-26 and #58 bound the RBS, consistent with their enhanced HAI breadth against pre-2009 and post-2009 H1N1 strains. In addition, the epitope of mAb #58 was conserved for three of four residues both in NC99 and MI15. mAb #58 bound close to the apex of the head domain of X6 and used both heavy and light chain residues to contact the RBS. In addition, we found that mAb BE1 binds to a distinct, non-RBS epitope on the head domain, instead binding to residues that are part of the previously described lateral patch at a near-horizontal angle^33,34. Although BE1 only binds pre-2009 HAs, the lateral patch epitope with which it overlaps is conserved across both pre- and post-2009 pandemic HAs³³. Structural comparisons to previously characterized lateral patch antibodies revealed that BE1 binds at an intermediate angle between that of Ab6649 and 045-09 2B05. We found that immunization with X6 elicited pre-2009 HA-binding mAbs with minimal HAI activity in mice, possessing similar binding and HAI profiles to mAb BE1 (Figure S5). This may indicate that the lateral patch may be another significant epitope in X6 although it is structurally skewed towards pre-2009 HA sequences and structures at this epitope. Therefore, the antibody epitopes found on X6 can likely accommodate a wide range of broadly reactive RBS- and lateral patch-targeting antibodies.

We also structurally characterized the stem of X6 through determining the structure of the Fab of CR6261 with this COBRA HA. CR6261 is a group 1-reactive antibody that possesses neutralizing activity against several subtypes, including H1 and H5³⁵. It has also been used previously to guide vaccine design of stem-based HAs⁴². Here we show that X6 possesses an intact central stem epitope that can bind CR6261 using conserved residues found across wild-type HAs. These CR6261:HA interactions span residues across the historic 1918 Spanish influenza pandemic, the seasonal pre-2009 A/Bayern/07/1995, and the recent swine-origin A/Ohio/09/2015 HAs. While it has been established that immunization with COBRA HAs elicits high amounts of HAI-active, head-focused neutralizing antibodies, the stem-based antibody response has been less characterized. We also showed here that X6 immunization also induces stem antibodies in mice, as observed for mAb B3, which bound all tested H1 HAs and the cH6/1 HA construct while lacking HAI activity against H1 viruses (Figure S5). Structural confirmation of the integrity of this epitope implies that COBRA HAs may indeed elicit stem-reactive antibodies. This was shown previously for another H1 subtype COBRA, Y2, where immunization in mice elicited serum antibodies that competed with CR6261²⁶. In addition, the Y2 COBRA also possesses the recently described anchor epitope, and the H3 COBRA TJ5 binds wild-type H3 HA cross-reactive stem antibodies^25,28. We further confirmed here that previously characterized broadly reactive antibodies against the head and stem still retain reactivity against both historic and modern H1 subtype COBRA HA constructs (Figure 7).

These data provide structural insights into a next-generation influenza vaccine, providing evidence that COBRA HAs are similar to wild-type HAs from the standpoints of glycosylation and antigenic integrity. Previously we have shown that COBRA HAs cross-react with functional antibodies isolated from seasonally vaccinated human populations^25,28. From the studies described here, we define the structural basis for this expanded antibody breadth which likely depends on glycan-dependent redirection of antibody responses and broadly reactive head and stem epitopes. COBRA HAs could mediate expanded antibody breadth through (1) eliciting a wide range of antibodies that synergistically expand virus breadth, or by (2) focusing antibody responses to narrow but highly conserved epitopes. Our data presented here, in addition to other structural studies on COBRA HAs^25,28,40, lend credence to this second mechanism. In addition, the broad binding of murine mAbs to wild-type H1 HAs further suggests that, at least in part, monoclonal breadth contributes to the wide antibody-dependent breadth from COBRA HA vaccination¹⁴ (Figure S5). As the COBRA HA methodology moves into late preclinical and early clinical studies, further structural characterization will be needed to probe correlates of an effective and long-lasting antibody response.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jarrod J. Mousa (jarrod.mousa@med.fsu.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

X-ray data for the X6 crystal structure have been deposited at the Protein Databank (PDB) with PDB ID 8SJ9. Cryo-EM data for the #58 Fab:X6, BE1 Fab:X6, and CR6261 Fab:X6 cryo-EM structures have been deposited to the PDB and the Electron Microscopy Database (EMDB). The corresponding PDB and EMDB IDs, respectively, are 8V7O and EMD-43008 for the #58 Fab:X6 structure, 8GHK and EMD-40046 for the BE1 Fab:X6 structure, and 8F38 and EMD-28833 for the CR6261 Fab:X6 structure. These are also listed in the Key Resources Table. ELISA, HAI, and neutralization data reported in this paper will be shared by the lead contact upon request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Goat anti-mouse IgG, human ads-HRP	Southern Biotech	Cat#1030-05
Goat anti-human IgG-HRP	Southern Biotech	Cat#2040-05
Goat anti-mouse IgG Fc-AP	Southern Biotech	Cat#1033-04
Goat anti-human IgG Fc-AP	Southern Biotech	Cat#2048-04
Goat anti-mouse peroxidase-labeled IgG	SeraCare	Cat#5220-0341
Chemicals, peptides, and recombinant proteins
PNPP powder, 25 g	Thermo Fisher Scientific	Cat#34045
PBS	Thermo Fisher Scientific	Cat#J61196.AP
Receptor-destroying enzyme (RDE)	Denka Seiken	Cat#DKA370013
Trypsin, TPCK treated	Thermo Fisher Scientific	Cat#20233
Avicel	Sigma-Aldrich	Cat#11365-1KG
TrueBlue substrate	SeraCare	Cat#5510-0030
ABTS substrate	VWR International	Cat#30931-67-0
Zeocin	Invivogen	Cat#ant-zn-1
Critical commercial assays
flashBAC™ Baculovirus Expression Systems Kit	Mirus Bio	Cat#MIR 6115
TransIT®-Insect Reagent	Mirus Bio	Cat#MIR 6104
Pierce Fab Preparation Kit	Thermo Fisher Scientific	Cat#44985
Sodium bicarbonate	Sigma-Aldrich	Cat#792519
Aluminum sulfate	Sigma-Aldrich	Cat#202614
Deposited data
X6 COBRA crystal structure	This paper	PDB: 8SJ9
#58 Fab:X6 COBRA cryo-EM structure	This paper	PDB: 8V7O
#58 Fab:X6 COBRA cryo-EM map	This paper	EMDB: EMD-43008
BE1 Fab:X6 COBRA cryo-EM structure	This paper	PDB: 8GHK
BE1 Fab:X6 COBRA cryo-EM map	This paper	EMDB: EMD-40046
CR6261 Fab:X6 COBRA cryo-EM structure	This paper	PDB: 8F38
CR6261 Fab:X6 COBRA cryo-EM map	This paper	EMDB: EMD-28833
Experimental models: Cell lines and viruses
Insect: Sf9 cells	Expression Systems	Cat#94-001S
Human: EXPI293F cells	Thermo Fisher Scientific	Cat#A14635
COBRA X6 virus	Sautto et al.2	N/A
Software and algorithms
XDS	Kabsch4,6	https://xds.mr.mpg.de
Aimless (CCP4 Software Suite v7.1.015)	Evans et al.8,10	https://www.ccp4.ac.uk/html/aimless.html
SWISS-MODEL	Waterhouse et al.12	https://swissmodel.expasy.org
Phenix v1.19.2_4158 and v1.20.1_4487	Liebshner et al.15	https://phenix-online.org
Coot v0.8.9.2, v0.9.5, v0.9.8.1	Emsley et al.16	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
MolProbity	Williams et al.18	http://molprobity.biochem.duke.edu
Privateer (CCP4 Software Suite v8.0.004)	Aguirre et al.20	https://www.ccp4.ac.uk/html/privateer.html
CryoSPARC	Punjani et al.22	https://cryosparc.com/
Other
Non-treated 384-well plate	VWR	Cat#10814-224
Anti-penta-HIS biosensors, HIS1K	Sartorius	Cat#18-5120
Immulon 4HBX 96 well plates	Thermo Fisher Scientific	Cat#3855
Microplate, 96 well, PS, V-bottom, clear	Greiner Bio-One	Cat#651101
96 well cell culture plate	Greiner Bio-One	Cat#655167
Turkey red blood cells	Lampire Biological Laboratories	Cat#7249407
HisTrap columns	Cytiva	Cat#17524801
HiTrap MabSelect SuRe columns, 1 mL	GE Healthcare	Cat#11003493

This paper does not report original code.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Expi293F cells were cultured at 37°C and 5% CO₂ after transient or stable transfection with HA-encoding plasmids. For stable transfection, Zeocin was used at 100 μg/mL to maintain selection.

METHOD DETAILS

Vaccinations to generate X6-specific mAbs

BALB/c mice (female, 8–10 week of age), antibody negative for circulating influenza A (H1N1 and H3N2) and influenza B viruses, were purchased from Envigo (Indianapolis, IN) and housed in microisolator units and fed ad libitum. Mice were handled in accordance with protocols approved by the University of Georgia Institutional Animal Care and Use Committee and were cared under U.S. Department of Agriculture guidelines for laboratory animals. The vaccination regimen was identical to that used previously^14,27. BALB/c mice (female, 8–10 week of age) antibody negative for circulating influenza A (H1N1 and H3N2) and influenza B viruses were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in microisolator units and fed ad libitum. Mice that showed signs of severe morbidity or lost >20% of their original weight were humanely euthanized. Five mice were primed intranasally with 10⁵ PFU of COBRA X6 influenza virus, generated as previously described² and then boosted intraperitoneally 21 days later with the same virus mixed with aluminum hydroxide adjuvant (1 M NaHCO₃ from Sigma-Aldrich cat. 792519 and 0.2 M Al₂(SO₄)₃ from Sigma-Aldrich cat. 202614), as previously described²⁷. For the generation of COBRA X6 B cell hybridomas, mice received an additional intraperitoneal boost of the same virus on day 42 containing 5 μg of HA quantified as previously described² and resuspended in PBS (Corning). On day 45, spleens were harvested and immediately processed for B cell hybridoma generation.

Expression of H1 rHAs for mAb binding studies

Truncated rHA encoding HAs from H1N1 A/Chile/1/1983 (Chile/83), A/Singapore/6/1986 (Sing/86), A/New Caledonia/20/1999 (NC/99), A/Brisbane/59/2007 (Brisb/07), A/California/04/2009 (CA/09), A/Michigan/45/2015 (Mich/15), A/Brisbane/02/2018 (Brisb/18), A/Guangdong-Maonan/SWL1536/2019 (GM/19), A/Wisconsin/588/2019 (Wisc/19), cH6/1 and COBRA X3, X6, P1 and Y2 were cloned, expressed and purified as previously described⁴³ and used for all the binding experiments. In brief, the different HA proteins were expressed through a transient transfection of the EXPI293F cells (Thermo Fisher Scientific) with the different COBRA and H1N1 HA pcDNA3.1/Zeo(+) encoding vectors following the instruction provided by the manufacturer. Alternatively, HA proteins were expressed through the generation of stable transfected cells supplemented with 100 µg/mL of Zeocin (Invivogen). rHA proteins were then purified through the ÄKTA Pure system using HisTrap columns (Cytiva, 17524801) according to the manufacturer’s instructions. The H1N1 A/Texas/36/1991 rHA protein was kindly provided by F. Krammer (Ichan School of Medicine at Mount Sinai, New York, NY) while A/Puerto Rico/8/1934 (PR/34) and A/Solomon Island/3/2006 were provided by BEI Resources.

Influenza viruses/virus-like particles (VLPs)

All the 7:1 recombinant PR/34 reassortant COBRA (P1, X3 and X6) and PR/34, Chile/83, NC/99, SI/06, Brisb/07 viruses and wild-type Sing/86, TX/91, CA/09, Mich/15, Brisb/18, GM/19, A/Victoria/2570/2019 (Vic/19) viruses were propagated in embryonated chicken eggs as previously described²⁴. The A/Philadelphia/1/2013 (Phil/13) was kindly provided by Scott Hensley (Department of Microbiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA). These viruses were titrated and used for immunization, HAI, and focus reduction assay (FRA) experiments described below. The pandemic A/South Carolina/1/1918 (SC/18) HA-expressing VLP were generated as previously described¹¹ and used for HAI assays.

Enzyme-linked immunosorbent assays (ELISAs)

ELISA was used to assess mAb and serum reactivity against different H1N1 HA strains. ELISA were performed as previously described¹⁴. In brief, Immulon 4HBX plates (Thermo Fisher Scientific, 3855) were coated overnight at 4 °C with 50 µL per well of a PBS solution containing 1 µg/mL of the different rHA in a humidified chamber. The mAbs were 3-fold serially diluted in blocking buffer starting from 20 µg/mL and plates were incubated for 1 h at 37 °C. Plates were washed five times with PBS, 100 µL per well of HRP-conjugated goat anti-mouse or anti-human IgG (Southern Biotech, 1030-05 or 2040-05, respectively) diluted 1:4,000 in blocking buffer was added, and plates were incubated at 37 °C for 1 h. Finally, plates were washed five times with PBS and ABTS substrate (VWR International, Radnor, PA, 30931-67-0) was added, and plates were incubated at 37 °C for 15–20 min. Colorimetric conversion was terminated by addition of 1% SDS (50 µl per well), and OD was measured at 414 nm (OD414) using a spectrophotometer (PowerWave XS; BioTek).

ELISA was also used to assess binding of BE1 to the MI15 HA, mutants MI15 I165A, MI15 I165V, and MI15 D167N, to X6 and mutants X6 A165I, X6 A165V, X6, N167A, and X6 N167D, as well as to NC99, Y2, and Ply. Briefly, non-treated 384-well plates (VWR, 10814-224) were coated with 25 μL/well of HAs diluted to 2 μg/mL in PBS at 4°C overnight. The following day, plates were washed once with water, then 25μL of 2% block (PBS+0.05% Tween-20, 2% non-fat dry milk, and 2% heat-inactivated goat serum) was added per well. After a 1 h incubation at 25°C, plates were washed three times with water, and three-fold serially diluted BE1 from 20 μg/mL concentration in PBS, P1-05 human mAb positive control, and MPV 458 human mAb negative control, was added in 25 μL/well. Plates were incubated at room temperature for 1 h. Afterwards, plates were washed three times with water, and 25 μL/well of goat anti-mouse IgG Fc-AP (for detection of mAb BE1) or goat anti-human IgG Fc-AP (Southern Biotech, 1033-04 or 2048-04, respectively) were added in 25 µL/well. Plates were incubated for one hour at 25°C, then washed five times with PBS-T (PBS+0.05% Tween-20). 1 mg/mL PNPP (Thermo Fisher Scientific, 34045) were then added and plates were allowed to develop in the dark for one hour prior to reading on a BioTek plate reader at 405 nm.

ELISAs were also used to measure the EC₅₀s of previously characterized human mAbs 5J8 (targeting the RBS epitope), Ab6649 (targeting the lateral patch), P1-05 (targeting the anchor epitope), CR6261 (targeting the central stem epitope), and negative control mAb MPV 458. All steps were followed as in the protocol for BE1, except the antigens used were the NC99 HA, the MI15 HA, the P1 COBRA HA, the X6 COBRA HA, and the Y2 COBRA HA. All mAbs were diluted from 20 μg/mL using three-fold serial dilutions, and detected using the secondary goat anti-human IgG Fc-AP antibody (Southern Biotech, 2048-04) and with 1 mg/mL PNPP (Thermo Fisher Scientific, 34045) for one hour prior to reading on the BioTek plate reader.

Biolayer interferometry (BLI)

An Octet RED 384 instrument (Sartorius) was used to measure the binding affinities of mAbs 5J8, Ab6649, P1-05, and CR6261 to wild-type HAs NC99, MI15, P1, X6, and Y2 HAs where possible. Briefly, Anti-penta-HIS HIS1K biosensors (Sartorius, 18-5120) were equilibrated in Octet buffer (PBS+0.5% BSA+0.05% Tween) for 5 min. Biosensors were then immersed into Octet buffer for 60 s for an initial baseline step to achieve an approximate signal of 1,0 nm. Afterwards, biosensors were immersed into the respective HA at either 200 μg/mL for the Y2 HA, or at 100 μg/mL for P1, X6, NC99, and MI15 HAs for 60 s to permit antigen loading. Afterwards, biosensors were returned to Octet buffer for another 60 s. Biosensors were then immersed in three-fold serial dilutions of mAb from 1000 nM to 37 nM for 300 s for the association step. Afterwards, biosensors were returned to Octet buffer for 600 s for the dissociation step. Finally, biosensors were regenerated through cycling between 0.1 M glycine, pH=2.7 and PBS three times.

Hemagglutination inhibition (HAI) assay

The HAI assay was performed as previously described²⁷. In brief, mAbs were diluted in a series of 2-fold serial dilutions in V-bottom microtiter plates (Greiner Bio-One, 651101) starting from 20 µg/mL. An equal volume of each H1N1 virus, adjusted to 8 hemagglutination units per 50 µL, was added to each well. The plates were covered and incubated at room temperature for 20 min, and then 0.8% of turkey red blood cells (RBCs) (Lampire Biologicals, Pipersville, PA, 7249407) in PBS was added. RBCs were stored at 4°C and used within 72 h of preparation. The plates were mixed by agitation and covered, and the RBCs were settled for 30 min at room temperature. The HAI titer was determined by the reciprocal dilution of the last well that contained non-agglutinated RBCs. Positive and negative controls were included for each plate.

Focus reduction assay

The FRA was performed similarly to previously described protocols²⁷. In brief, MDCK-SIAT1 cells were seeded at a density of 2.5–3 × 10⁵ cells/mL in a 96-well plate (Greiner Bio-One) the day before the assay was run. The following day, the cell monolayers were rinsed with PBS (Thermo Fisher Scientific, J61196.AP), followed by the addition of 2-fold serially diluted mAbs at 50 µL per well starting with 20 µg/ml dilution in virus growth medium containing 1 µg/mL of L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK)–treated trypsin (Thermo Fisher Scientific, 20233). Afterwards, 50 µL of virus (CA/09) standardized to 1.2 × 10⁴ focus forming units per milliliter, and corresponding to 600 focus forming units per 50 µL, was added to each well, including control wells. Following a 2 h incubation period at 37 °C with 5% CO₂, the cells in each well were then overlaid with 100 µL of equal volumes of 1.2% Avicel (Sigma-Aldrich, 11365-1KG) in 2X MEM (Thermo Fisher Scientific) containing 1 µg/mL TPCK-treated trypsin, 0.1% BSA, and antibiotics. Plates were incubated for 18–22 h at 37 C, 5% CO₂. The overlays were then removed from each well and the monolayer was washed once with PBS to remove any residual Avicel. The plates were fixed with ice-cold 4% formalin in PBS for 30 min at 4°C, followed by a PBS wash and permeabilization using 0.5% Triton X-100 in PBS/glycine at room temperature for 20 min. Plates were washed three times with PBS supplemented with 0.1% Tween-20 (PBS-T) and incubated for 1 h with a mAb against influenza A nucleoprotein (IRR) in ELISA buffer (PBS containing 10% horse serum and 0.1% Tween-80 [Thermo Fisher Scientific]). Following washing three times with PBS-T, the cells were incubated with goat anti-mouse peroxidase-labeled IgG (SeraCare, Milford, MA, 5220-0341) in ELISA buffer for 1 h at room temperature. Plates were washed three times with PBST, and infectious foci (spots) were visualized using TrueBlue substrate (SeraCare, 5510-0030) containing 0.03% H₂O₂ incubated at room temperature for 10–15 min. The reaction was stopped by washing five times with distilled water. Plates were dried and foci were enumerated using an ImmunoSpot^® S6 Ultimate Analyzer and the CTL ImmunoSpot SC Studio software (Version 1.6.2, Shaker Heights, OH, USA). The FRA IC₅₀ titer was reported as the mAb concentration corresponding to 50% foci reduction compared with the virus control minus the cell control.

Expression of X6 for crystallographic studies

The X6 gene was synthesized and cloned into the pBacPAK8 vector (GenScript). The X6 construct consisted of a GP67 secretion signal, the COBRA X6 HA ectodomain, fused to a C-terminal thrombin cleavage site, T4 fibritin foldon, His-tag, and Strep-tag⁴⁴. Recombinant baculovirus containing the X6 gene was generated using the flashBAC^™ system (Mirus Bio, MIR 6115). Protein expression was performed in Sf9 cells cultured in ESF921 medium (Expression Systems, 94-001S) by infecting with ~23 mL of virus per liter of culture. After 3 days, the supernatant was harvested by centrifugation and stored at −20°C.

The medium containing the X6 protein was thawed at 4°C. It was then subjected to filtering through glass microfiber filter, followed by buffering with concentrated NaCl and Tris pH 8. This solution was then sequentially filtered through 0.45 µm, and 0.22 µm filters and concentrated by tangential flow using VivaFlow 200^® cassettes (Sartorius). BioLock (IBA Life Sciences) was added to bind free biotin, then the sample was filtered and loaded onto a 5 mL StrepTrap column (GE Healthcare). The column was washed with 150 mM NaCl, 50 mM Tris [pH 8], 1 mM EDTA, and the protein eluted with 150 mM NaCl, 50 mM Tris [pH 8], 1 mM EDTA, 2.5 mM desthiobiotin. The fractions were pooled, concentrated, and supplemented with glycerol (5%) prior to snap freezing in liquid nitrogen and storage at −80°C.

Crystallization and structural solution of X6

Aliquots of X6 protein were thawed on ice, supplemented with 2 mM CaCl₂, and digested with ~7 µg of trypsin per 1 mg of protein to remove the Foldon trimerization domain and cleave the furin site to generate active HA. The trypsin-activated X6 HA was further purified by size exclusion chromatography (SEC) using a Superdex 200 column (GE Healthcare) equilibrated with 50 mM NaCl, 10 mM Tris [pH 7.5]. The fractions containing HA based on an SDS-PAGE gel were pooled and concentrated to 9.8 mg/mL. The protein was crystallized in 2 µL hanging drops with a 1:1 ratio of well solution to protein in a condition consisting of 0.1 M Tris pH 8.4, 22% PEG 3350. Crystals were cryoprotected in a solution consisting of 0.1 M Tris pH 8.4, 22% PEG 3350, 6% ethylene glycol, 6% DMSO, 6% glycerol and flash-cooled in liquid nitrogen. Data collection was performed at the GM/CA beamline 23ID-D at the Advanced Photon Source (APS). Data from a single crystal was indexed and integrated using XDS^4,6, followed by scaling and merging in Aimless^8,10. The structure was phased using Phaser⁴⁵ in the PHENIX suite using a homology model generated by SWISS-MODEL¹². Three monomer copies were placed in the asymmetric unit to form a single trimer. Refinement and model building were performed in PHENIX⁴⁶ and COOT¹⁶. The final model was validated with MolProbity¹⁸ and Privateer²⁰.

Fab generation

The CA09-26 mAb, isolated from a subject who received the 2017–2018 Fluzone seasonal vaccine, was purified from hybridoma culture on a Protein GE column as described previously³⁷ (GE Healthcare). The #58 mAb was generated by single-cell sequencing of the plasmablast repertoire of a Fluzone-vaccinated individual using a previously described strategy^47,48. Heavy and light chains of mAb #58 were synthesized and cloned in the pcDNA3.4 vector (Thermo Fisher Scientific) by GenScript (Piscataway, NJ, USA) and expressed in human embryonic kidney (HEK) 293F cells and purified as previously described⁴⁹. These mAbs were then digested to Fab fragments using the Fab Preparation Kit (Thermo Fisher Scientific, 44985) according to the manufacturer’s instructions. Fab was purified from the digestion reaction on a MabSelect column (GE Healthcare, 11003493) and buffer exchanged into 20 mM Tris, pH=7.5, 100 mM NaCl for generation of the complex.

For the isolation of BE1 and other X6-specific murine mAbs, splenocytes from X6-immunized mice were used to generate B cell hybridoma cell lines, from which the mAbs were purified using previously described methods^14,27. The BE1 Fab was prepared in a similar manner as the CA09-26 and #58 Fabs but a final buffer of 1x PBS pH=7.4, 300 mM NaCl and a final concentration of 7.5 mg/mL were used.

The CR6261 antibody Fab fragment was expressed in HEK293 cells as a secreted protein with a C-terminal His₆ tag on the heavy chain and harvested 5 days post-induction. The secreted mammalian media was buffer-exchanged and concentrated by passing over a 10 kDa filter using hollow fiber tangential flow filtration with 20 mM Tris pH=7.4, 250 mM NaCl. The protein was purified by nickel affinity chromatography followed by size exclusion chromatography, then concentrated to 9.8 mg/mL in 25 mM Tris pH 7.4, 150 mM. The Fab was then flash frozen in liquid nitrogen in 100 µL aliquots and stored at −80°C until used in cryo-EM.

Cryo-EM sample preparation

For the CA09-26 Fab:X6 and #58 Fab:X6 structures, the X6 COBRA HA was purified from 293 cells as described previously¹¹. X6 COBRA HA, in 20 mM Tris, pH=7.5, 100 mM NaCl, was mixed with CA09-26 or #58 Fab in a molar excess of CA09-26 or #58 Fab. Following incubation of the components at 4°C overnight, the mixture was subjected to size exclusion chromatography on a Superdex 200 column (GE Healthcare) to isolate the Fab:X6 complex from excess Fab.

For the BE1:X6 structure, the complex was prepared in a ratio of 1:1.2 theoretical equivalents of Sf9 cell-expressed X6 HA to Fab. After incubation on ice for 30 minutes, the mixture was subjected to size exclusion chromatography to remove extra BE1 Fab. The peaks that corresponded to the X6 COBRA HA-BE1 Fab complexes were kept for cryo-EM grid preparation. We did not have the BE1 Fab protein sequence at the time and used a generic protein molar extinction coefficient for the BE1 Fab, which resulted in a sub-stoichiometric ratio and a 3:2 HA:Fab complex in the cryo-EM structure.

For the CR6261:X6 structure, Sf9 cell-expressed X6 COBRA HA and CR6261 antibody Fab were mixed in a 1:1.2 molar ratio. Again, after incubation on ice for 30 minutes, the mixture was subjected to size exclusion chromatography to remove extra CR6261 Fab. The peaks that corresponded to the X6 COBRA HA-CR6261 Fab complexes were kept for cryo-EM grid preparation.

Cryo-EM grid preparation and data collection

For CA09-26 Fab:X6 and #58 Fab:X6, Fab:X6 complex at concentrations of 1.47 mg/mL and 1.0 mg/mL, respectively, were applied to Quantifoil 1.2/1.3 (400 mesh) grids previously glow-discharged for 45 s at 25 mAmp current on the carbon side. For CA09-26 Fab:X6, grids were blotted for 10 s with 100% humidity and plunge-frozen in liquid ethane using a FEI Vitrobot Mark IV instrument. For #58 Fab:X6, grids were blotted for 8 and 10 s with 100% humidity using a FEI Vitrobot Mark IV instrument. Cryo-EM data were collected on a Glacios (Thermo Fisher) equipped with a Falcon 4 camera. Cryo-EM movies were acquired using a nominal magnification of 190,000x, with a pixel size of 0.526 Å. Movies were recorded as 30-frame videos in counting mode, with a defocus range from −0.9 to −2.0 µm.

For the BE1 Fab:X6 complex, 4 µL of HA-Fab complex directly purified from SEC were applied to a C-flat R 2/1 (300 mesh) grid after glow discharging. Grids were blotted for 2.5 s with approximately 90% humidity and plunge-frozen in liquid ethane using a FEI Vitrobot Mark IV instrument. Cryo-EM data were collected at liquid nitrogen temperature on a Titan Krios at 300 kV equipped with a K3 summit direct electron detector (Gatan) at McGill University. All cryo-EM movies were recorded using SerialEM software. Specifically, images were acquired at a nominal magnification 135,000x, corresponding to a pixel size of 0.675 Å. Movies were recorded as 40-frame videos using 2-w exposures in counting mode with a defocus range from −1.0 µm to −2.5 µm.

For CR6261 Fab:X6, 4 µL of HA-Fab complex at a concentration of 0.5 mg/ml were applied to a lacey carbon grid (300 mesh) coated by a layer of graphene oxide (SPI supplies). Grid blotting conditions were the same as the BE1 Fab:X6 complex.

Cryo-EM image processing and model refinement

For CA09-26 Fab:X6 and the #58 Fab:X6 structure, the data were processed in CryoSPARC²² for patch motion correction, patch CTF correction, particle picking, and particle extraction. For CA09-26 Fab:X6, this was followed by two rounds of 2D classification to generate class averages. For #58 Fab:X6, this was followed by multiple rounds of 2D/3D class averaging and reconstruction and refinement. The #58 Fab:X6 structure was manually built in COOT and refined in Phenix.

For the BE1 Fab:X6 cryo-EM dataset, raw movies were directly imported into CryoSPARC 3.2.0 for the analysis. In CryoSPARC, motions were corrected by batch motion correction. Contrast transfer function (CTF) parameters were estimated using patch CTF estimation. Movies with CTF Fit resolution (Å) worse than 8 Å were removed. Blob picker was applied for automatic particle picking. Particles were extracted using a box size 558 pixels and Fourier-cropped to 140 pixels in order to save computation resources. Two rounds of 2D classification were applied to clean the particles, which yielded about 794,000 particles. Ab-initio 3D reconstruction was done by asking for three ab-initio models. The class that contains the greatest number of particles contained two copies of Fabs. The particles in this class were selected and used for Topaz training and particle picking in order to pick more particles. The newly picked particles were cleaned by additional rounds of 2D classification. Then, homogeneous refinement and non-uniform refinement were applied to refine the particles. Lastly, per-particle based CTF refinement was applied to further polish the particles, which were used for local refinement by using a focused mask covering the two Fab regions and the well resolved HA region. No symmetry was applied during the refinements as there are only two copies of Fab bound. The final map was achieved to 3.47 Å.

The sharpened map was obtained using the sharpening tool in CryoSPARC. The CR6261 structure described below was divided into a HA part and a Fab part, after which each of them was then individually docked into the density using ChimeraX. The HA trimer could be placed, along with two copies of the Fab, associated with chains A and B of the HA trimer. The density for Fab bound to chain C was too weak to model. The initial model was then improved by iterative cycles of manual model building in Coot and real-space refinement in Phenix. Cryo-EM data collection, reconstruction and model statistics are summarized in Table S4.

For the CR6261 Fab:X6 structure, movie frames were corrected for their motions using MotionCor2 in Relion 3.1. The motion-corrected movies were imported into CryoSPARC 2.2.0 for downstream analysis. In CryoSPARC, CTF parameters were estimated using patch CTF estimation. Movies with CTF Fit resolution (Å) worse than 8 Å were removed. The blob picker was applied for automatic particle picking. Particles were extracted using a box size of 558 pixels and Fourier-cropped to 140 pixels in order to save computation resources. Three rounds of 2D classification were applied to clean the particles, which yielded about 500,000 particles. Ab-initio 3D reconstruction was done by asking for five ab-initio models. The initial model containing three Fab molecules was selected for further 3D refinement using C1 symmetry. The refined model confirmed the identity of the expected HA-Fab complex. These selected particles were re-retracted using their original pixel size. After per-particle CTF refinement, and non-uniform refinement using C3 symmetry, the final map was refined to 2.64 Å. The sharpened map was obtained using the sharpening tool in CryoSPARC. A sequence alignment was performed between X6 COBRA HA and influenza A virus hemagglutinin from A/Ohio/09/2015 (PDB code: 6UYN) first, then the side chains of X6 COBRA HA were truncated and aligned on the monomeric crystal structure of the influenza A virus hemagglutinin from A/Ohio/09/2015 using Chainsaw from CCP4. Three copies of the generated monomeric X6 COBRA HA were manually fit into the cryo-EM map. The structure of a single heterodimeric CR6261 Fab was extracted from PDB code 6UYN and then three copies of the molecules were manually fit into the corresponding EM densities. All the glycosylation sites were modeled in coot using a blurred map, which was generated by the sharpen/blur tool under cryo-EM module in coot. The initial model was then subjected to rounds of model building in coot and real-space refinement in Phenix. The final model was validated with statistics from MolProbity and EMRinger. Cryo-EM data collection, reconstruction, and model statistics are summarized in Table S5.

Cryo-EM data processing procedures, FSC curves, local resolution maps, and particle orientation distributions are shown in Figure S6.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical details are described in the figure legends. X-ray crystallography data collection and refinement statistics are shown in Table S1. Cryo-EM data collection and refinement statistics are summarized in Tables S2, S3, S4, and S5.

Highlights.

X6 captures glycosylation and antigenic features of pre-2009 and post-2009 H1 HAs

Vaccine-induced COBRA-reactive mAbs can have broad H1 reactivity and neutralization

X6 possesses intact head and stem epitopes characteristic of several H1 viruses