Altering the Immunogenicity of Hemagglutinin Immunogens by Hyperglycosylation and Disulfide Stabilization

Influenza virus alters glycosylation patterns on its surface exposed glycoproteins to evade host adaptive immune responses. The viral hemagglutinin (HA), in particular the H3 subtype, has increased its overall surface glycosylation since its introduction in 1968. We previously showed that modulating predicted N-linked glycosylation sites on H3 A/Hong Kong/1/1968 HA identified a conserved epitope at the HA interface. This epitope is occluded on the native HA trimer but is likely exposed during HA “breathing” on the virion surface. Antibodies directed to this site are protective via an ADCC-mediated mechanism. This glycan engineering strategy made an otherwise subdominant epitope dominant in the murine model. Here, we asked whether cysteine stabilization of the hyperglycosylated HA trimer could reverse this immunodominance by preventing access to the interface epitope and focus responses to the HA receptor binding site (RBS). While analysis of serum responses from immunized mice did not show a redirection to the RBS, cysteine stabilization did result in an overall reduction in immunogenicity of the interface epitope. Thus, glycan engineering and cysteine stabilization are two strategies that can be used together to alter immunodominance patterns to HA. These results add to rational immunogen design approaches used to manipulate immune responses for the development of next-generation influenza vaccines.


INTRODUCTION
Viruses use antigenic evolution to introduce glycans on their surface glycoproteins to evade immune surveillance by the host adaptive response; HIV and its envelope (Env) glycoprotein is one notable example (1). Indeed, for HIV Env, glycans account for approximately half the mass of the protein, and glycan density and heterogeneity can affect induction of broadly neutralizing antibodies (bnAbs) (1)(2)(3). Influenza (4,5), ebola (6,7), hepatitis C (8), Lassa (9)(10)(11), and coronavirus (12) are other examples of viruses that vary their glycosylation to "shield" conserved epitopes. Influenza has varied its hemagglutinin (HA) glycosylation profiles across subtypes since its introduction into the human population. Circulating H1 HAs have maintained approximately 8 glycans since their re-introduction in 1977 until the introduction of the pandemic H1N1 in 2009 which lost 4 glycans relative to previous historical H1s. H3 HAs, however, have increased overall glycans since the 1968 pandemic from 7 to 13.
Conserved epitopes on influenza HA are targets for broadly protective humoral immunity. These epitopes include the receptor binding site (RBS), stem, and "interface" regions (13). Antibodies that engage the RBS neutralize by impeding viral entry through sterically blocking attachment to its receptor, sialic acid. However, cross-reactive RBS-directed antibodies are subdominant, and their breadth is often dependent on having conserved features in their complementary determining region 3 (HCDR3) to present sialic acid-like contacts (14)(15)(16)(17)(18)(19)(20). The stem region of HA is similarly conserved, likely due to functional constraints necessary for undergoing conformational changes required for membrane fusion (21). Antibodies targeting the stem show breadth across nearly all HA subtypes (22)(23)(24)(25). However, these antibodies targeting the stem are encoded by a relatively restricted set of variable genes and are difficult to elicit (21). More recently, we and others identified an additional conserved epitope at the HA trimer interface (26)(27)(28)(29)(30). This epitope lies at an otherwise occluded site between HA protomers in the native trimeric structure present on the virion surface. Nonetheless, antibodies targeting this epitope have been isolated from both human and murine sources and confer protection via an Fc-dependent mechanism (26)(27)(28). This implies that this occluded epitope is transiently exposed enough on the virion surface or the infected cell to be targeted by such antibodies (31).
Current immunogen design efforts for next generation influenza vaccines are focused on eliciting protective antibodies targeting these conserved epitopes (32). Here, we analyzed whether glycan engineering along with cysteine stabilizing mutations could modulate elicited humoral immune responses. We used the H3 A/Hong Kong/1/1968 HA as a model antigen to develop glycan shielded HA immunogens that block variable surface-exposed epitopes; we selectively exposed or concealed the RBS epitope with glycans and further stabilized the trimeric HA by introducing cysteine residues at both the head and stem interfaces. We immunized mice with stabilized or non-stabilized, glycan-modified immunogens and characterized elicited immune responses. We observed a reduction in serum responses to the stabilized immunogens, indicating that blocking surface epitopes with glycans or disulfide stabilization can significantly alter their overall immunogenicity. These data show how different protein engineering strategies can be combined to influence humoral responses to conserved epitopes on the influenza HA.

RESULTS
Based on our previous work, we designed second-generation glycan-modified hemagglutinin (gHA) immunogens based on the historical H3A/Hong Kong/1/1968 HA (HK-68). Our firstgeneration gHAs included three immunogens with varying degrees of glycosylation: (1) an "RBS-exposed" immunogen (gHA RBS ), in which the entire surface is hyperglycosylated but leaves the RBS exposed, (2) an "RBS-concealed" (gHA cRBS ) immunogen with a single glycan at position 133 to abrogate binding of RBS-directed antibodies, and (3) a "full-shield" immunogen (gHA shield ) which obscures all surface-exposed epitopes. For this study, we modified our initial gHA RBS immunogen to include two additional predicted N-linked glycans (PNGs) at residues 122 and 126 to create gHA RBSv2.0 ; this construct differs from gHA shield by two PNGs at residues 133 and 158, both in the RBS pocket ( Figure 1). We used these two immunogens, gHA RBSv2.0 and gHA shield , to discern RBS-from non-RBS directed responses and as templates for further modifications.
Antibodies elicited by our previous immunization with firstgeneration gHAs focused predominantly to the trimer interface; therefore, we hypothesized that occluding the interface epitope, could re-focus the immune response to the RBS. This may have been a consequence of the engineered glycans destabilizing the HA trimer resulting in mixed trimer and monomeric HAs; the latter readily exposing the interface epitope (Supplemental Figure 1). To that end, we stabilized the HA trimer by engineering cysteines at residues 212 and 216 (33) (H3 numbering) on the HA head and at residues 30 and 376 (34) on the HA stem, in order to occlude immune responses directed to the HA interface ( Figure 1A). These stabilizing mutations were engineered onto both gHA RBSv2.0 and gHA shield , to yield their cysteine-stabilized counterparts, referred to as gHA cys RBSv2.0 and gHA cys shield ( Figures 1B-D). We confirmed disulfide presence using SDS-PAGE under reducing and nonreducing conditions ( Figure 1E). The cysteine modified hyperglycosylated HA proteins showed predominantly trimeric bands that collapsed into a monomeric species upon reduction. They also are stable trimers under native conditions, as confirmed by size-exclusion chromatography (Supplemental Figure 1).
We then assessed the immunogenicity and antigenicity of gHA RBSv2.0 , gHA cys RBSv2.0 , gHA shield , and gHA cys shield . We immunized mice and collected serum at day 0, 7, 14, 21, and 28. Serum IgG responses were assessed via multiplex bead array (Luminex platform) for binding to each of the four immunogens ( Figure 3 and Supplemental Figure 3). While the gHA RBSv2.0 and gHA shield immunogens elicited a robust self-directed serum response, there was a comparatively weak reactivity to the cysteine modified counterparts indicating a predominantly interface-directed response ( Figures 3A, B). This pattern persisted throughout the course of the experiment (Supplemental Figure 3). These data are consistent with our previous observations of the role of glycan shielding and immunodominance of the interface epitope (27). While immunizations with cysteine gHA cys RBSv2.0 and gHA cys shield elicited a nearly undetectable response at day 7, by day 28 we observe a similar overall magnitude of response comparable to the non-cysteine stabilized immunogens (Figures 3C, D and Supplemental Figure 3). However, unlike gHA RBSv2.0 and gHA Shield , this response was comparable across all four immunogens for both gHA cys RBSv2.0 and gHA cys shield . This potentially indicates that serum responses elicited by disulfide-stabilized gHAs were directed towards a common epitope exposed across all four immunogens. Notably, we did not observe differential reactivity to our "RBS-exposed" and "shield" immunogens at any timepoint, indicating that there was likely not a robust RBS-directed humoral response elicited by our exposed RBS immunogens. Overall, these data indicate that the combination of hyperglycosylation and disulfide stabilization alters the immunogenicity kinetics of gHA immunogens but cannot significantly redirect humoral responses to the exposed RBS epitope.
Next, we performed competition ELISAs with a panel of conformational specific antibodies with d28 serum to epitope map the elicited responses to our immunogens ( Figure 4). Our antibody panel included the FI6 (stem), HC45 (vestigial esterase), HC19 (RBS), and 8H10 (interface) epitopes ( Figure 2). The H1 RBS-directed 5J8 (38), which does not bind to H3 HK-68, was included as a negative control. Serum from gHA RBSv2.0 , gHA cys RBSv2.0 , and gHA cys shield showed significant competition with HC45 ( Figures 4A-C), that was not present in gHA cys shield ( Figure 4D). While not as significant as the HC45-like response, both gHA cys RBSv2.0 , and gHA cys shield immunogens apparently elicited a HC19-like and 8H10-like response, despite the presence of glycans and inter-protomer disulfides. There was no FI6-like response detected in sera from any of the immunogens; this potentially indicates that despite fewer glycans present to obscure responses targeting the stem, this region remains subdominant. Finally, we analyzed day 28 serum for breadth and influenza virus neutralization activity. We assessed H3 breadth using a panel of antigenically distinct, historical H3s spanning 1968-2012 ( Figure 5 and Supplemental  Additionally, serum elicited by all four immunogens had weak neutralization activity, with no appreciable difference between "RBS-exposed" and "shield" immunogens ( Figure 5B). This is consistent with the above data that these immunogens do not elicit robust RBS-directed serum responses that would contribute to neutralization. Furthermore, the lack of neutralization by serum from gHA RBSv2.0 and gHA shield immunogens is likely a consequence of a focused response to HC45-like or interface epitopes that do not contribute to neutralization.

DISCUSSION
In this work, we determined whether modifying HA using hyperglycosylation and disulfide stabilization could focus antibody responses to the subdominant RBS epitope. We hypothesized that by both shielding surface-exposed epitopes and by restricting access to the immunodominant interface epitope by glycans, theoretically the only accessible epitope is the RBS, and thus should become immunodominant, or at minimum, be able to elicit a detectable   (43), and HIV (44). To our knowledge, however, this is the first example to study the effect of inter-protomer disulfide stabilization offull-length influenza A HA in combination with hyperglycosylation on immunogenicity. However, in this study we showed that obscuring all surfaceexposed epitopes except the RBS did not robustly direct responses to the RBS. The observed lack of an RBS-directed serum response could be due to a variety of factors. Human-derived bnAbs that engage the RBS of group 1 and group 2 HAs have been isolated and appear to be relatively abundant (14-17, 19, 20). These RBS-directed antibodies require long HCDR3s (~19 amino acid in length) to engage the recessed RBS. Because the C57BL/6 mice used in this study tend to have shorter HCDR3s, it is likely that this restricts humoral responses targeting the RBS epitope. Additionally, these antibodies could be rare in the mouse antibody repertoire, and thus may not be readily detected in serum. Alternative murine models with human variable heavy genes knocked-in [e.g., JH6 (45)] may be better suited for characterizing immunogens designed to elicit RBS-directed responses.
Despite being nearly "immune silent" at early time points, the disulfide-stabilized immunogens eventually elicited a weak, but detectable, antibody response at days 21 and 28. This may indicate that the engineered glycans do not completely shield surfaceexposed epitopes. In other words, it is possible that antibodies target "holes" in the engineered glycan shield that may ultimately contribute to the humoral response. From competition ELISA data, this could be an epitope that partially overlaps with HC45 within the vestigial interface region but is not obscured by glycans on gHA constructs, or another epitope(s) not covered by our antibody panel. A similar phenomenon has been observed in HIV-1, where antibodies targeting the CD4 binding site on gp120 such as VRC01-like antibody, VRC08 (46), as well as N6 (47), could evolve around a glycan at position N276, present on~95% of circulating HIV-1 strains. Future immunogen design iterations may vary the pattern of engineered glycans to optimize placement and shielding.
Overall, these data show how hyperglycosylation and disulfide stabilization can effectively modulate humoral responses. There appears to be a "limit" on available surface-exposed epitopes on our immunogens that can be effectively shielded by glycans to dampen overall humoral response. This parallels how viruses use glycosylation of their envelope proteins to evade host adaptive immune responses [e.g., HIV (1-3), ebola (6), hepatitis C (8),

Fab and IgG Expression and Purification
Variable domain genes for IgG production were codon optimized and synthesized by Integrated DNA Technologies (IDT) and subcloned into pVRC vectors containing humanized heavy and light constant domains for expression in mammalian cells (HEK/Expi 293F cells). DNA constructs were sequenced confirmed using Genewiz. Humanized HC19 (35), HC45 (36), FI6 (22) and 8H10 (27) IgGs were expressed transiently in suspension 293F cells using polyethyleneamine (PEI) or ExpiFectamine 293 for 5 days. After 5 days, IgGs were purified from clarified supernatants on a Protein G resin, and eluted into Tris-Glycine buffer with 150mM NaCl.

Size Exclusion Chromatography
To assess the oligomeric state(s) of wild-type H3 X31 HA and gHAs,~200µg of purified protein with foldon trimerization tag was passed over a Superdex 200 10/300 increase column (GE Healthcare) in Dulbecco's Phosphate-Buffered Saline (D-PBS) (pH 7.5). The foldon tag was then removed using His-tagged HRV 3C protease overnight at 4°C overnight. The cleavage reaction was then re-purified over a cobalt-TALON resin to remove non-cleaved HA, cleaved foldon tag, and protease. The cleaved HA was passed over the same Superdex 200 10/300 increase column. All traces were normalized to their respective peak absorbance value and plotted using Prism 9 (Graphpad).
ELISA 100 ng of hyperglycosylated or wild-type recombinant hemagglutinin proteins were adhered to high-capacity binding, clear bottom, 96-well plates (Corning) in PBS overnight at 4°C. Plates were blocked with 1% BSA in PBS containing 0.1% Tween-20 (PBS-T) for 1 hour at room temperature (RT), shaking. Blocking solution was removed, and plates were incubated with 10-fold dilutions of structurally characterized IgGs in PBS for 1 hour at RT, shaking. After 1 hour, plates were washed three times with PBS-T, then incubated with 1:20,000 diluted HRP-conjugated anti-human IgG (Abcam) for 1 hour at RT. Plates were washed three times with PBS-T. 1-Step ABTS substrate (ThermoFisher) was added to plates for 20-30 mins at RT, then stopped with 100 µL of 1% SDS. Absorbance measurements from each well at 410 nm were recorded using a plate reader. Data were plotted using Prism 9 (Graphpad) and relative binding was determined.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Duke University Institutional Animal Care and Use Committee.

AUTHOR CONTRIBUTIONS
DT, AS, GS, and AK designed research. DT, AM, ES performed research. DT and AS analyzed data. DT and AS wrote the paper. DT, AS, GS, AM, and AK edited and commented on the paper. All authors contributed to the article and approved the submitted version.

FUNDING
We acknowledge support from R01 AI146779 and P01 AI089618 (to AS). This research has been funded in whole or part with federal funds under a contract from the National Institute of Allergy and Infectious Diseases, NIH contract 75N93019C00050. Multiplex binding and microneutralization assays were performed in the Immunology and Virology Units, respectively, of the Duke Regional Biocontainment Laboratory, which received partial support for construction from the NIH/NIAID (UC6 AI058607; to GS).

ACKNOWLEDGMENTS
We thank the Schmidt Lab for helpful discussions.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2021.737973/ full#supplementary-material Supplementary Figure 1 | Hyperglycosylation increases tendency to present "exposed" HA trimer interface epitope-Wild-type (WT) H3 HK-68 and gHAs were analyzed via size-exclusion FPLC for conformational changes pre-and post-Foldon (Fd) trimerization tag cleavage. WT (A) and cysteine-stabilized gHA (B) showed a small shift in elution volume (<0.5mL) due to the small change in molecular weight following Fd tag cleavage. gHA RBS (C) and gHA shield (D) showed an additional elution volume shift of~1.5mL, indicating the presence of dimer or monomeric HA and an "exposed" interface epitope.