Shared IGHV1-69-encoded neutralizing antibodies contribute to the emergence of L452R substitution in SARS-CoV-2 variants

ABSTRACT SARS-CoV-2 variants continue to emerge facing established herd immunity. L452R, previously featured in the Delta variant, quickly emerged in Omicron subvariants, including BA.4/BA.5, implying a continued selection pressure on this residue. The underlying links between spike mutations and their selective pressures remain incompletely understood. Here, by analyzing 221 structurally characterized antibodies, we found that IGHV1-69-encoded antibodies preferentially contact L452 using germline-encoded hydrophobic residues at the tip of HCDR2 loop. Whereas somatic hypermutations or VDJ rearrangements are required to acquire L452-contacting hydrophobic residues for non-IGHV1-69 encoded antibodies. Antibody repertoire analysis revealed that IGHV1-69 L452-contacting antibody lineages are commonly induced among COVID-19 convalescents but non-IGHV1-69 encoded antibodies exhibit limited prevalence. In addition, we experimentally demonstrated that L452R renders most published IGHV1-69 antibodies ineffective. Furthermore, we found that IGHV1-69 L452-contacting antibodies are enriched in convalescents experienced Omicron BA.1 (without L452R) breakthrough infections but rarely found in Delta (with L452R) breakthrough infections. Taken together, these findings support that IGHV1-69 population antibodies contribute to selection pressure for L452 substitution. This study thus provides a better understanding of SARS-CoV-2 variant genesis and immune evasion.

Before global dominance of Omicron BA.1 variant, Delta variant featuring a L452R substitution was the dominant strain worldwide. L452R was first observed in B.1.427/B.1.429 variants which were dominant in California between November and December 2020 [4]. Studies on Delta variant found that L452R mutation reduced neutralizing activities of 14 out of 34 isolated RBD-specific monoclonal antibodies (mAbs). It has milder effect in reducing neutralizing activity of convalescent sera [2,5,6]. Most recently, L452Q, L452M, and L452R have been identified in Omicron subvariants BA.2.12.1, BA.2.13, and BA.4/ BA.5, respectively [7], indicating active substitution events at this site. It is currently understood that although mutations are result of error-prone viral replication, dominant variants were selected for transmission advantages. It is conceivable that population immune pressure established from prior infections or vaccinations plays crucial roles in driving virus evolution [8], and variants with ability to evade population antibody responses would possess a considerable transmission advantage.
To help understanding virus evolution, it is therefore important to analyze the prevalence of antibody lineages that are shared among the human population. Variants of SARS-CoV-2 bear mutations that allow escape from neutralization by different classes of antibodies, especially those belonging to classes widely distributed in the human population [9]. Previously, we and others have reported that IGHV3-53-encoded neutralizing antibodies are commonly elicited by SARS-CoV-2 infection and they share similar binding mode to RBD [10][11][12]. It has been identified that recurrent mutation K417N abolishes neutralization by this class of shared antibodies [12]. Similar to K417N, L452R has been also recurrent in various VoCs. However, it is unclear whether the appearance of L452 substitution is related to the immune pressure of share neutralizing antibodies induced by SARS-CoV-2. In this report, we identified that IGHV1-69 gene encodes antibodies of 3 major clonotypes and these antibodies are widely induced in vaccinees or convalescents of the ancestral SARS-CoV-2 strain (wildtype, WT). We identified that IGHV1-69 antibodies preferentially recognize L452 using preexisting hydrophobic residues at the tip of heavy chain complementary determining region 2 (HCDR2) of germline IGHV1-69 gene. By antigen binding and virus neutralization experiments, we confirmed that L452R can specifically evade binding and neutralization by many IGHV1-69 antibodies. We provide evidence that IGHV1-69-encoded L452contacting mAbs are differentially recalled in Delta and Omicron BA.1 breakthrough infections, further revealing L452R's ability to impair immune memory recall in addition to evasion of physical antibody binding. These findings point to an immune selection pressure for L452R mutation contributed by IGHV1-69-encoded neutralizing antibodies. The findings reported here facilitate better understanding of SARS-CoV-2 variant genesis and will help rationalize consequences of immune evasion by generated variants.
Analysis of SARS-CoV-2 spike variants and published SARS-CoV-2 RBD antibody sequences Per-site variation of SARS-CoV-2 spike was calculated by sequence entropy using the COVID-19 Viral Genome Analysis Pipeline (https://cov.lanl.gov/ components/sequence/COV/int_sites_tbls.comp?t=2) that enabled by data from the Global Initiative for Sharing Avian Influenza Data (GISAID; https:// gisaid.org). A total of 2438500 SARS-CoV-2 sequences collected before 16 June 2022 were included in this analysis. The germline usage distribution of RBD-targeting mAbs was calculated using the data from COV-AbDab database (the coronavirus antibody database) (http://opig.stats.ox.ac.uk/webapps/covabdab/). A total of 4002 RBD-targeting mAbs isolated from ancestral SARS-CoV-2 infected individuals or vaccinees (accessed before June 22, 2022) were included in this analysis. The frequency distribution of IGHV genes in RBD-targeting mAbs were compared to the IGHV distributions in healthy human antibody repertoires (PRJCA007067 and PRJCA00375) [10,13,14].

Structural analysis
Human SARS-CoV-2 RBD-specific mAbs with available structures were downloaded from the PDB (https://www.rcsb.org/). A total of 221 antibody-RBD/Spike complexes were curated. Buried surface area (BSA) was calculated using the PDBe PISA server (https://www.ebi.ac.uk/msd-srv/prot_int/). Antibody with a non-zero BSA on L452 were identified as a L452-contacting antibody. Structure figures were generated by PyMOL Molecular Graphics System (Schrödinger).

Expression of monoclonal antibody
The antibody heavy-and light-V genes (VH/VL) were synthesized (Genscript, China) and were cloned into human IgG1 expression vectors using Clone Express II One Step Cloning Kit (Vazyme, China). When density of HEK293F cells reached 1 × 10 6 cells/mL, equal amounts of heavy-and light-chain plasmids were transfected into HEK293F cells using EZ cell transfection reagent (Life-iLab Biotech, China). Following transfection, HEK293F cells were cultured in CD 293 TGE medium (ACRO, China) containing 10% CD Feed X supplement (ACRO, China) at 37°C in a humidified 5% CO 2 incubator shaking at 120 rpm. 6 days post transfection, supernatants were harvested and clarified by centrifugation. Supernatants were filtered through 0.22-µm filters (Merck Millipore, Germany) before incubated with Protein A Resin (Genscript, China) at room temperature for 2 h for antibody affinity purification. After washing, antibodies were eluted from the Protein A Resin using 0.1 M Na-Citrate (pH 3.25) and eluents were neutralized immediately with 1 M Tris-HCl (pH 8.8). Antibodies were concentrated in PBS using Amicon Ultrafilter (Merck Millipore, USA) (GIBCO, USA) and stored at −80°C.

Biolayer interferometry
Binding between SARS-CoV-2 RBDs and antibodies was measured by biolayer interferometry (BLI) using an Octet RED96e system (FortéBio, USA). All experiments were performed at 25°C with shaking at 1,000 r.p.m, and anti-human IgG biosensors were pre-equilibrated in Q buffer containing PBS (10 mM pH7.4), 0.02% Tween-20, and 0.2% BSA for at least 600 s before use. Antibodies were diluted to 11 μg/ mL before loaded onto protein A biosensors for 60 s. Antibody loaded biosensors were dipped into wells containing 200 nM SARS-CoV-2 RBD-His recombinant proteins or Q buffer (as control) for 300 s to monitor association. Dissociation was monitored by moving biosensors into Q buffer for 300 s. Binding data was analyzed using FortéBio data analysis software.
ELISA assay to determine antibody binding activity to SARS-CoV-2 RBD 96 well assay plates were coated with 100 µL per well of SARS-CoV-2 WT-RBD, E484K-RBD, E484Q-RBD, or E484A-RBD recombinant proteins at 1 µg/mL in PBS overnight at 4°C, respectively. After standard washing, 5% skim milk (200 µL per well) was added to block for 2 h at 37°C. After washing wells three times with PBS-0.05% Tween 20 (MP), 100 µL semilogarithmic dilutions in PBS of antibodies were added to each well and incubated at 37°C for 2 h. After washing wells 3 times with PBS-0.05% Tween 20, plates were incubated with 1:5000 dilutions of HRP-labelled Goat Anti-Human IgG (H + L) (Beyotime, China) in 5% skim milk at 37°C for 1 h. After washing wells 6 times with PBS-0.05% Tween 20, 100 µL per well of TMB/E solution (Merck Millipore) was added and developed at room temperature for 15 min. Reactions were stopped by adding 50 µL 1 M sulphuric acid and OD values at 450 nm were measured.

SARS-CoV-2 foci reduction neutralization test
Antibodies were serial diluted with DMEM and mixed with 200 FFU Wuhan-Hu-1 (wildtype) or Delta variant authentic SARS-CoV-2 viruses. After incubation at 37°C for 1 h, antibody-virus mixtures were added to a 96-well plate cultured with Vero E6 cells and incubated at 37°C in 5% CO 2 for 1 h. After removing the inocula, plates were overlaid with 100 μL 1.6% carboxymethylcellulose warmed to 37°C per well After culturing for 24 h, overlays were removed and the cells were fixed with 4% paraformaldehyde (Biosharp, China) and permeabilized with 0.2% Triton X-100 (Sigma, USA). Cells were incubated with a human anti-SARS-CoV-2 nucleocapsid protein monoclonal antibody (obtained by laboratory screening) at 37°C for 1 h. After washing with 0.15% PBST three times, cells were incubated with an HRP-labeled goat antihuman secondary antibody (Cat. No.: 609-035-213, Jackson ImmunoResearch Laboratories, Inc. West Grove, PA) at 37°C for 1 h. Plates were washed with 0.15% PBST three times, before the foci were visualized by TrueBlue Peroxidase Substrate (KPL, Gaithersburg, MD), and counted with an ELISPOT reader (Cellular Technology Ltd. Cleveland, OH). The foci reduction neutralization test (FRNT50) was calculated by Spearman-Karber method. The SARS-CoV-2 WT was isolated from a COVID-19 patient [25]. This virus is stored in Guangzhou Customs District Technology Center BSL-3 Laboratory. The SARS-CoV-2 Delta (B.1.617.2) strain is stored at Guangdong Provincial Center for Disease Control and Prevention, China. Experiments using SARS-CoV-2 authentic viruses were conducted in Guangzhou Customs District Technology Center BSL-3 Laboratory.

Pseudovirus-neutralization assay
SARS-CoV-2 WT, Omicron BA.1.1.529, BA.2.12.1, BA.2.13, and BA.4/BA.5 spike plasmids were constructed using the pcDNA3.1 vector. G*ΔG-VSV virus (VSV G pseudotyped virus) was used to infect 293T cells, and spike protein-expressing plasmid was used for transfection at the same time. After culture, the supernatant containing pseudovirus was collected, filtered, aliquoted, and frozen at −80°C for further use. Monoclonal antibodies were serially diluted (threefold) in DMEM (GIBCO, USA) and mixed with pseudovirus in 96-well plates. After incubation at 5% CO2 and 37°C for 1 h, digested Huh-7 cells were seeded. After 24 h of culture, supernatant was discarded and d-luciferin reagent (PerkinElmer, USA) was added to react in the dark. The luminescence value was measured using a microplate spectrophotometer (PerkinElmer, USA). IC50 was calculated by a four-parameter logistic regression model using PRISM v 8.0.1.

L452-contacting mAbs show IGHV germline gene preference
Analysis of 2,438,500 SARS-CoV-2 sequences deposited in GISAID database (https://www.gisaid.org/) showed that L452 is one of the most active mutation hotspot on RBD ( Figure S1A). Structural analysis showed that L452 is in proximity to the RBD-ACE2 binding interface ( Figure S1B), but pointing away from the interface to form a hydrophobic patch with F490 (we noted 490 is also a mutation hotspot) and L492 ( Figure S1C). Substitution of L452 to a charged residue R led to a moderate increase in affinity towards ACE2 receptor ( Figure S1C,D), increased ACE2 affinity was shown to enhance infectivity of human cells [5].
Sequence analysis showed that IGHV1-69-encoded mAbs isolated from vaccination and Omicron BA.1 breakthrough infection but not from Delta breakthrough infection are highly similar with the structurally characterized L452-contacting mAbs encoded by IGHV1-69 ( Figure S6A). Notably, most of the IGHV1-69-encoded clones isolated from Omicron BA.1 breakthrough infection utilize IGLV1-40 lightchain gene and have a preferred length of 17 aminoacid HCDR3s and a GYSGYG/D-like motif, which are related to our previous reported R1-32-like antibody lineage (non-ACE2 competitive) that isolated from primary SARS-CoV-2 ancestral strain infection or vaccination (Figure 4(D), Figure S2A, Figure  S6B). By reanalyzing the binding affinity data [16], we found that the IGHV1-69-encoded clones isolated from Omicron BA.1 breakthrough infection showed around 10-fold reduced binding affinity to the Delta variant (Figure 4(C,D)). Structural analysis revealed that the epitope residues of R1-32-like mAbs (R1-32 and FC08) are not affected by mutations in Omicron BA.1, whereas most of ACE2-competitive IGHV1-69 mAbs largely interact with E484 ( Figure S6B). We showed by ELISA that ACE2-competitive IGHV1-69 mAbs, LY-CoV555, MW01, and DH1043, are generally sensitive to E484 mutations (Omicron BA.1 has E484A) but the non-ACE2-competitive IGHV1-69 mAbs, R1-32 and FC08, are relatively insensitive to the E484 mutations (Figure 4(E)). The above observations suggest a preferential recalling of the non-ACE2-competitive IGHV1-69-encoded memory B cell response following Omicron BA.1 breakthrough infection, thereby increasing selection for L452.

Discussion
In summary, we report detection of three shared IGHV1-69-encoded antibody clonotypes in broad population exposed to SARS-CoV-2 antigen.
Structural and genetic analysis revealed that germline IGHV1-69 genes encode multiple hydrophobic residues at the tip of HCDR2 loop to allow contact to the L452 formed hydrophobic patch. This binding (E) IC50 values of the IGHV1-69-encoded L452-contacting mAbs against WT or B.1.617.2 (Delta) viruses. green, IC50 ≤ 100 ng/mL; blue, 100 ng/mL < IC50 < 1,000 ng/mL; red, IC50 ≥ 1,000 ng/mL; *, IC50 ≥ 5,000 ng/mL. Statistical tests in panel B and D are performed using two-tailed Wilcoxon signed-rank tests of paired samples. *, p < 0.05; **, p < 0.01, p < 0.001, ***. mode is shared by most characterized IGHV1-69encoded neutralizing antibodies. Unlike other IGHVs-encoded L452-contacting antibodies, IGHV1-69-encoded antibodies appear to have preexisting ability to bind L452 without requiring VDJ recombination or somatic hypermutation. These unique characteristics and frequent presentation of IGHV1-69-encoded mAbs in human antibody repertoires led us to propose that this class of antibodies contributes to immune pressure driving the L452R mutation. Consistently, we demonstrated experimentally that L452R was able to abrogate binding and neutralization by most known IGHV1-69-encoded L452contacting mAbs.
In addition to natural SARS-CoV-2 infection, SARS-CoV-2 vaccination also induced the enrichment of L452-contacting mAbs encoded by IGHV1-69, suggesting an established population immune pressure from prior infections or vaccinations. However, the preferential induction of IGHV1-69 was not observed in breakthrough infections by the L452R bearing Delta variant, indicating that L452R substitution not only confer virus ability to evade neutralization by IGHV1-69-encoded mAbs but also ability to suppress reactivation of immune memory. In this respect, IGHV1-69 antibody response is different to IGHV3-53/3-66 responses that even though Omicron variants contain the K417N mutation, IGHV3-53/3-66-encoded antibodies (also a shared antibody response) could be recalled following Omicron BA.1 breakthrough infection with some recalled mAbs showing K417N resistance [15]. These observations confirmed that certain spike mutations may not only evade physical binding by antibodies but also modulate antibody responses by suppressing induction of certain antibody classes.
Existing studies have shown that certain antibodies against a particular epitope share a restricted set of immunoglobulin heavy chain variable region genes and are often quite similar in overall structure and function [30,31]. For example, broadly neutralizing antibodies targeting the E2 antigenic region 3 (AR3) of HCV are mainly derived from IGHV1-69 [32,33]. IGHV1-69 has been also identified as one of the major classes of antibodies against the Influenza hemagglutinin (HA) stem region [34,35]. In these studies, it has been demonstrated that IGHV1-69 germline-encoded HCDR2 hydrophobic residues are essential for HCV and Influenza epitope recognition [31], consistent with our finding on IGHV1-69encoded SARS-CoV-2 S-specific antibodies. Therefore, a recurring theme emerges that hydrophobic epitopes of diverse viral envelope glycoproteins can be preferentially recognized by germline-encoded hydrophobic residues at the tip of the HCDR2 loop of IGHV1-69-encoded antibodies. Such class of antibodies can be readily produced by multiple individuals, representing an immunological solution in response to emerging pathogens, and demonstrating an evolutionary advantage for primates to possess the IGHV1-69 gene [36]. However, in the case of SARS-CoV-2, the virus demonstrates a remarkable ability to adapt and evade the host immune response.
Immune evasion by SARS-CoV-2 is highly complex, and other selection pressures could exist for the emergence of L452R. Nevertheless, our analysis points to a link between L452 substitution and herd immunity that is constituted by IGHV1-69-encoded neutralizing antibodies. Insights into antibody-mediated selection pressure on SARS-CoV-2 should provide theoretical framework for better understanding of variant genesis and variant immune evasion.