Molecular basis of SARS-CoV-2 Omicron variant evasion from shared neutralizing antibody response

Abstract

Understanding the molecular features of neutralizing epitopes is important for developing vaccines/therapeutics against emerging SARS-CoV-2 variants. We describe three monoclonal antibodies (mAbs) generated from COVID-19 recovered individuals during the first wave of the pandemic in India. These mAbs had publicly shared near germline gene usage and potently neutralized Alpha and Delta, poorly neutralized Beta, and failed to neutralize Omicron BA.1 SARS-CoV-2 variants. Structural analysis of these mAbs in complex with trimeric spike protein showed that all three mAbs bivalently bind spike with two mAbs targeting class 1 and one targeting a class 4 receptor binding domain epitope. The immunogenetic makeup, structure, and function of these mAbs revealed specific molecular interactions associated with the potent multi-variant binding/neutralization efficacy. This knowledge shows how mutational combinations can affect the binding or neutralization of an antibody, which in turn relates to the efficacy of immune responses to emerging SARS-CoV-2 escape variants.

Graphical abstract

Patel et al. describe how a combination of certain mutations affect the binding or neutralization of an antibody and thus have implications for predicting structural features of emerging SARS-CoV-2 escape variants and to develop vaccines or therapeutic antibodies against these.

Introduction

SARS-CoV-2 Omicron subvariants are continuously emerging and escaping therapeutic monoclonal antibodies (mAbs) and vaccines.^{1 , 2 , 3} Mutations acquired in the spike protein of SARS-CoV-2 variants, a target for neutralizing antibodies (nAbs), are primarily responsible for this immune escape.^{1 , 4 , 5} Identifying nAbs/non-nAbs to these variants and determining their prevalence in the human population allows us to understand the shared mechanisms of immune protection among diverse populations.^{6 , 7} Since the emergence of COVID-19, >11,000 SARS-CoV-2 mAbs have been identified.⁸ Among these, nAbs encoded by human antibody heavy-chain variable germline genes such as IGHV3-53/3-66, IGHV1-58, IGHV3-30, and IGHV1-69 are commonly observed in many individuals across the globe.⁸ These related rearrangements, known as a public antibody response, suggest a shared immune response with a similar genetic makeup and modes of antigen recognition that has been found in large number of individuals infected with influenza, dengue, malaria, HIV, and SARS-CoV-2.^{6 , 7 , 9 , 10 , 11 , 12 , 13 , 14} Mapping the immunogenetic makeup, structure, and function of these public clonotypes allows us to better understand how certain mutations affect the binding of an antibody and thus potentially expedite antibody re-purposing for emerging variants. It is established that SARS-CoV-2 variants bearing K417N/N501Y mutations evade IGHV3-53/3-66 RBD mAbs.^{6 , 14} These antibodies are primarily encoded by near germline sequences and are commonly found in populations residing in distinct geographical regions.^{6 , 13 , 14} However, SARS-CoV-2 variant evasion from the IGHV3-30 shared antibody response is unclear.

We recently published a panel of 92 RBD-binding mAbs isolated from five individuals infected with the ancestral SARS-CoV2 strain in India and identified a potent class 3 broad-spectrum antibody capable of neutralizing all highly evasive Omicron variants.^{15 , 16} Here, we focused on three mAbs that potently neutralize the ancestral WA.1 strain, but differentially neutralize SARS-CoV-2 variants for further characterization. The immunogenetic analysis confirms that all three mAbs were encoded by IGHV3-53/66 and IGHV3-30 genes and were publicly shared.¹⁵ While the cryo-EM structure of all three mAbs showed bivalent spike binding, two mAbs (002-02 and 034-32) targeted the class 1 RBD epitope, whereas mAb 002-13 targeted a relatively conserved class 4 epitope. A Detailed look at the molecular interactions at each mAb’s epitope-paratope surface allowed us to predict how mutations of certain residues in key variants of concern (VOCs) might impact antibody functionality and their role in immune evasion.

Results

Identification and characterization of shared human mAbs to SARS-CoV-2

In this study, we have selected 3 out of 92 previously identified RBD-specific mAbs for further characterization.¹⁵ These three mAbs, referred to as 002-13, 002-02, and 034-32 have heavy-chain VJ pairings encoded by IGHV3-30, IGHD2-8, IGHJ4; IGHV3-66, IGHD4-17, IGHJ4; and IGHV3-53, IGHD1-1, IGHJ6 immunoglobulin genes, respectively, whereas their light-chain VJ pairings were encoded by IGLV6-57, IGLJ2; IGK3-20, IGKJ4; and IGK1-9, IGKJ3 genes, respectively (Figure 1A). Genetic analysis of these three mAbs showed that heavy-chain variable (V) genes of all three mAbs were encoded by a shared public antibody response (Figures 1B–1E and S1) as documented in the CoV-AbDab database of all RBD-specific mAbs (n = 6,520) isolated from SARS-CoV-2-infected/vaccinated individuals.⁸ Public clonotypes were defined as groups of sequences with the same VH gene, the same JH gene, and the same junction length and high CDRH3 amino acid sequence identity (>80%) between donors.⁶ Interestingly, the antibody gene IGHV3-30, IGHJ4 of 002-13 mAb is the most frequent VJ pairing used by SARS-CoV-2 RBD mAbs (Figure 1D). Heavy-chain V gene IGHV3-30 of mAb 002-13 is the second most frequently used IGHV gene among all RBD mAbs (Figure 1B). In total, 71 of 002-13-like shared mAbs that exhibit the presence of a conserved CxGGxC motif in their complementarity-determining region (CDR) H3 (CDRH3) (Figure S1A; Table S1) encoded by an IGHD2-8 gene were found in different cohorts.⁸ Out of 71 mAbs, 55 mAbs (77%) have a CDRH3 length of 20, 13 mAbs (18%) have a CDRH3 length of 21, and the 2 remaining mAbs have a CDRH3 length of 18 and 19 (Table S1). The IGHV genes of 002-02 (IGHV3-66) and 034-32 (IGHV3-53) have already been described earlier in detail as a shared clonotype antibody response that shows the characteristic motifs of NY and SGGS in their CDRH1 and CDRH2 regions, respectively, preferred IGHD4-17 gene and a short CDRH3 length of 9–12 amino acids with high sequence diversity (5, 12, 13) (Figures S1B and S1C).⁸

Figure 1.

Genetic information of SARS-CoV-2 RBD-specific shared mAbs

(A) Immunogenetic information of the three SARS-CoV-2 mAbs.

(B) Heavy-chain variable gene distribution of SARS-CoV-2 RBD-specific human mAbs (n = 6,520) documented in the CoV-AbDab dataset.

(C) Light-chain variable gene distribution of SARS-CoV-2 RBD-specific human mAbs documented in the CoV-AbDab dataset.

(D) Heavy-chain VJ gene bar plot of SARS-CoV-2 RBD-specific human mAbs documented in the CoV-AbDab dataset.

(E) Light-chain VJ gene bar plot of SARS-CoV-2 RBD-specific human mAbs documented in the CoV-AbDab dataset.

Next, we revealed that all three mAbs strongly bind the spike protein with K_d values in low nM to pM range, by both BLI (Figure S2) and mesoscale binding assay (Mesoscale Discovery) (Figure 2A). In addition, in agreement with binding data they all potently neutralize the ancestral WA.1 live virus in a focus reduction neutralization mNeonGreen (FRNT-mNG) assay (Figures 2B and 2C).^{15 , 16} Taken together, these results confirm high binding affinity and potent neutralizing capacity of all three shared mAbs against the SARS-CoV-2 WA.1 strain.

Figure 2.

Binding, neutralization, and affinity analysis of selected mAbs toward the WA.1 strain

(A) Three SARS-CoV-2 mAbs were tested for binding to the WA.1 RBD protein.

(B) Live virus neutralization curves of the three mAbs against live WA.1 SARS-CoV-2. Neutralization was determined on using a focus reduction neutralization mNeonGreen (FRNT-mNG) assay on Vero cells.

(C) Fifty percent focus reduction neutralization titers (FRNT-mNG₅₀) for the three SARS-CoV-2 mAbs against WA.1 are shown.

Epitope mapping of mAbs 002-02, 002-13, and 034-32

To delineate the molecular determinants conferring epitope recognition and to understand the mechanism of their potent neutralization against WA.1 strain, we solved the cryo-EM structures of WA.1 spike-6P (spike-hexapro) in complex with each of the three mAbs (002-13, 002-02, and 034-32) in their native full-length IgG form (Figures 3 and 4 ). The structures show bivalent binding modes for all three mAbs, revealing two distinct neutralization mechanisms (Figure S4). Below we summarize our observations.

Figure 3.

Cryo-EM structure of 002-13 in complex with WA.1 spike trimer explains its broad neutralization activity

(A) Cryo-EM structure of WA.1 spike-6P trimer in complex with mAb 002-13. Overall density map at contour level of 5.4 σ showing the antibody binding in the RBD up-conformation. Each spike protomer is shown in gray, yellow, or green; light and heavy chains of each Fab region are shown in blue/magenta and light blue/pink, respectively. A model for one complex between Fab and RBD is shown to the right. The positions of all Fab complementarity-determining region (CDR) regions are labeled.

(B) Surface representation of RBD with relative positions of all CDR loops. The mapped epitope surface in the RBD is highlighted in magenta.

(C, E, and F) Interaction details at the 002-13-RBD interface.

(D) Heavy-chain CDR3 loop in density map.

(G) 002-13 binding on RBD sterically block ACE2 binding.

(H) Comparison of 002-13 binding mode with other class 4 mAbs.

(I) Zoom-in view comparing the heavy-chain CDR3 loop positions of 002-13 vs. COVA1-16. CDR3 amino acid sequence of 002-13 and COVA1-16 is shown below.

(J) Locations of Beta (yellow), Delta (red), and Omicron (green) mutations on the RBD relative to the 002-13 epitope site (black outline). See also Figures S3 and S4.

Figure 4.

Cryo-EM structure of 002-02 in complex with WA.1 spike trimer

(A) Cryo-EM structure of WA.1 spike-6P trimer in complex with mAb 002-02. Overall density map at contour level of 3.6 σ showing the antibody binding two RBDs in the up-conformation. Each protomer of spike is shown in gray, yellow, or green; the light and heavy chains of each Fab region are shown in blue/magenta and light blue/pink, respectively. A model one Fab-RBD complex is shown to the right and the positions of all Fab CDR regions are labeled.

(B) Surface representation of the RBD showing the relative positions of all CDR loops. The mapped epitope surface in the RBD is highlighted in orange.

(C–F) Interaction details of the 002-02-RBD interface.

(G) Locations of Beta (yellow), Delta (red), and Omicron (green) mutations on RBD relative to the 002-02 epitope site (black outline). See also Figures S4 and S5.

mAb 002-13

The cryo-EM structure of 002-13 in complex with WA.1 spike-6P (Figures 3 and S3) resolved at 3.8 Å global resolution revealed a conserved epitope on the inner face of the RBD, aligning with RBD-7/class 4 epitopes only accessible in up-conformation. There is clear intra-spike bivalent binding, where each Fab region of the full-length IgG recognizes adjacent RBDs in a spike trimer (Figures 3A and S4). The 002-13 mAb belongs to the public clonotype encoded by IGHV3-30 and IGLV6-57 that has not been structurally characterized before. Notably, the 22-residue long CDRH3 region encoded by the IGHD2-8 gene of the 002-13 mAb contains a CxGGxC motif, which is shared by the other 81 RBD-specific mAbs (Figure S1A) documented in the CoV-AbDab database.⁸ Like most of the other class 4 antibodies, 002-13 RBD binding is dominated by the heavy chain contributing ∼76% of total interaction with a total buried surface area of ∼887 Å² (Figure 3B).^{17 , 18} Most of the heavy-chain interactions are mediated through the CDR3 region that forms a foot-like loop, stabilized by an intra-loop disulfide bond between residues C105 and C110 of the CxGGxC motif (Figures 3C and 3D). We observe that multiple interactions involving the residues in the RBD region S371-C379 and the heavy-chain CDR3 loop are responsible for epitope recognition (Figure 3C). Heavy-chain CDR2 residues D57 and S56 engage the RBD residue K386 through a salt bridge and hydrogen bond (Figure 3E). The light chain of 002-13 contributes minimally to RBD binding, only the CDR2 loop of the light chain comes into RBD proximity to make a hydrogen bond with the side chain of RBD residue T415 (Figure 3F). Although 002-13 binds outside the receptor-binding motif (RBM) surface, it can sterically block ACE2 binding through its light-chain orientation, as previously observed in ACE2 competition profiling (Figure 3G).¹⁵

Structural comparison of 002-13 with another two class 4 mAbs (COVA1-16 and CR3022) shows a distinct binding pose for 002-13 (Figure 3H); in addition, a unique small side-chain-containing amino acid sequence in the CxGGxC motif of the heavy-chain CRD3 loop allows it to go much deeper into the RBD pocket facilitating extensive interactions in this region compared with other class 4 mAbs (Figure 3I).

We then marked key mutations present in Beta (yellow), Delta (red), and Omicron (green) variants within the 002-13 epitope surface and observed that, while all VOCs except for Omicron carry no mutations, Omicron carries three mutations (S371L, S373P, and S375F) within the 002-13 epitope (Figure 3J). This suggests that, while the binding and neutralization of 002-13 mAb will be preserved for most SARS-CoV-2 variants, it might be impacted toward Omicron as these mutations are known to induce a local conformational change in the Omicron RBD structure and thus could exclusively evade Omicron.^{19 , 20}

mAbs 002-02 and 034-32

Antibodies 002-02 and 034-32 were isolated from two different individuals and are encoded by public clonotype genes IGHV3-53/3-66 (Figure 1A). They both show very similar properties with high binding specificity toward SARS-CoV-2 RBD, effectively competing with ACE2 and potently neutralizing WA.1¹⁵ (Figure 2). To define the details of epitope recognition, we determined the cryo-EM structure of both 002-02 (Figures 4 and S5) and 034-32 (Figures S6 and S7) in complex with WA.1 spike-6P at a resolution of 3.8 and 4.3 Å, respectively. For both complexes, we observe intra-spike bivalent binding, where each Fab region of IgG binds two neighboring RBDs in the spike trimer in the up-conformation, as observed previously for similar class-specific mAbs.²¹ The RBD that does not engage in binding Fab remains in the down conformation (Figure 4A). Both mAb structures recognize epitopes in the top RBD pocket that align with the RBM surface, suggesting direct ACE2 competition and based on this they are classified as RBD-2/class 1 antibodies. Since 002-02 and 034-32 recognize the RBD in a very similar manner, we focus our structural analysis on mAb-RBD recognition in the locally refined map for 002-02.

While all CDR loops are involved in epitope recognition (Figures 4A and 4B), most RBD contacts are dominated by the heavy chain, contributing ∼70% of the total of 1,058 Å² of buried surface between mAb and RBD (Figure 4B). Primary interactions in the heavy chain are mediated by CDR1 and CDR2 regions. Most mAbs that belong to class 1 antibodies are encoded by public clonotype genes IGHV3-53/IGHV3-66.^{6 , 14} The common features among these mAbs include a conserved NY and SGGS motif in the CDR1 and CDR2 regions, respectively, that contribute significantly toward RBD binding.¹³ We also observed a network of hydrogen bonds with the RBD through the CDR2 SGGS motif. The side chain of S53 and S56 in the CDR2 heavy chain engages in a hydrogen bond with side chains of Y421 and D420 in RBD, respectively (Figure 4C). However, the CDR3 loop heavy-chain residues in this mAb class varies more. In 002-02, the heavy-chain CDR3 residue D101 forms a hydrogen bond with K417 and Y453 in the RBD (Figure 4D). In the light chain, CDR1 and CDR3 make some contact with the inner left side of the RBD. The S30 and Y32 residues in the CDR1 region of the light chain make a hydrogen bond with Q498 and R403 in the RBD, respectively (Figure 4E). Also, S93 in the CDR3 region of the light chain interacts with Y505 and D405 (Figure 4F).

Like 002-13, we also mapped mutations found in Beta (yellow), Delta (red), and Omicron (green) variants onto the 002-02/034-32 epitope (Figure 4G). While Delta carries no mutation within the 002-02/034-32 epitope surface, three of the Beta mutations (K417N, E484K, and N501Y) fell within its epitope, suggesting no variation in binding and neutralization for Delta but weakened binding and neutralization for Beta. However, six Omicron mutations (K417N, S477N, Q493R, G496S, Q498R, and N501Y) lay within the 002-02/034-32 epitope surface and are predicted to evade Omicron binding and neutralization. Collectively, based on these observations both 002-02 and 034-32 mAbs will be less or ineffective toward both Beta and Omicron variants.

Assessing binding and neutralization breadth toward SARS-CoV-2 variants

To link the paratope mutation landscape in VOCs to the antibody function, we tested binding and neutralization of these three mAbs against SARS-CoV-2 variants. In agreement with the structure-based prediction, the binding of 002-13 (class 4 antibody) remained unaffected toward Alpha, Beta, and Delta variants as these variants contain no mutations within the 002-13 epitope and showed moderately reduced (∼2.7-fold) binding to Omicron (Figures 3I, 5A, and 5G). In agreement with binding data, the neutralization potency of 002-13 remained unperturbed in Alpha, Beta, and Delta and showed no observable neutralization of the Omicron virus (Figures 5E and 5G). Along that line, binding of 002-02 and 034-32 (class 1 antibodies) retained for Alpha and Delta variants to the same affinity as of WA.1, showed 3- and 150-fold reduced affinity to Beta, respectively, and no observable binding to Omicron (Figures 5B, 5C, and 5G). Following this thread, both 002-02 and 034-32 neutralize Alpha and Delta variants with the same potency as WA.1 and showed 4- and 17-fold reduced potency to Beta, respectively, and showed a complete loss of neutralization against Omicron (Figures 5E–5G). This is further supported by the fact that the unique K417N mutation (present in Beta but not in Delta) would result in a loss of a hydrogen bond and possibly a salt bridge with D101 in the heavy-chain CDR3 (Figure 4D) and subsequent Beta-variant-specific loss of binding and neutralization for 002-02 and 034-32. This was also confirmed by a 2-fold decrease in the calculated ddG value of −46.23 ± 10.5 kcal/mol based on molecular mechanics/Poisson-Boltzmann surface area free energy for the single K417N mutant in the 002-02-spike structure compared with the WA.1 ddG value of −82.62 ± 9.57 kcal/mol (Figure S8).

Figure 5.

Binding affinity and neutralization analysis of selected mAbs against SARS-CoV-2 variants

(A–C) Three potent neutralizing mAbs were tested for binding to spike proteins of SARS-CoV-2 WA.1, Alpha, Beta, Delta, and Omicron (BA.1) variants of concern (VOCs). Curves shown are the best fit one-site binding curves calculated by Prism 9.0.

(D–F) Live virus neutralization curves and FRNT₅₀ values of three potent mAbs for WA.1, Alpha, Beta, Delta, and Omicron (BA.1) SARS-CoV-2 VOCs are shown. Neutralization was determined on Vero-TMPRSS2 cells using a focus reduction neutralization assay. Values are shown for two independent replicates. Error bars indicate SD.

(G) Table summarizing the apparent dissociation constant (K_D) and neutralization potency of mAbs against SARS-CoV-2 variants derived by electrochemiluminescence antibody binding assay.

(H) Table summarizing the properties of RBD targeting antibodies and their neutralization toward SARS-CoV-2 variants (variant in bold show loss of neutralizing potency with specified mAb).

Altogether, these data catalog the epitope class-specific antibody susceptibility toward existing SARS-CoV-2 variants and can inform their action on a newly emerging variant.

Discussion

Understanding how SARS-CoV-2 mAbs achieve broad neutralization or are rendered ineffective by viral mutations provides insight not only about natural immunity, but is critical to develop broadly effective therapeutic mAbs and guide vaccine design.^{6 , 13 , 22 , 23 , 24} Moreover, defining antibody-antigen interactions is critical for the rapid re-evaluation of existing antibody-based therapeutics toward continuously emerging SARS-CoV-2 variants. This, overlaid with the immuno-genetic makeup of the antibodies shared by a large population further informs our understanding of the public immune response and their antigenic drift from variants. For example, certain antibody responses are repeatedly shared among large numbers of individuals regardless of their genetic origins, as has been observed previously during different pathogen infections including influenza, dengue, HIV, and malaria.^{9 , 10 , 11 , 12} With SARS-CoV-2, these are encoded by IGHV3-53/66, IGHV1-58, IGHV3-30, and IGHV1-69, which are found both following natural infection and post-vaccination.^{6 , 7 , 13} Such information can be collectively used to fine-tune the immune response focused on broad and potent neutralizing epitopes through antigen design for a universal vaccine.^{22 , 24} Recently, based on the information from shared public clonotypes of HIV-1 bnAbs, a V2-apex region-specific immunogen has been successfully designed.²⁵

We recently reported the isolation of 92 SARS-CoV-2 RBD-specific mAbs from COVID-19 recovered individuals from India during the first wave of the pandemic and identified a broadly neutralizing class 3 antibody (002-S21F2), capable of neutralizing BA.1-BA.5 Omicron subvariants.¹⁵ Out of 92, three SARS-CoV-2 nAbs (002-13, 002-02, and 034-32) characterized in this study belong to shared public antibody responses. Sequence analysis of 6,520 published SARS CoV-2 RBD-specific mAbs define 002-13 as a public clonotype encoded by IGHV3-30, IGHJ4 genes with >80% of these exhibiting IGHD2-8 gene usage and the presence of a CxGGxC motif in their CDRH3 region that has not been structurally characterized.⁸ While the other two mAbs, 002-02 and 034-32, are encoded by shared IGHV3-53/3-66 antibody genes as previously shown by others,^{6 , 13 , 14} the cryo-EM structures for these three mAbs in complex with trimeric spike protein show class 4 epitope recognition by 002-13 and class 1 epitope recognition by 002-02 and 034-32. The structures further allowed us to define their epitope-paratope interfaces in detail in relation to the locations of SARS-CoV-2 variant mutations to predict viral immune escape. While there was no observable difference in the antibody functionality for variants containing mutations that lie outside the mapped epitope surface of a particular antibody, there was a remarkable drop in binding affinity and neutralization of the antibody when the mutations mapped to the antibody footprint. Most broad nAbs recognize all variant antigens that either carry no mutations within their epitopes or the mutations in epitope region are favored by mAb-specific molecular interactions, as we observed for class 3 mAb 002-S21F2.¹⁵ Here, we show all three mAbs potently neutralized the ancestral WA.1 strain, but differentially neutralize other variants, primarily due to the presence of evading mutations present in their epitope antigenic sites, similar to the other well-characterized mAbs recognizing the same epitope classes (Figure 5H).⁸ Major mutations responsible for Beta evasion are K417N and E484K for 002-02 and 034-32 mAbs, also observed previously for IGHV3-53/3-66 shared antibody responses.^{6 , 13} Omicron, which contains six epitope mutations (K417N, S477N, Q493R, G496S, Q498R, and N501Y) within 002-02/034-32 and three mutations (S371L, S373P, and S375F) within the 002-13 binding site would collectively lead to major immuno-escape, especially as some mutation residues participate in direct interaction with mAbs. Although 002-13 showed only moderate reduction in binding affinity, it showed no neutralization toward Omicron, suggesting that additional factors might play a role in 002-13-specific Omicron escape. One explanation could be that Omicron mutations that favor spike up-conformation would likely promote ACE2 interaction and reduce 002-13 mAb competition.²⁶ Our findings suggest that immune pressures exerted by the shared antibody response to SARS-CoV-2 are likely to cause evolution variants with mutations in the class 4 antibody epitope residues S371, S373, and S375. These mutations must be tracked to find effective solutions to combat emerging variants. Furthermore, the structure guided prediction made for three SARS-CoV-2 shared nAbs that potently neutralized the WA.1 strain holds true toward the functional efficacy of these mAbs against SARS-CoV-2 variants, including Omicron.

In summary, this study vastly improves our understanding of how Omicron escaped from shared antibody responses to SARS-CoV-2 elicited during the natural infection and has implications toward concepts for fast-tracking effective broad-range therapeutics against continuously emerging SARS-CoV-2 variants.

Limitations of the study

Our work is focused on the binding, neutralization, and structural characterization of three shared SARS-CoV-2 antibodies isolated from individuals recovered from COVID-19 in India. While the proposed bivalent binding mode of the three antibodies to the spike protein is consistent with our cryo-EM data and earlier studies, additional biophysical measurements would be needed to further strengthen this conclusion. These antibodies recognize different epitope classes on the spike protein RBD leading to differential variant neutralization, and likely immune evasion. All conclusions are based on structure-based predictions and results from live virus neutralization experiments of three selected antibodies, which may show different efficacy in animal model experiments.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Peroxidase AffiniPure F(ab’)2Fragment Goat Anti-Human IgG, Fcγ fragment specific	Jackson ImmunoResearch	Cat# 109-036-098; RRID: AB_2337596
AF647-CR3022	Dr. Jens Wrammert (Emory University, Atlanta, GA)	IEDB Cat# CR3022; RRID: AB_2848080
Bacterial and virus strains
nCoV/USA_WA1/2020	Dr. Mehul Suthar (Emory University, Atlanta, GA)	N/A
B.1.1.7	Dr. Mehul Suthar (Emory University, Atlanta, GA)	N/A
B.1.351	Dr. Mehul Suthar (Emory University, Atlanta, GA)	N/A
B.1.617.2	Dr. Mehul Suthar (Emory University, Atlanta, GA)	N/A
B.1.1.529	Dr. Mehul Suthar (Emory University, Atlanta, GA)	N/A
Chemicals, peptides, and recombinant proteins
Recombinant binding protein (SARS-CoV-2 spike)	Dr. Jens Wrammert Emory University	N/A
Methylcellulose	Sigma-Aldrich	Cat. #: M0512-250G
TrueBlue Peroxidase Substrate	KPL	Cat. #: 5067428
O-phenylenediamine (OPD) substrate	Sigma	Cat# P8787
Phosphate-citrate buffer	Sigma	Cat# P4809
Hydrogen peroxide	Fisher Scientific	Cat# 18755
Paraformaldehyde (16%)	Thermo Fisher Scientific	Cat# 28906
DMEM	Thermo Fisher Scientific	Cat# 11965118
Tween 20	Fisher Scientific	BP337-500
Pierce™ Protein A Agarose resin	Thermo Fisher Scientific	Cat# 20334
ExpiFectamineTM 293 transfection reagent	Thermo Fisher Scientific	Cat# A14524
Expi293™ Expression Medium	Thermo Fisher Scientific	Cat# A1435102
Critical commercial assays
V-PLEX SARS-CoV-2 Panel 24 (IgG) Kit	Mesoscale Discovery	Cat# K15575U
Deposited data
mAb sequences	This study	GenBank: ON882061 - ON882244
WA.1 spike-6P in complex with mAb 002-02	This study	PDB: 7U0Q and EMDB: 26263
WA.1 spike-6P in complex with mAb 002-013	This study	PDB: 7U0X and EMDB: 26267
WA.1 spike-6P in complex with mAb 034-32	This study	PDB: 7UOW and EMDB: 26656
Experimental models: Cell lines
VeroE6 C1008 cells	ATCC	Cat# CRL-1586; RRID: CVCL_0574
Vero TMPRSS2 cells	Dr. Mehul Suthar Emory University	RRID: CVCL_YQ48
Human: Expi293F	ThermoFisher	Cat # A14527; RRID: CVCL_D615
Oligonucleotides
Multiplex antibody cloning primer set	Davis et al. and Milligan et al.27,28	N/A
HuIgG-const-anti 5′ TCTTGTCCACCTTGGTGTTGCT-3′	Smith et al.29	N/A
Ab Vec sense 5′ GCTTCGTTAGAACGCGGCTAC-3′	Smith et al.29	N/A
Recombinant DNA
WA.1-spike-6P plasmids	Kumar et al.15	RRID: Addgene_154754
AbVec antibody expression vectors	Davis et al. and Milligan et al.27,28	N/A
Software and algorithms
GraphPad Prism (v9)	GraphPad Software, Inc.	GraphPad Prism https://www.graphpad.com:443/; RRID: SCR_002798
Discovery WorkBench 4.0	Mesoscale Discovery	RRID: SCR_019192
Viridot	Katzelnick et al.	https://github.com/leahkatzelnick/Viridot
Coot	Emsley et al.30	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/; RRID:SCR_014222
PHENIX	Adams et al.31	https://phenix-online.org; RRID: SCR_014224
UCSF ChimeraX	Goddard et al.32	https://www.cgl.ucsf.edu/chimerax/; RRID: SCR_015872
PDBePISA	EMBL-EBI33	https://www.ebi.ac.uk/pdbe/pisa/; RRID: SCR_015749
Molprobity	Williams et al.34	http://molprobity.biochem.duke.edu/; RRID: SCR_014226
IMGT/HighV-QUEST	Lefranc et al.35	http://imgt.org/HighV-QUEST; RRID: SCR_010749
Octet analysis studio software version 13.0	Sartorius	RRID: SCR_023267
PyMOL	Schrödinger36	https://pymol.org/2/; RRID: SCR_000305
Relion 3.1	Scheres,37	https://www3.mrc-lmb.cam.ac.uk/relion/index.php/Main_Page; RRID: SCR_016274
Cryosparc v3.3.2	Punjani et al.38	https://cryosparc.com/updates; RRID: SCR_016501
Isolde	Croll,39	https://isolde.cimr.cam.ac.uk/what-isolde/
Privateer	Agirre et al.40	https://www.ccp4.ac.uk/html/privateer.html
Alpha-Fold	Jumper et al.41	https://alphafold.ebi.ac.uk/
Other
96-well MaxiSorp plates	Thermo Fisher Scientific	Cat# 439454
MSD QuickPlex plate reader	Mesoscale Discovery	MESO QuickPlex SQ 120
Octet Red96	Fortebio	N/A
HisPur Ni-NTA column	Thermo Fisher Scientific	Cat# 88229
Superose-6 Increase 10/300 column	GE Healthcare	Cat# 17-5172-01
ELISPOT reader	Immunospot	CTL ImmunoSpot S6 Universal Analyzer
Vitrobot IV	Thermo Fisher Scientific	N/A

Resource availability

Lead contact

Further inquiries and requests for data, plasmids and resources should be directed to the lead contact Eric Ortlund (eortlun@emory.edu).

Materials availability

Antibody expression plasmids generated in this study (see key resources table) are available upon request from the lead contact with a completed Materials Transfer Agreement.

Experimental model and subject details

Cell lines

VeroE6-TMPRSS2 cells were generated and cultured as previously described.^{15 , 42} VeroE6-TMPRSS2 cells were cultured in complete DMEM in the presence of Gibco Puromycin 10 mg/mL (# A11138-03). VeroE6-TMPRSS2 cells were used to propagate all virus stocks. Cells were incubated at 37°C in the presence of 5% CO₂. The expi293F cells were maintained in Expi293 Expression Medium for SARS-CoV-2 spike, RBD and mAb protein expression. The expi293F cells were incubated at 37°C in the presence of 8% CO₂ with shaking at 125 RPM.

Viruses

All the SARS-CoV-2 viruses were expanded in a confluent VeroE6-TMPRSS2 cells T175 flask to generate a working stock. All viruses used in this study were deep sequenced and confirmed as previously described.^{15 , 42}

Method details

Expression and purification of SARS-CoV-2 RBD

The recombinant SARS-CoV-2 RBD gene was cloned, expressed and purified as previously described.^{15 , 16 , 43} Briefly, a recombinant form receptor-binding domain (RBD) from SARS-CoV-2, Wuhan-Hu-1 (GenPept:QHD43416) was cloned into the EcoRI-HindIII cloning site of a mammalian expression vector containing a CMV promoter (GenBank: FJ475055). Plasmid DNA was purified using the QIAGEN PlasmidPlus Maxi purification system. The expression plasmid DNA was mixed with the expifectamine transfection reagent. Complex was added to the expi293 cell suspensions shaking at 125 RPM and incubated overnight at 37°C in an 8% CO₂ humidified incubator. After 18 h, protein expression enhancers and antibiotics were added. Cultures were then incubated for an additional four days to allow for expression into the supernatant. Cell culture supernatants were harvested by centrifugation at 16,000xg for 15 min. Supernatants were sterile filtered through a 0.2 μm before purification. Purification was performed according to manufacturer’s instructions using 5 mL HisTALON Superflow Cartridges (Clontech Laboratories). Briefly, an additional 11.7 g/L of sodium chloride and 0.71 g/L of cobalt(II) chloride hexahydrate were added to culture supernatants, which were adjusted to pH 7.5. The supernatant was then loaded on to the column equilibrated with 10 column volumes of 50 mM phosphate 300 mM sodium chloride buffer pH 7.5 (equilibration buffer). The column was washed with 8 column volumes of equilibration buffer supplemented with 10 mM imidazole. Protein was eluted with 6 column volumes of equilibration buffer supplemented with 150 mM imidazole. The eluted protein was dialyzed overnight against 80 volumes of phosphate-buffered saline pH 7.2. The protein was filter-sterilized (0.2 mm) and normalized to 1 mg/mL by UV spectrophotometry using an absorption coefficient of 1.3 AU at 280 nm = 1 mg/mL. Proteins were aliquoted and stored at −80°C prior to use.

SARS-CoV-2 RBD-specific ELISA binding assays

The ELISAs were performed as described previously.^{15 , 16 , 43} Briefly, purified RBD was coated on 96-well MaxiSorp plates (Thermo Fisher, #439454) at a concentration of 1 μg/mL in phosphate-buffered saline (PBS) at 4°C overnight. The plates were washed with PBS containing 0.05% Tween 20. 3-fold serially diluted purified mAb was added and incubated at room temperature for 1 h. Plates were washed and the SARS-CoV-2 RBD specific IgG signal was detected by incubating with horseradish peroxidase (HRP) conjugated - anti-human IgG (Jackson ImmunoResearch Labs, #109-036-098). Plates were then washed thoroughly and developed with o-phenylenediamine (OPD) substrate (Sigma, #P8787) in 0.05M phosphate-citrate buffer (Sigma, #P4809) pH 5.0, containing 0.012% hydrogen peroxide (Fisher Scientific, #18755). Absorbance was measured at 490 nm.

Live SARS-CoV-2 neutralization assay

Neutralization titers to SARS-CoV-2 were determined based on either a focus-reduction neutralization mNeonGreen (FRNT-mNG) assay on Vero cells or FRNT assays based on Vero TMPRSS2 cells as previously described.^{15 , 16} Briefly, 100 pfu of SARS-CoV-2 (2019-nCoV/USA_WA1/2020), Alpha, Beta, Gamma, Delta and Omicron variants were used on Vero TMPRSS2 cells. Purified monoclonal was serially diluted 3-fold in duplicate starting at 10 μg/mL in a 96-well round-bottom plate and incubated for 1 h at 37°C. This antibody-virus mixture was transferred into the wells seeded with Vero-TMPRSS2 cells the previous day at a concentration of 2.5 × 10⁴ cells/well. After 1 h, the antibody-virus inoculum was removed and 0.85% methylcellulose in 2% FBS containing DMEM was overlaid onto the cell monolayer. Cells were incubated at 37°C for 16–40 h. Cells were washed three times with 1X PBS (Corning Cellgro) and fixed with 125 μL of 2% paraformaldehyde in PBS (Electron Microscopy Sciences) for 30 min. Following fixation, plates were washed twice with PBS and 100 μL of permeabilization buffer, was added to the fixed cells for 20 min. Cells were incubated with an anti-SARS-CoV spike primary antibody directly conjugated with alexaflour-647 (CR3022-AF647) for up to 4 h at room temperature. Plates were then washed twice with 1x PBS and imaged on an ELISPOT reader (CTL Analyzer). Foci were counted using Viridot (counted first under the “green light” set followed by background subtraction under the “red light” setting). IC₅₀ titers were calculated by non-linear regression analysis using the 4PL sigmoidal dose curve equation on Prism 9 (Graphpad Software). Neutralization titers were calculated as 100% x [1- (average foci in duplicate wells incubated with the specimen) ÷ (average number of foci in the duplicate wells incubated at the highest dilution of the respective specimen).

Immunogenetic analyses of antibody genes

The plasmid sequences were verified by Sanger sequencing (Macrogen sequencing, South Korea). The immunogenetic analysis of both heavy chain and light chain germline assignment, framework region annotation, determination of somatic hypermutation (SHM) levels (nucleotides) and CDR loop lengths (amino acids) was performed with the aid of IMGT/HighV-QUEST (www.imgt.org/HighV-QUEST).³⁵

Expression of human monoclonal antibodies

All transfections were done as described earlier.¹⁵ Briefly, expi293F cells were transfected with antibody expression plasmids at a density of 2.5 million cells per/mL using 1 mg/mL PEI-Max transfection reagent 40,000 MW (Polysciences). Supernatants were harvested 4–5 days post-transfection and tested for their SARS-CoV-2 RBD binding potential by enzyme-linked immunosorbent assay (ELISA). Supernatant with positive RBD binding signals was next purified using Protein A/G beads (Thermo Scientific), concentrated using a 30 kDa or 100 kDa cut-off concentrator (Vivaspin, Sartorius) and stored at 4°C for further use.

Electrochemiluminescence antibody binding assay

Binding analysis of SARS-CoV-2 mAb to spike protein was performed using an electrochemiluminescence assay as described earlier.¹⁵ Briefly, V-PLEX COVID-19 Panel 24 (Meso Scale Discovery) was used to measure the IgG1 mAb binding to SARS-CoV-2 spike antigens following the manufacturer’s recommendations. antigen coated plates were blocked with 150 μL/well of 5% BSA in PBS for 30 min. Plates were washed 3x with 150 μL/well of PBS with 0.05% Tween between each incubation step. mAbs were serially diluted for concentrations ranging from 10 μg/mL to 0.1 pg/mL and 50 μL/well were added to the plate and incubated for 2 h at room temperature with shaking at 700rpm. mAb antibody binding was then detected with 50 μL/well of MSD SULFO-TAG anti-human IgG antibody (diluted 1:200) incubated for 1 h at room temperature with shaking at 700rpm. 150 μL/well of MSD Gold Read Buffer B was then added to each plate immediately before reading on an MSD QuickPlex plate reader.

Octet biolayer interferometry (BLI) analysis

Octet BLI was performed using an Octet Red96 instrument (ForteBio, Inc.) as described earlier.^{15 , 44} A 5 μg/mL concentration of each mAb was captured on a protein A sensor and its binding kinetics were tested with serial 2-fold diluted RBD (600 nM–37.5 nM) and spike hexapro protein (100 nM–6.25 nM). The baseline was obtained by measurements taken for 60 s in BLI buffer (1x PBS and 0.05% Tween 20), and then, the sensors were subjected to association phase immersion for 300 s in wells containing serial dilutions of RBD or trimeric spike hexapro protein. Then, the sensors were immersed in BLI buffer for as long as 600 s to measure the dissociation phase. The apparent KD and KD2 values of the mAbs binding affinities for RBD and spike hexapro were calculated from all the binding curves based on their global fit to a 2:1 heterogeneous ligand binding model using Octet analysis studio software version 13.0.

Spike protein expression and purification

SARS-CoV-2 spike-6P trimer protein carrying WA.1 was expressed and purified by transfecting expi293F cells using WA.1-spike-6P plasmids as described previously.¹⁵ Transfections were performed as per the manufacturer’s protocol (Thermo Fisher). Briefly, expi293F cells (2.5 × 10⁶ cells/mL) were transfected using ExpiFectamine 293 transfection reagent (ThermoFisher, cat. no. A14524). The cells were harvested 4–5 days post-transfection. The spike protein was purified using His-Pur Ni-NTA affinity purification method. Column was washed with Buffer containing 25 mM Imidazole, 6.7 mM NaH₂PO₄.H2O and 300 mM NaCl in PBS followed by spike protein elution in elution buffer containing 235 mM Imidazole, 6.7 mM NaH₂PO₄.H2O and 300 mM NaCl in PBS. Eluted protein was dialyzed against PBS and concentrated. The concentrated protein was loaded onto a Superose-6 Increase 10/300 column and protein eluted as trimeric spike collected. Protein quality was evaluated by SDS-PAGE and by Negative Stain-EM.

Negative Stain – Electron microscopy (NS-EM)

Spike protein was diluted to 0.05 mg/mL in PBS before grid preparation. A 3 μL drop of diluted protein (∼0.025 mg/mL) was applied to previously glow-discharged, carbon-coated grids for ∼60 s, blotted and washed twice with water, stained with 0.75% uranyl formate, blotted and air-dried. Between 30-and 50 images were collected on a Talos L120C microscope (Thermo Fisher) at 73,000 magnification and 1.97 Å pixel size. Relion-3.1³⁷ or Cryosparc v3.3.2³⁸ was used for particle picking and 2D classification.

Sample preparation for Cryo-EM

SARS-CoV-2 spike-6P trimer incubated with the mAb (full-length IgG) at 0.7 mg/mL concentration. The complex was prepared at a 0.4 sub-molar ratio of mAb to prevent inter-spike crosslinking, mediated by bivalent binding of intact antibody. The complex was incubated at room temperature for ∼5 min before vitrification. Three μL of the complex was applied onto a freshly glow-discharged (PLECO easiGLOW) 400 mesh, 1.2/1.3 C-Flat grid (Electron Microscopy Sciences). After 20 s of incubation, grids were blotted for 3 s at 0 blot force and vitrified using a Vitrobot IV (Thermo Fisher Scientific) under 22°C with 100% humidity.

Cryo-EM data acquisition

Single-particle Cryo-EM data for WA.1 spike-IgG complexes of mAb 002-02, 002–13 and 034-32 were collected on a 300 kV Titan Krios transmission electron microscope (ThermoFisher Scientific) equipped with Gatan K3 direct electron detector behind 30 eV slit width energy filter. Multi-frame movies were collected at a pixel size of 1.0691 Å per pixel with a total dose of 63 e/Ų at defocus range of −0.7 to −2.7 μm.

Cryo-EM data analysis and model building

Cryo-EM movies were motion-corrected by Patch motion correction implemented in Cryosparc v3.3.2.³⁸ Motion-corrected micrographs were corrected for contrast transfer function using Cryosparc’s implementation of Patch CTF estimation. Micrographs with poor CTF fits were discarded using CTF fit resolution cutoff to ∼6.0 Å. Particles were picked using a Blob picker, extracted, and subjected to an iterative round of 2D classification. Particles belonging to the best 2D classes with secondary structure features were selected for heterogeneous 3D refinement to separate IgG bound spike particles from non-IgG bound spike particles. Particles belonging to the best IgG bound 3D class were refined in non-uniform 3D refinement with per particle CTF and higher-order aberration correction turned on. To further improve the resolution of the RBD-IgG binding interface a soft mask was created covering one RBD and interacting Fab region of IgG and refined locally in Cryosparc using Local Refinement on signal subtracted particles. All maps were density modified in Phenix³¹ using Resolve Cryo-EM. The combined Focused Map tool in Phenix was used to integrate high resolution locally refined maps into an overall map. Additional data processing details are summarized in Figures S3, S4, S5, S6 and Table S2.

The initial spike models for WA.1 (PDB:7lrt) as well as individual heavy and light chains of the Fab region of an IgG (generated with Alpha-fold)⁴¹ were docked into combine focused Cryo-EM density maps using UCSF ChimeraX.³² The full spike-mAb model was refined using rigid body refinement in Phenix, followed by refinement in Isolde.³⁹ The final model was refined further in Phenix using real-space refinement. Glycans with visible density were modeled in Coot and validated by Privateer.^{30 , 40} Model validation was performed using Molprobity.³⁴ PDBePISA³³ was used to identify mAb-RBD interface residue, to calculate buried surface area and to identify polar interaction. Figures were prepared in ChimeraX³² and PyMOL.³⁶

Molecular dynamics simulation

Molecular dynamics simulations were carried out to understand the effect of RBD mutations on the AB binding. MD simulations were carried using AMBER99SB force field⁴⁵ as implemented in GROMACS 2019.⁴⁶ The system was solvated with TIP3P water model and neutralized with salts ([NaCl] = 0.15 M). Electrostatics were calculated using the PME method⁴⁷ with a real space cut-off of 10 Å. van der Waals interactions were modeled using Lennard–Jones 6–12 potentials with a 14 Å cut-off. The temperature was maintained at 300 K using V-rescale; hydrogen bonds were constrained using the LINCS algorithm.⁴⁸ Energy minimization was carried out to reach a maximum force of no more than 10 kJ/mol using steepest descent algorithm. The time step in all molecular dynamics simulations was set to 2 fs. Prior to the production run, the minimized systems were equilibrated for 5ns with NVT and followed with NPT at 300 K.

To calculate the Binding energies for the wild and the mutants, 200 snapshots were extracted from the last 20 ns of the 80ns production run. The extracted 400 frames for the wild and the variant subjected to MM/PBSA calculations using the gmx_MMPBSA tool with the standard input parameters.⁴⁹ For the selected frames, PBC conditions were removed from the GROMACS output trajectory and protein-mAb complex were indexed.

Quantification and statistical analysis

Statistical analysis was performed with Prism 9.0.

Published: July 6, 2023

Data and code availability

• •Atomic coordinates and Cryo-EM maps for reported structures are deposited into the Protein DataBank (PDB) and the Electron Microscopy DataBank (EMDB) with accession codes PDB: 7U0Q and EMDB: 26263 for WA.1 spike-6P in complex with mAb 002-02, PDB: 7U0X and EMDB: 26267 for WA.1 spike-6P in complex with mAb 002–013, PDB: 7UOW and EMDB: 26656 for WA.1 spike-6P in complex with mAb 034-32.
• •Immunoglobulin sequences are available in GenBank under accession numbers Genbank: ON882061 - ON882244.
• •This paper does not report original code.
• •All data needed to evaluate the conclusions in the paper are present in the paper and/or the supplemental information.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.