Broad SARS-CoV-2 neutralization by monoclonal and bispecific antibodies derived from a Gamma-infected individual

Summary The pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has remained a medical threat due to the evolution of multiple variants that acquire resistance to vaccines and prior infection. Therefore, it is imperative to discover monoclonal antibodies (mAbs) that neutralize a broad range of SARS-CoV-2 variants. A stabilized spike glycoprotein was used to enrich antigen-specific B cells from an individual with a primary Gamma variant infection. Five mAbs selected from those B cells showed considerable neutralizing potency against multiple variants, with COVA309-35 being the most potent against the autologous virus, as well as Omicron BA.1 and BA.2, and COVA309-22 having binding and neutralization activity against Omicron BA.4/5, BQ.1.1, and XBB.1. When combining the COVA309 mAbs as cocktails or bispecific antibodies, the breadth and potency were improved. In addition, the mechanism of cross-neutralization of the COVA309 mAbs was elucidated by structural analysis. Altogether these data indicate that a Gamma-infected individual can develop broadly neutralizing antibodies.

between up and down conformations, thus representing a target where some epitopes are differentially exposed depending on RBD state. 4][8][9] More recently, the evolution of the wild-type (WT) virus into multiple variants that have acquired resistance to vaccines and mAb therapies has raised concerns about the longevity and efficacy of current treatment options, which are based on the original Wuhan Hu-1 S sequence.Some of these variants are characterized by higher transmissibility, virulence, and/or immune evasion compared to the WT virus, resulting in the WHO declaring the Alpha (Pango nomenclature B.1.1.7),Beta (B.1.351),Gamma (P.1), Delta (B.1.617.2), and Omicron (B.1.1.529)lineages as variants of concern (VOCs). 10In addition, variants of interest (VOIs) have also been reported worldwide.While the Alpha, Beta, Gamma, and Delta variants have a moderate number of mutations in the S protein, the latest Omicron variant, with its sub-lineages, contains more than 30 amino acid changes, of which almost half are located in the RBD, underpinning its substantial shift in antigenicity compared to early variants. 11,124][15][16] In addition, a substantial or complete loss in neutralizing activity has also been observed for most commercial therapeutic mAbs, with sotrovimab, bebtelovimab, and the combination of tixagevimab with cilgavimab being the only antibodies still showing activity against Omicron BA.1, albeit with a 5-to 100-fold loss of potency.5][16][17][18][19][20][21] In addition, the emergence of resistance-associated mutations after sotrovimab administration has been recently reported. 22,23he marked reduction in VOCs neutralization by the current mAbs and vaccine-induced sera highlights the importance and the need for the discovery of novel broadly reactive neutralizing mAbs covering different SARS-CoV-2 strains and potential future emerging variants.In addition, although the isolation of mAbs from vaccinated and convalescent patients infected with the WT SARS-CoV-2 virus has been reported, 6,24,25 reports describing the discovery of such antibodies from VOC-infected individuals are limited. 26Here, we describe the isolation of a set of human mAbs isolated from a convalescent unvaccinated COVID-19 patient who experienced a confirmed primary infection with the Gamma variant.The antibodies display a variety of functionalities against the autologous virus but also against heterologous VOCs, including WT, Alpha, Beta, Delta, Omicron BA.1, BA.2, BA.4/5, BQ.1.1,and XBB.1.The specificities were corroborated by structural analyses.In particular, COVA309-35 had very potent neutralizing activity against Omicron BA.1 and BA.2 but less against Delta, while COVA309-22 retained binding and neutralization, although to a lesser extent, against more recent Omicron sub-lineages.Combining COVA309-35 with COVA1-18 and COVA1-16, previously isolated from a Wuhan Hu-1-infected individual, 6 yielded bispecific antibodies (bsAbs) with increased breadth and potency.

Selection of SARS-CoV-2 spike-specific B cells and antibodies from a Gamma-infected individual
To study the B cell response induced by infection with the SARS-CoV-2 Gamma variant and isolate mAbs with novel specificities, we selected samples from an unvaccinated 32-year-old female, COSCA309, who experienced upper respiratory tract infection but did not require hospital admission, following a primary sequence-confirmed Gamma infection.COSCA309 was enrolled in the COSCA study (NL 73281.018.20) that was set up to facilitate investigation of VOC-specific B cell responses and isolation of broad and potent neutralizing mAbs against SARS-CoV-2 and its variants. 6,11e collected blood samples from COSCA309 approximately 40 days after symptom onset.After blood sample collection, peripheral blood mononuclear cells (PBMCs) were isolated and Gamma S-specific B cells were selected by flow cytometry-based single cell sorting using a stabilized Gamma S probe 6 in two colors, in addition to the WT S protein in a third color.The flow cytometry analysis of the donor PBMCs showed a frequency of 0.58% Gamma S-specific B cells (Gamma S-AF647 + , Gamma S-BV421 + ) among the total pool of B cells (CD19 + Via À ), which were shown to be predominantly memory B cells (CD27 + IgD À , 61.5%) (Figure 1A).Within these Gamma S-specific B cells, the majority expressed immunoglobulin G (IgG + ) (57.3%), although a considerable portion of the S-specific B cells were IgM + (34.7%).Co-staining with WT SARS-CoV-2 S protein indicated that 57.7% of Gamma S-specific B cells cross-reacted with the WT strain (Figure 1A).After gene amplification and single-cell cloning, we obtained a total of 45 productive IgG heavy and light chain (HC/LC) pairs.The genetic signatures of the Gamma S-specific B cells were compared to the International Immunogenetics Information System (IMGT) germline repertoire 27 (Table S1).9][30][31][32][33][34][35] Strikingly, none of the mAbs isolated in this study has been found to use the most dominant IGHV3-53 gene segment (Table S1), whereas IGHV1-69 was found to be frequently used by the sorted Gamma S-specific B cells (15.6%), followed by IGHV3-7 and IGHV3-23 (11.1% each) and IGHV4-59 (8.9%) (Figure S1A).The median somatic hypermutations (SHM) was 2%, in line with SHM levels observed after infection with the WT SARS-CoV-2. 6,25,36,37We did not find any substantial difference in the heavy complementary determining region 3 (CDRH3) length of isolated Gamma S-specific B cells (mean 16 amino acids) compared to the average length of 15 amino acids generally present in the human naive repertoire. 38Following expression of the 45 mAbs in human embryonic kidney (HEK) 293T cells and subsequent screening of the supernatants by a flow cytometry-based binding assay against the autologous Gamma S, as well as against Beta and WT S, a total of 14 mAbs were selected based on binding potency and breadth for large-scale protein production and purification (Figure 1B).SARS-CoV-2 Gamma S-specific B cells.Gamma S-specific B cells were selected by double staining of Gamma S labeled with two different fluorescent dyes (Gamma S-AF647, Gamma S-BV421).In addition, Gamma S-positive B cells were stained for IgD, CD27, IgG, and IgM expression.Frequency of WT S cross-reactive B cells is also indicated.SSC-A, side scatter area; FSC-H, forward scatter height; FSC-A, forward scatter area.(B) Flow cytometry-based screening of HEK293T-produced non-purified COVA309 supernatants against Gamma, Beta, and WT S expressed on HEK293T cells, shown in the percentage of binding.293T NT, non-transfected cells.COVA1-16 and CH3I antibodies were included as positive and negative controls, respectively.14 mAbs (in bold) were selected for larger scale expression.
We first assessed COVA309 mAbs binding specificities.Using flow cytometry, we tested the 14 antibodies for binding to HEK293T-transfected cells expressing a large panel of full-length membrane-expressed S, including WT D614G, VOCs (Alpha, Alpha with E484K, Beta, Gamma, Delta, Omicron BA.1, BA.2, BA.4/5, BQ.1.1,and XBB.1), and VOIs (Epsilon, Lambda, and Iota) (Figure 1C).Besides binding to the autologous Gamma S protein, 11 out of the 14 mAbs also recognized the Beta variant, in line with the presence of high similarity between the S mutations in these strains.In general, COVA309 mAbs showed reduced binding to the WT D614G S compared to Beta and Gamma variants, while most of them still recognized the Alpha S.Even though some of the mAbs showed decreased binding to Delta and the Omicron BA.1 and BA.2 variants, almost half of the tested mAbs still recognized the S of Omicron BA.4/5, BQ.1.1,and XBB.1.This may indicate that approximately 30%-40% of the sorted Gamma S-specific B cells were likely to be cross-reactive to more recent Omicron sub-variants.Furthermore, COVA309-35 was unique since it showed the strongest binding to the Omicron BA.1 and BA.2 sub-lineages.In total, 7 mAbs recognized Epsilon, Lambda, and Iota VOIs.In addition to the flow cytometry binding data to full-length S, we performed an enzyme-linked immunosorbent assay (ELISA) with soluble RBD proteins of the WT, Alpha, Beta, Gamma, Delta, Omicron BA.1, BA.2, BA.4/5, BQ.1.1,and XBB.1 variants and showed that 5 COVA309 mAbs are directed toward the RBD, while the others probably target other epitopes on the S, such as the NTD and the S2 subunit (Figure 1D).Among the RBD-targeting mAbs, COVA309-22, -35, and -38 were the broadest, with COVA309-22 binding to all RBDs of the tested variants, while COVA309-03 and -10 seemed to be more strain specific.Overall, these data indicate that following a primary infection with the Gamma variant, broadly reactive mAbs against other viral strains can be elicited.
Next, we assessed the ability of the five COVA309 mAbs to neutralize authentic SARS-CoV-2 viruses, including WT, Alpha, Beta, Gamma, Delta, Omicron BA.1, BA.2, and BA.5 strains (Figures 1F and S1C).COVA309-03 neutralized primary Beta and Gamma viruses with an IC 50 of 1.03 and 0.3 mg/mL, respectively, whereas COVA309-10 only neutralized Beta (IC 50 3.2 mg/mL).COVA309-35 and -38 were still highly effective in neutralizing multiple primary viruses.COVA309-35 neutralized autologous Gamma virus efficiently (IC 50 0.07 mg/mL) but showed a decrease in neutralization against Alpha and Delta viruses (IC 50 9.7 mg/mL and >10 mg/mL, respectively), while maintaining neutralization of Omicron BA.1 and BA.2 variants (IC 50 0.6 and 0.2 mg/mL, respectively).COVA309-38 neutralized primary variants with IC 50 ranging from 0.3 mg/mL to 4.7 mg/mL but was not effective against the Omicron BA.1 and BA.2 strains (IC 50 > 10 mg/mL).None of the tested mAbs showed neutralization against the primary Omicron BA.5 virus with IC 50 values below the cutoff of 10 mg/mL.Compared to the pseudovirus neutralization, the neutralization potency against authentic SARS-CoV-2 strains was generally somewhat reduced, consistent with previous studies, 6,41 but the five neutralizing mAbs were overall similar between primary virus and pseudovirus neutralization assays in terms of activity and patterns of neutralization, which is confirmed by a strong correlation between the two assays (Figure S1D, Spearman correlation, r = 0.8460, p < 0.0001) and is also in line with previous reports. 41,42hen we examined the genetic signatures of the five neutralizing versus non-neutralizing COVA309 mAbs, we found that COVA309-03, -10, -22, and -38 used genes belonging to the IGHV3 family (IGHV3-48, IGHV3-23, IGHV3-33, and IGHV3-7, respectively), while COVA309-35  6,39,40 The cutoff was set at 10 mg/mL (light gray bar) for variants up to Omicron BA.2, and at 50 mg/mL (dark gray bar) for Omicron BA.4/5, BQ.1.1,and XBB.1.Each dot represents the mean value from two or three experiments in which three or four replicates per mAb were performed.(F) Neutralization of authentic SARS-CoV-2 viruses by COVA309 mAbs.The cutoff was set at 10 mg/mL (gray bar).Color code is the same as for the pseudovirus neutralization.COVA1-16 and COVA1-18 mAbs were tested against WT, Alpha, Beta, Gamma, Omicron BA.1, BA.2, and BA.5 to corroborate the pseudovirus neutralization findings.Each dot represents the mean value from one or two experiments performed in duplo.

Structural analysis of COVA309 mAbs explains their ability to cross-neutralize
To study the mechanism of binding of the neutralizing COVA309 mAbs, we performed a biolayer interferometry (BLI) assay measuring competition between COVA309 mAbs and recombinant soluble human ACE-2 receptor for binding to autologous Gamma S (Figure 2A).COVA309-03, -10, and -35 were able to block ACE-2 binding to Gamma S, suggesting that they target an epitope in close vicinity to the ACE-2-binding site on the apical area of the RBD.Alternatively, depending on the angle of approach to the RBD, they may cause steric hindrance with the receptor, therefore precluding ACE-2 interaction with S.These data are consistent with observations that many potent RBD-specific mAbs neutralize the virus by binding to a region that mediates ACE-2 receptor engagement, or by inhibiting the receptor attachment through specific approaching angles. 43,44On the contrary, mAbs COVA309-22 and -38 were unable to block ACE-2 binding, suggesting that they might have different binding sites and mechanisms of action than ACE-2 blocking (Figures 2A and 2C).
Next, to acquire more insight into the binding site specificities of the COVA309 mAbs, we tested their binding by BLI to the WT SARS-CoV-2 S coated on biosensors and determined cross-competition with previously characterized mAbs.The mAbs directed to the RBD can be classified based on their binding epitope, targeting either sub-sites in receptor binding site, (RBS)-A, -B, -C, -D, or the more conserved S309 and CR3022 sites. 45,46We included mAbs belonging to distinct classes, comprising COVA2-14 against S2, COVA1-22 against NTD, COVA2-04 against RBS-A, J08 against RBS-B, COVA2-15 against RBS-D, S309 against the S309 site, and COVA-16 and CR3022 against the CR3022 epitope 6,45-47 (Figures 2B, 2C, and S2A).COVA309-22 and -38 strongly interfered with binding of COVA1-16 and CR3022, indicating that they target the CR3022 site on the lateral face of the RBD, which is highly conserved among variants (Figures 2B and 2C).COVA309-35 competed with COVA2-04 and COVA2-15 (46% and 47% residual binding, respectively), suggesting that its epitope partially overlaps with the regions targeted by the mentioned mAbs, or that the binding of COVA309-35 causes steric hindrance or induces a conformation change of the S which reduces the binding of COVA2-04 and COVA2-15.While we did not observe substantial competition of COVA309-03 with any of the tested mAbs, COVA309-10 showed a partial competition with RBS-B-targeting mAb J08.
We next determined the structural features and antigen recognition of the isolated COVA309 mAbs.We only obtained a crystal structure of the COVA309-22 antibody binding fragment (Fab) in complex with SARS-CoV-2 WT RBD at 3.7 A ˚resolution.In agreement with the BLI data, COVA309-22 interacts with a conserved RBD site that largely overlaps with that of CR3022 but in a different binding mode 46 (Figure 2D).The HC and LC of COVA309-22 and CR3022 bind the RBD with a swapped orientation, suggesting that, although the footprint is the same, the residues involved in the antigen recognition and interaction are different.
The three CDRs of the COVA309-22 HC (CDRH1, H2, and H3) account for the majority of the interactions, burying 31 A ˚2, 377 A ˚2, and 232 A ˚2 of surface area (BSA), respectively, while the CDRL1, CDRL3, and HFR3 contribute to 73 A ˚2, 280 A ˚2, and 64 A ˚2 of the total 1057 A ˚2 BSA.Aromatic residues of COVA309-22 IGHV W52, H58, and IGLV P95 stack with RBD residues Y365, F377, and P384, whereas IGLV Y92 contacts RBD-Y380 and P412.Moreover, IGHV M99 forms hydrophobic interactions with RBD-V382, F392, and T430 (Figure 2E).In accordance with competition experiments, the structure confirmed that the ACE-2 receptor binding site is far from the epitope recognized by this antibody and that COVA309-22 does not interfere with receptor binding (Figure 2F).In parallel, we generated lower-resolution negative-stain electron microscopy (NS-EM) reconstructions of the other COVA309 mAbs.COVA309-03, -10, and -38 were analyzed in complex with 6P-stabilized Gamma S, while COVA309-35 was complexed with 6P-stabilized Omicron BA.1 S protein (Figures 2G and S2B).By using reference models from other mAbs, including CV07-270, CV07-250, C110, and DH1047, 48,49 we could determine the epitope for each mAb (Figure S2C).In accordance with BLI competition data, COVA309-35 Fab in complex with the Omicron 6P S trimer showed that this antibody binds the apical part of the open, up-state RBD, targeting a region which encompasses the RBS-C and RBS-D regions 50,51 (Figure 2G).This corroborates the competition of COVA309-35 with COVA2-15 that binds RBS-D.In addition, RBS-C harbors L452, mutated in the Delta strain, explaining why COVA309-35 does not neutralize Delta.This finding is consistent with the ACE-2 competition data and shows that COVA309-35 resembles the binding characteristics of mAb C110, which was also reported to target a similar epitope 52 (Figure S2C).The NS-EM maps obtained from COVA309-03, -10, and -38 Fabs complexed with the Gamma 6P trimer demonstrated that all the antibodies could be aligned on the RBD in the up conformation, although COVA309-38 was also found to recognize the RBD in a partial down state.COVA309-10 binds RBS-B, on the apical part of the RBD, which includes amino acids at positions 478, 484, and 452, in accordance with our ACE-2 competition results and explains the loss in neutralizing activity against Delta and the Omicron sub-lineages that have different amino acids at these positions compared to Gamma.COVA309-03 targets RBS-C which also includes residues 484 and 452, thereby explaining the narrow neutralizing activity of this antibody.Lastly, NS-EM structures revealed that COVA309-38 binds the CR3022 site, in agreement with the competition to COVA1-16 and CR3022 itself (Figures 2B, 2C, and 2G).
In addition to pseudovirus neutralization data, we also examined the neutralizing activity of the bsAbs against primary, authentic SARS-CoV-2 viruses, including WT, Alpha, Beta, Gamma, Delta, Omicron BA.1, BA.2, and BA.5 (Figure 3B, left panel).COVA309-35 in combination with either COVA1-18 or COVA309-38 appeared to have the highest potency and breadth, showing neutralization of all primary SARS-CoV-2 strains up to Omicron BA.1 and BA.2.COVA309-35 together with COVA1-16 also neutralized all variants except Delta and Omicron BA.5, with a slightly lower potency (IC 50 from 0.2 to 3.4 mg/mL).In accordance with the lack of neutralization capacity of COVA309-38 against Omicron BA.1 and BA.2, we did not observe any activity of the bsAbs involving COVA309-38 against these strains.Moreover, similarly to the pseudovirus neutralization data, the bsAbs did not neutralize Omicron BA.5 with IC 50 values < 10 mg/mL.Neutralization data of antibody cocktails against authentic SARS-CoV-2 viruses are also reported in Figure 3B, right panel.

DISCUSSION
SARS-CoV-2 has been shown to rapidly evolve into antigenically distinct variants, reducing the protective effect of natural infection as well as vaccination.4][15][16][17][18] Therefore, it is important to have potent therapeutics available, especially for the elderly and the immune-compromised.Different therapeutics have seen market introduction including several cocktails and mAbs.The mAbs, mAb cocktails, or bsAbs should ideally be able to prevent viral escape and be broad enough to recognize diverse future SARS-CoV-2 lineages.8][19][20][21] Since we mainly depend on memory B cells for protection against serious disease and future outbreaks, it is important to study the immune repertoire of convalescent individuals to understand the breath and potency of the memory B cell response.In this study, we were able to demonstrate that Omicron and Delta neutralizing mAbs derived from memory B cells were present in a Gammainfected individual.This finding is in contrast with the relatively poor cross-reactive serum response observed for COSCA309, 11 suggesting that the memory B cell response might be more versatile and therefore more adequately prepared to respond to future variants and be able to protect against severe disease.
The Gamma-elicited COVA309 mAbs differ substantially when comparing VOC binding, pseudoviru,s and primary virus neutralization.For example, COVA309-35 and -38 mAbs appear to have distinct S binding and neutralizing patterns against Omicron and Delta variants.Together, these mAbs can cover a broad range of antigenic variation within current and possible future SARS-CoV-2 strains.Therefore, we tested COVA309-35 and -38 together with COVA1-18 and COVA1-16 mAbs, 6 either as bsAbs or as cocktails in a pseudovirus neutralization assay (Figure 3A).For bsAbs, lack of bivalent binding can be a disadvantage since the two different Fabs then should cooperate in binding to different epitopes on the same or adjacent S or S protomer.Although loss of bivalent binding can be observed for bsAbs compared to the combination therapy, the difference is small and, in general, both approaches provide broader antigenic coverage compared to the corresponding mAbs, especially the combination of COVA309-35 and COVA1-18.As a bsAb, this combination neutralizes all pseudovirus variants up to Omicron BA.2 with IC 50 between 0.01 and 0.3 mg/mL, with the latter value being for BA.1, while as a cocktail, the range is from 0.001 to 0.2 mg/mL, again the least against BA.1.
In line with previous studies, 51,52 the binding epitopes of the five neutralizing COVA309 mAbs are mainly located in the immunodominant RBD of the S (Figures 2C-2G).More specifically, we show that our SARS-CoV-2 mAbs target distinct RBD sites, within and outside the RBS.The differences in neutralization of the VOCs by the COVA309 mAbs could be explained by the structural analysis and binding studies.COVA309-35, which showed the highest potency against the autologous Gamma strain, as well as against Omicron BA.1 and BA.2, interferes with ACE-2 receptor binding and recognizes an epitope covering RBS-C and RBS-D when the RBD is in the up state, thereby being highly effective against the infectious S conformation.COVA309-03 was also found to bind RBS-C, but its binding specificity appears to be narrower as it neutralizes only the Beta and Gamma variants.This indicates that its binding strongly depends on 484K, which was confirmed by recognition of an Alpha strain that we engineered to contain the E484K mutation (Figure 1C).
In addition to mAbs targeting the RBS, COVA309-38 was found to target a more conserved cryptic site at the base and lateral face of the RBD.mAbs directed at this RBD region are generally broadly reactive but weakly neutralizing, although few exceptions exist like COVA1-16 6,39 and ADI-62113. 53COVA1-16 exhibits a special angle of approach and is able to directly compete with the receptor binding through steric hindrance, thereby neutralizing SARS-CoV-2 much more potently.As confirmed by BLI experiments and NS-EM, we showed that COVA309-38, in addition to COVA1-16, competes with CR3022 and COVA309-22 (Figures 2B, 2C, and S2D).In addition, ACE-2 competition data revealed that, despite targeting the base and lateral face of the RBD, this antibody is still able to partially compete with the receptor binding to the S, thereby in part resembling the mechanism of action of COVA1-16.The crystal structure of COVA309-22 showed that the antibody targets the same epitope as COVA309-38, explaining the breadth of the antibody, albeit with a lower potency compared to mAbs targeting the RBS. 51The epitope of COVA309-22 also extensively overlaps with that of previously described mAbs, including CR3022 46 and EY6A, 54 indicating once again that this RBD region represents a key site of vulnerability on the S protein.
Here, we demonstrate that potent and broadly reactive Gamma-elicited antibodies can be generated and can still neutralize other variants, including the highly distant Omicron sub-lineages.Despite the lack of neutralization of all variants tested by a single COVA309 mAb, the overall humoral response of the COSCA309 patient, represented by the sum of the isolated antibodies targeting diverse viral epitopes, appears to be very broad and would be expected to confer a benefit in the protection against more recent SARS-CoV-2 lineages.We also show that, when tested in bsAb and cocktail combinations with other mAbs, all antibodies show enhanced neutralization breadth and potency, suggesting that combining antibodies with different RBD epitopes and mechanisms of action can provide a better resistance to viral mutants and therefore lead to more effective therapeutics.In addition, it is important to study how SARS-CoV-2 evolution shapes and alters the antibody response compared to the ancestral strain.In this way, we can increase our knowledge on the specific features and signatures of broadly reactive mAbs that cover diverse viral strains and may aid in development of next-generation vaccines and mAb therapeutics.gBlock gene fragments (Integrated DNA Technologies) of the different constructs were ordered and cloned into a pPPI4 avidin-tagged and/ or hexahistidine-tagged vector by PstI-BamHI digestion and ligation with Gibson Assembly (Thermo Fisher Scientific).The recombinant human ACE-2 receptor was obtained in the same way after ordering the corresponding gBlock gene fragment (Integrated DNA Technologies).Sanger sequencing was used to verify the constructs.Proteins were then produced in human embryonic kidney (HEK)293F cells (Thermo Fisher Scientific) maintained in Freestyle medium (Life Technologies).Briefly, HEK293F cells were transfected with a 1:3 ratio of expression DNA plasmids (312.5 mg/l) and Polyethylenimine Hydrochloride (PEI)max (1 mg/ml) in OptiMEM.Supernatants containing the produced proteins were harvested six days after transfection, centrifuged at 4000 rpm for 30 min and filtered using 0.22 mM Steritop filter units (Merck Millipore).Affinity chromatography with Ni-NTA agarose beads (Qiagen) was used to purify the proteins.Eluates were subsequently concentrated and buffer exchanged to phosphate-buffered saline (PBS) or TN75 buffer (75 mM NaCl and 20 mM Tris HCl, pH 8.0) using VivaSpin20 filters (Sartorius).Protein concentrations were determined by the Nanodrop.
After single cell sorting, the plates were stored at -80 C for at least 1 h before a reverse transcriptase (RT)-PCR was performed to convert the mRNA of the lysed B cells into cDNA, as described previously. 6Analysis was performed using FlowJo X software (BD Biosciences).

Single-cell immunoglobulin gene amplification, cloning and antibodies expression
After RT-PCR, additional PCR rounds were performed to amplify the V(D)J variable region of the HC and LC of the antibodies, as described previously. 6PCR products were then cloned into corresponding human IgG1 expression vectors with Gibson Assembly (Thermo Fisher Scientific) and the mixture was subsequently transformed into DH5a cells.After DNA purification, the sequences were verified by Sanger sequencing.For small scale expression of the antibodies, adherent HEK293T cells (ATCC, CRL-11268) were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% foetal calf serum and a mixture of penicillin/streptomycin (100 U/ml and 100 mg/ml, respectively).HEK293T cells were seeded in 24-well plates at a density of 2.75x10 5 cells/well 24 h prior to transfection.The transfection was performed with a 1:1 (w:w) HC/LC ratio using a 1:2.5 ratio with 1 mg/l PEImax (Polysciences) in 200 ml Opti-MEM.The transfection mix was incubated for 15 min at RT and then added onto the cells.Supernatants of transfected cells were harvested 48 h post-transfection and tested for preliminary screening by flow cytometry.

Larger expression of antibodies in HEK293F cells
Suspension HEK293F cells (Invitrogen) were maintained in FreeStyle medium (Gibco) and co-transfected with a 1:3 ratio of the two HC/LC DNA plasmids and 1 mg/l PEImax (Polysciences).Five days post-transfection, the cell suspension was centrifuged at 4000 rpm for 30 min, followed by filtration of the supernatant using 0.22 mm pore size SteriTop filters (Millipore).The filtered supernatant containing the recombinant IgG antibodies was run over a 10 ml protein G column (Pierce) and the antibodies were then eluted with 0.1 M glycine pH 2.5, into the neutralization buffer (1 M TRIS pH 8.7).50 kDa VivaSpin20 columns (Sartorius) were used to concentrate and buffer exchange the antibodies to PBS.The IgG concentration was determined by the NanoDrop.

Flow cytometry-based screening of supernatants and purified recombinant mAbs
Cell surface-expressed SARS-CoV-2 ancestral and variant S were obtained by transfecting 8 mg of SARS-CoV-2 full-length DNA plasmid with 25 ml PEImax in 400 ml OptiMEM onto 12 to 15 ml HEK293T cells in a petri dish (seeded the day before at a density of 3.0x10 6 ).After 48 h, cells were harvested and frozen until further use.After thawing, HEK293T cells expressing the S of interest were seeded at 20.000 to 30.000 cells per 96-well in FACS buffer (PBS/0.5% FCS) and were incubated 1:1 with non-purified supernatants from small scale IgG production or a dilution of purified HEK293F-produced mAbs for 30 min at 4 C. Subsequently, cells were washed twice with FACS buffer and incubated for 30 min at 4 C with 1:1000 diluted PE-conjugated goat F(ab)'2 anti-human IgG (Southern Biotech 2042-09) in the dark.Cells were washed and analysed on the FACS Canto II analyser.Samples were analysed by FlowJo X software (BD Biosciences) and percentage of cells that showed binding were plotted correspondingly.

Figure 1 .
Figure1.Selection of B cells and antibodies from a SARS-CoV-2 Gamma-infected individual (A) Sorting strategy of singlet viable CD19 + SARS-CoV-2 Gamma S-specific B cells.Gamma S-specific B cells were selected by double staining of Gamma S labeled with two different fluorescent dyes (Gamma S-AF647, Gamma S-BV421).In addition, Gamma S-positive B cells were stained for IgD, CD27, IgG, and IgM expression.Frequency of WT S cross-reactive B cells is also indicated.SSC-A, side scatter area; FSC-H, forward scatter height; FSC-A, forward scatter area.(B) Flow cytometry-based screening of HEK293T-produced non-purified COVA309 supernatants against Gamma, Beta, and WT S expressed on HEK293T cells, shown in the percentage of binding.293T NT, non-transfected cells.COVA1-16 and CH3I antibodies were included as positive and negative controls, respectively.14 mAbs (in bold) were selected for larger scale expression.

Figure 1 .
Figure 1.Continued (C) Heatmap showing the percentage of mean fluorescence intensity (MFI) of HEK293F-produced and purified COVA309 mAbs binding to SARS-CoV-2 variant S expressed on HEK293T cells, as assessed by flow cytometry.VOCs, variants of concern; VOIs, variants of interest.COVA1-16 and D25 (an RSV F specific mAb) are used as positive and negative controls, respectively.(D) Heatmap depicting the binding of COVA309 mAbs to SARS-CoV-2 variant RBDs, as determined by ELISA.Color scale indicates the area under the curve (AUC) for each mAb.(E) Half maximal inhibitory concentrations (IC 50 ) of SARS-CoV-2 VOC pseudoviruses neutralization for COVA309 mAbs, COVA1-16, and COVA1-18.6,39,40The cutoff was set at 10 mg/mL (light gray bar) for variants up to Omicron BA.2, and at 50 mg/mL (dark gray bar) for Omicron BA.4/5, BQ.1.1,and XBB.1.Each dot represents the mean value from two or three experiments in which three or four replicates per mAb were performed.(F) Neutralization of authentic SARS-CoV-2 viruses by COVA309 mAbs.The cutoff was set at 10 mg/mL (gray bar).Color code is the same as for the pseudovirus neutralization.COVA1-16 and COVA1-18 mAbs were tested against WT, Alpha, Beta, Gamma, Omicron BA.1, BA.2, and BA.5 to corroborate the pseudovirus neutralization findings.Each dot represents the mean value from one or two experiments performed in duplo.

Figure 2 .
Figure 2. Continued (D) Crystal structure of COVA309-22 Fab in complex with SARS-CoV-2 WT RBD at a 3.7 A ˚resolution (left).The HC and LC are colored in orange and yellow, respectively.CR3022 Fab (right) is reported as a comparison (PDB: 6W41).(E) Detailed representation of the main residues involved in the COVA309-22-WT RBD interaction.(F) COVA309-22 HC (orange) and LC (yellow) bind the base and lateral face of the RBD, far from the ACE-2 receptor binding site (green; PDB: 6M0J).(G) Front view (top row) and top view (bottom row) of low-resolution NS-EM reconstructions of COVA309-03, -10, -35, and -38 in complex with either Omicron 6P S or Gamma 6P S. Composite NS-EM maps of COVA309 Fabs indicate that all COVA309 mAbs align on one RBD in the up conformation.

Figure 3 .
Figure 3. Characterization of bsAbs and antibody cocktails (A) IC 50 values for pseudovirus neutralization of SARS-CoV-2 VOCs by bsAbs (left) and corresponding antibody cocktails (right).For antibody cocktails, a 1:1 ratio was used.The cutoff was set at 10 mg/mL (gray bar).Each dot represents the mean value from two or three experiments in which three or four replicates per mAb were performed.(B) Neutralization of authentic SARS-CoV-2 viruses by bsAbs and antibody cocktails.The cutoff was set at 10 mg/mL (gray bar).Each dot represents the mean value from one or two experiments performed in duplo.