Varying neutralization of Spike by antibodies
We performed biophysical characterization of nine human monoclonal IgG antibodies (‘HuMAbs’), namely LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016, LSI-CoVA-017, 4A833, 5A634, CR302235, CoVA-02, and CoVA-3936. These monoclonal antibodies were principally discovered from convalescent patients of COVID-19 with the exception of 5A6 which was derived from a naïve human phage-FAB library. The binding activity of each antibody to SARS-CoV-2 Spike trimer Wuhan strain (‘Spike’) and isolated RBD (‘RBD’) was determined using Quartz Crystal Microbalance (QCM) and enzyme-linked immunosorbent assay (ELISA). As observed by the half-maximal effective concentration (EC50) values, seven of the nine antibodies bound strongly to both Spike trimer and RBD (Fig. 1b). Antibody LSI-CoVA-017 binds strongly to Spike but not RBD, suggestive of an epitope outside RBD. Antibody 4A8 binds weakly to Spike and showed negligible binding to RBD. Next, we determined the binding kinetics of these HuMAbs against Spike and observed high affinity binding with slow off-rates (Extended data Fig. 1a, Table S1). The affinity constants (KD) were in sub-nM range, with LSI-CoVA-017 being the lowest (0.088 nM). We next investigated their neutralization efficacies using a pseudotyped virus neutralization test (PVNT). A neutralization capacity of >50% was considered significant as reference to WHO standards. Correspondingly, we observed differential levels of neutralization, wherein LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016, and CR3022 showed less than 50% efficacy. On the other hand, 4A8, LSI-CoVA-017, and CoVA2-04 showed significant neutralization capacities, while the highest neutralization was observed for CoVA2-39 and 5A6. On this basis, the antibodies were classified as (i) weak, (ii) moderate, and (iii) strong neutralizing HuMAbs (Fig. 1c).
Cryptic site binding IgGs destabilize Spike
The epitopes of the HuMAbs were mapped by comparative hydrogen-deuterium exchange mass spectrometry (HDXMS) analysis of complexes with Spike and RBD. We observed extensive protection against deuterium exchange across peptides spanning RBD of Spike and isolated RBD, in the presence of LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016. This indicates binding to either receptor binding motif (RBM) or at a site distal to RBM (Fig. 1d, 1e, Extended data Fig. 1b). Overlapping peptides covering residues 361-395 showed large-scale protection against deuterium exchange in both Spike (Fig. 1d) and RBD complexes with LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 (Extended data Fig. 1b), indicating that these three antibodies bind RBD at a site distal to the RBM site. In the trimeric Spike, the region spanning residues 361-395 becomes accessible only when the RBD adopts an up-position. These changes indicate that the epitope sites identified for LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 are similar to the site observed for the CR3022 antibody (Extended data Fig. 1c), previously characterized as a “cryptic” site binder37. Interestingly, increased deuterium exchange was observed across the peptides spanning the RBM/ACE2 binding site upon binding of either of these three antibodies (Fig. 1d, Extended data Fig. 1d). This correlates to increased conformational dynamics at the ACE2 binding site, suggesting that binding of LSI-CoVA-014, LSI-CoVA-015, or LSI-CoVA-016 induces allosteric destabilization at RBM (Fig. 1e).
Similar effects were observed across other regions of Spikeupon binding of LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 (Extended data Fig. 2a). Notably, peptides spanning residues 516-533 showed increased deuterium exchange (Fig. 2a i), confirming that antibody-binding stabilized RBD in an ‘up’-conformation. This is accompanied by the loss of inter- and intra- monomer contacts between RBD and NTD, with residues 166-182 which interact with RBD, and 289-305 that connect NTD to the central Spike core showing the most significant changes (Fig. 2ai). Further, multiple regions of the S2 subunit including the fusion peptide (FP), heptad repeats (HR1 and HR2) and residues 902-916 also showed increased deuterium exchange in the presence of these three HuMAbs (Fig. 2aii, Extended data Fig. 2a), except the S1/S2 cleavage site which was associated with a decreased deuterium exchange. These sites are essential for inter-monomer interactions. Taken together, the conformational changes observed at the NTD and the S2 subunit suggest antibody-binding at RBD induces allosteric changes across the Spike trimer, resulting in its global destabilization that may lead to dissociation of adjacent monomers.
These observations were further supported by the HDX changes at the paratope sites of LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 (Extended data Fig. 2b, 2c). Peptides spanning CDRL1-3 and CDRH1-3 showed greater differences in deuterium exchange in the RBD-bound complex, as compared to Spike-bound antibody complexes (Table S2). The epitope sites are not hidden in the isolated RBD construct, but are readily accessible to stably bind the paratope sites. On the other hand, in the Spike trimer, the RBDs must move from a down- to an up- position, and antibody binding is further hindered spatially by the NTD and the S2 subunit, leading to less stable antibody binding.
NTD-binding LSI-CoVA-017 stabilizes Spike trimer
We next investigated the effects of LSI-CoVA-017, which shows moderate neutralization. In the LSI-CoVA-017-bound state, protection against deuterium exchange was observed across the Spike trimer, with only a few peptides showing deprotection (Fig. 2B, Extended data Fig. 3a). Upon closer examination, peptides spanning residues 92-110, 136-143 (N3 loop), and 243-265 (N5 loop) showed large-scale decreased deuterium exchange in the LSI-CoVA-017-bound state (Fig 2b). These peptides are positioned towards the outer edge of NTD, and are likely the epitope sites binding to LSI-CoVA-017. This region also corresponds to the NTD antigenic supersite 20,38. Reduction in deuterium exchange at short labeling times was observed across residues 36-48, 166-182, and 303-318, while increased deuterium exchange was observed for residues 60-83, 107-117, 213-228, and 266-276. These differences, mapped onto the structure of NTD, revealed that peptides encompassing the epitope site are clustered closely to form a structural epitope and facilitate complexation of Spikewith LSI-CoVA-017. These peptides also showed reduction in deuterium exchange in our comparative HDX analysis of free Spike and its complex with 4A8, which has been previously characterized as an NTD-binding antibody 39 (Extended data). Thus, our binding assays and HDX data identify LSI-CoVA-017 as an NTD-binding antibody, with both LSI-CoVA-017 and 4A8 being moderate neutralizers.
Large-magnitude decreases in deuterium exchange were observed across all peptides (including 320-350, 516-533) spanning the RBD of Spike bound to LSI-CoVA-017 (Extended data Fig. 3a, Table S3).This indicates significantly reduced conformational dynamics across RBD, suggesting restricted domain motions in the LSI-CoVA-017-bound state. HDXMS analysis of LSI-CoVA-017 and 4A8 with isolated RBD, revealed no significant changes in deuteration levels of RBD. Hence, it is clear that the antibodies binding at NTD induce a distinct allosteric effect across RBD compared to RBD-binding antibodies. Furthermore, allosteric effects were propagated to the S2 subunit as well. Decreased deuterium exchange was observed for peptides spanning the S2 subunit of the Spike-LSI-CoVA-017 complex (Fig. 2b, Extended data Fig. 3b). Upon LSI-CoVA-017 binding, notable changes in conformational dynamics were observed at the S1/S2 cleavage site, FP, central helix, and HR (Fig. 2c ii, inset). While both LSI-CoVA-017 and 4A8 binding resulted in similar effects on the Spike trimer, the changes induced by 4A8 HuMAb were less prominent. Overall, these HDXMS results reveal that LSI-CoVA-017 binding at NTD induces global stabilization of the Spike trimer.
HDXMS analysis of the LSI-CoVA-017 antibody showed significant changes across both heavy and light chains in the presence of Spike (Table S2). Peptides overlapping CDRH2 (residues 48-70), CDRH3 (96-103), and CDRL2 (48-71) showed protection against deuterium exchange, while CDRL3 (101-129) showed increased deuterium exchange. Interestingly, similar changes were observed for 4A8 complexed to Spike. No significant changes were observed for the light chain of 4A8, consistent with available high-resolution structures 33.
The commonalities in effects of 4A8 and LSI-CoVA-017 upon Spike suggest similar modes of neutralization, as reflected in their neutralization capacities. However, our biophysical data showed LSI-CoVA-017 binds Spike trimer with an affinity much greater than 4A8 (Fig. 1). To rationalize this, we determined the stoichiometry of the Spike-LSI-CoVA-017 complex by size-exclusion chromatography (Extended data Fig. 3c, Table S3). Densitometry analysis of different amounts of peak B suggested a binding stoichiometry of three LSI-CoVA-017 antibodies per Spike trimer. With a 1:3 Spike:IgG stoichiometry, two models are plausible where: (i) Fab arms from three LSI-CoVA-017 antibodies bind to three monomers of Spike; or (ii) two Fab arms of the same LSI-CoVA-017 bind monomers of two different Spike trimers. We further probed this by molecular docking, as discussed in later sections.
RBM-binding strong neutralizers stabilize Spike
Multiple studies have reported high-resolution structures of HuMAbs bound to RBM, including CoVA2-04, 5A6, and CoVA2-39, showing direct competition with ACE2 binding 40. However, given that these HuMAbs display various virus-entry neutralization potencies whilst binding to overlapping epitopes, a mechanistic explanation for their contrasting behavior remains elusive, particularly for CoVA2-04, a weak neutralizer, as opposed to 5A6 and CoVA2-39 that are strong neutralizers. We therefore monitored the binding of CoVA2-04, 5A6, and CoVA2-39 to the Spike trimer and observed a distinct impact on its conformational dynamics (Fig. 2c, Extended data Fig. 4). A large-magnitude decrease in deuterium exchange was observed across RBD, particularly the peptide clusters spanning RBM (485-502) of Spike and RBDcomplexes with 5A6, CoVA2-04 and CoVA2-39 (Fig. 2, Extended data Fig. 2d, 4a). Interestingly, HDXMS analysis of CoVA2-04 and CoVA2-39 complexes with the isolated RBD construct showed lower deuterium exchange across RBM, and only minor changes at other regions. These results indicate that binding of HuMAbs at RBM induces localized changes that lead to a significant reduction in the structural dynamics of RBD, including the peptides spanning the base and linker regions that connect RBD to the Spike trimer. Notably, the Spike VoCs contain mutations at different sites including E484K, N501Y- or K417N/E484K/N501 that are localized at RBM, and are reported to reduce the neutralization efficacy of antibodies 1,26,41.
Binding of CoVA2-04, 5A6, and CoVA2-39 to Spikeresulted in increased deuterium exchange across peptide clusters covering residues 31-42, 92-110, 177-191, 265-276 of NTD (Fig. 2c, Extended data Fig. 4a). Some of these peptides span the interface interacting with the RBD and the C-terminal region of NTD. As binding of these HuMAbs leads to RBD domain movement, it induces NTD movement as well, disrupting their interaction. Significant protection against deuterium exchange was observed at the S1/S2 cleavage site (672-695), regions flanking the FP (residues 770-782, 878-898), HR1 (927-962), central helix (1003-1031), and 1103-1117 of the S2 subunit (Fig. 2c, Extended data Fig. 4b). These sites are essential for the Spike trimer to transition from its pre-fusion state to post-fusion state. Decreased deuterium exchange across the S2 subunit suggests that binding of these strong neutralizing HuMAbs leads to global reduction in the conformational dynamics of the Spike trimer, which may prevent the transition to the fusogenic intermediate. Collectively, the HDXMS results provide detailed insights into the mechanism of action of CoVA2-04, 5A6, and CoVA2-39, wherein they compete with ACE2 binding and induce stabilization throughout the Spike trimer to restrict its mobility.
Glycan-Fab interaction stabilizes Spike-antibody complexes
Next, we performed molecular docking to model the Fab domains of LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016, and LSI-CoVA-017 at their respective epitope sites of the Spike protein (RBD or NTD) using HDXMS footprints as restraints (Extended data Fig. 5, 6, Table S4), followed by atomistic molecular dynamics (MD) simulations. Out of multiple Fab:RBD/NTD complexes, five top-scoring binding poses were selected for 200 ns simulations. Among the simulated models of Fab:RBD/NTD complexes, multiple models were observed to either displace from the epitope site (Model 3, RBD-014) or completely detach from RBD (Model 4, RBD-014 and Model 3 of RBD-016) (Extended data Fig. 5a), and were not considered for further analysis. To select the best models from the stable complexes with each antibody, we next calculated the root mean square deviation (RMSD) of backbone atoms of the Fab domain (Extended data Fig. 5b,c) and the model with the lowest RMSD was selected for additional replicate simulations to improve the conformational sampling (Table S4). A stable Fab binding orientation from the most populated cluster was identified via cluster analysis, of each Fab:RBD/NTD complex (Extended data Fig. 5d). This identified 34, 79, 52 and 65 clusters sampled for LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016 and LSI-CoVA-017, respectively (Extended data Fig. 6). Further, contact frequencies calculated between the glycan moieties and Fab from these triplicate MD trajectories revealed that N-glycans interact with the residues across Fab arms of LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016 and LSI-CoVA-017 (Extended data Fig. 6). Interestingly, the glycan moieties at N343 of RBD and N74, N122 and N149 on NTD were observed to interact with the Fab (Extended data Fig. 7a, b). For the residues interacting with the N-glycans at N331 and N343, contact frequency maps showed the highest number and magnitude of contact frequencies made by LSI-CoVA-014 Fab, as compared to those by LSI-CoVA-015 and LSI-CoVA-016. Similarly, simulation trajectories of the NTD-LSI-CoVA-017 complex showed prominent interactions between the N74, N122 and N149 N-glycans and the Fab. The contact frequency measured for the NTD-LSI-CoVA-017 Fab complex indicated stable interactions with a larger surface compared to any of the RBD-Fab complexes,suggesting a potential role for glycans in forming the antibody epitope.
To verify this, we tested the binding of four novel HuMAbs (“LSI-CoVA”) with a deglycosylated Spike trimer. Binding of LSI-CoVA-017 was completely abolished with deglycosylated Spike, in contrast to the minor changes observed for LSI-CoVA-014, LSI-CoVA-015 and LSI-CoVA-016 (Extended data Fig. 7c). Furthermore, significant reduction in binding kinetics of these four HuMAbs with deglycosylated Spike was observed, as compared to the glycosylated Spike trimer (Extended data Fig. 7d). To further validate the significance of glycosylation to RBD-binding HuMAbs, we tested 5A6 as a control, which showed a partial reduction. Collectively, these results demonstrate that the LSI-CoVA-017 epitope encompasses glycan moieties on the Spike protein surface. For other antibodies (LSI-CoVA-014/LSI-CoVA-015/LSI-CoVA-016/5A6) the primary binding sites were non-glycosylated epitopes, as identified by HDXMS, with only secondary interactions contributed by glycans. These results provide a view contrary to the prevailing notion that glycans only act as a shield for Spike protein to hide epitope sites from host immune recognition42,43 and suggest that non-specific interactions of glycans with the antibodies play a substantial role in stabilizing Fab arm binding.
Capture ELISA to access combinatorial therapy
Capture ELISA was performed to evaluate the extent of epitope site overlap among antibodies (Fig. 2d) and also to characterize the cooperative binding to Spike monomers in the trimeric Spike. This allowed us to distinguish the mechanisms of binding and neutralization of RBD-specific antibodies that share the same or highly overlapping epitopes, yet have different affinities and neutralization activities. Competitive binding ELISA results indicated similar OD450 values between LSI-CoVA-015 and LSI-CoVA-016 as detection or capture antibodies, suggesting a significant overlap in their binding orientation, which is in-line with our HDXMS-constrained docking and MD simulations (Extended data Fig. 7e, f). On the other hand, LSI-CoVA-014 did not prevent binding of LSI-CoVA-015 or LSI-CoVA-016. Our simulation cluster analysis trajectories showed that LSI-CoVA-014 bound to Spike in a different orientation than LSI-CoVA-015 or LSI-CoVA-016, and thus can pair with either LSI-CoVA-015 or LSI-CoVA-016 (Fig. 2d, Extended data Fig. 7g). Competition between RBD- and NTD- recognizing antibodies showed that the binding sites for these two antibody classes do not overlap with each other, as observed for LSI-CoVA-014 and LSI-CoVA-017 with LSI-CoVA-015/LSI-CoVA-016 antibodies.
To infer stoichiometry and plausible mechanisms of neutralization, we then modelled the binding of full-length IgGs to Spike using a representative structure of Fab:RBD/NTD from the cluster analysis described above (Extended data Fig. 7g). Modelled full length antibodies showed that RBD-binding antibodies specifically bind to RBD in the ‘up’-position. Full-length LSI-CoVA-015 and LSI-CoVA-016 bound to Spike trimer showed that IgG binding to a single RBD of a Spike monomer sterically hinders the binding of a second IgG to the same Spike trimer. In the case of LSI-CoVA-014 and LSI-CoVA-017, the predicted orientation allows the respective full-length antibody to bind all three RBDs or NTDs of the same Spike protein trimer. These results are consistent with our competitive ELISA and neutralization assays. Additionally, the second Fabarm of LSI-CoVA-017 and LSI-CoVA-014 can bind to a second Spike protein trimer, crosslinking two Spike trimers (Extended data Fig. 7g vi, vii). Taken together, the Spike-IgG complex models suggest that the novel antibodies characterized in this study indirectly interfere with ACE2 binding by either crosslinking Spike trimers on the viral surface (LSI-CoVA-014 and LSI-CoVA-017), or by blocking RBD-ACE2 interaction on a single Spike trimer (LSI-CoVA-015 and LSI-CoVA-016).
Overall, our study has identified four novel antibodies isolated from convalescent patients with suboptimal levels of neutralisation efficacy compared to RBM binding antibodies. Considering the mutually exclusive epitope sites complemented by high affinity binding to Spike protein, it would be of interest to explore the use of antibody cocktails. We explored the possibility to induce destabilisation in individual monomers or stabilization to reduce the hinge dynamics in order to effectively neutralise the SARS-CoV-2. Synergistic effects of selected HuMAbs used in this study were thus evaluated. The selected HuMAbs were used in a pairwise cocktail to study the potential synergistic enhancement of neutralization efficacy, amongst which the paired MAb cocktail of LSI-CoVA-017 and CoVA2-04 displayed a significantly higher percentage neutralization in comparison to the treatment of either single HuMAb (Fig. 2e). We did not observe any enhancement in the neutralization efficacies of the two potent HuMAbs (CoVA2-39, 5A6) with NTD-binding LSI-CoVA-017. Surprisingly, a combination of NTD- (LSI-CoVA-017) with RBD- (LSI-CoVA-014) antibodies resulted in lower neutralization, than when added alone.
Cryptic site binding antibodies explain immune escape by Spike variants
Emergence of new variants as a result of mutations of the Spike protein, have led to many antibody-mediated therapies faltering. Therefore, we assessed the impact of defined VOC-linked mutations on binding and neutralization of the novel antibodies characterized in this study with the RBD and Spike proteins of the two dominant VoCs - Delta (δ) and Omicron (o1 for BA.1 and o2 for BA.2 lineages). LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 bind to Spike and RBD of all strains tested, although the binding activities to Omicron variants were slightly lower (Fig. 3ai), akin to CR3022 (Extended data Fig. 8a). Binding of these antibodies was preserved as the mutations among the variants are distant from the cryptic epitopes site. Interestingly, 4A8 bound to Spike of all strains tested, although the binding activity was drastically reduced among all variants compared to Wuhan-Hu-1 (Extended data Fig. 8). On the other hand, the other NTD-binding antibody LSI-CoVA-017 bound only to SpikeWuhan (Fig. 3aii). This lack of binding of LSI-CoVA-017 and 4A8 to Spikeδ or Spikeo is due to the deletions and mutations spanning the NTD antigenic supersite. Most importantly, 5A6, CoVA2-04, and CoVA2-39 which strongly bind and neutralize SpikeWuhan, bound only to Spikeδ and RBDδ, but not the Omicron variants (Fig. 3aiii). We also observed that upon deglycosylation, the binding activity of LSI-CoVA-017 and 4A8 was lost for all Spike variants, while minimal reduction was detected for anti-RBD antibodies against the Spike variants (Extended data Fig. 8c).
Many studies have reported higher binding affinities of Spike variants with the ACE2 receptor44,45. We therefore set out to probe if any of the nine HuMAbs competed with ACE2 binding to Spike variants. We performed ACE2-binding inhibition assays and observed a lack of any inhibitory activity by the cryptic site binders or the NTD-binding antibodies (Fig. 3b, Extended data Fig. 8). Interestingly, LSI-CoVA-015 and LSI-CoVA-016 seemed to enhance ACE2 interaction with RBDo1, as indicated by the negative inhibition. These antibodies bind at the cryptic site, and maintain RBD in an up-position, making the RBM site accessible, which may lead to increased ACE2 binding. This is similar to the effects of antibodies (e.g. S309) recognizing epitopes outside the RBM locus, and show some efficacy against the Omicron variant.46,47 Also, 5A6, CoVA2-04, and CoVA2-39 inhibited interactions between ACE2 and Spike Wuhan-Hu-1/Delta strains, but this was not the case for in Spikeo1 or RBDo1 (Fig. 3bii). For ACE2 inhibition assays, neutralizers that bind RBD often exhibit >40% inhibition. Therefore, the binding and ACE2-inhibition results suggest that only the cryptic-site binding antibodies retain binding to Spikeδ and Spikeo2, and hence only their interactions were further explored.
We examined the effects of ACE2 binding on the conformational dynamics of the RBD and Spike variants by HDXMS, and compared this with ACE2-binding footprints previously reported5,44,48. Binding of ACE2 elicited large-scale protection against deuterium uptake across all regions of isolated RBDδ, RBDo1 and RBDo2 (Fig. 3c, d). This altered the conformational dynamics of RBD variants upon ACE2 binding, and is reflective of their higher binding affinities44,48. Despite this, mutations of key residues of RBM disrupted specific contacts between the RBD and ACE2, as reflected by available cryo-electron microscopy structures45. Upon closer examination of the HDX results, the ACE2 binding footprints were smaller for variants of RBD, as compared to RBDWuhan. Peptides spanning the mutation sites of the loop region (475-495) showed a lower degree of deuterium exchange, while residues 445-455 and 493-510 showed greater protection (Fig. 3c). We further determined the effects of ACE2 using trimeric Spikeδ and Spikeo2 (Extended data Fig. 9). Binding of ACE2 elicited conformational changes across RBD of Spikeδ akin to those of isolated RBDδ, as well as RBD of SpikeWuhan. Surprisingly, we observed marked differences between the ACE2-bound states of RBDo2 and Spikeo2. While domain-wide decreased deuterium exchange was observed for the RBDo2-ACE2 complex, for RBD of the Spikeo2-ACE2 complex decreased deuterium exchange was observed only at residues 390-417 and 450-467, and at 1 min labeling time for residues 488-507,as observed in high-resolution structures. Peptides spanning 373-384, 429-446, and 468-483 showed significantly increased deuterium exchange at all labeling time points, and residues 488-507 showed increased deuterium exchange at longer labeling times. Peptides showing deprotection overlapped the RBD-specific mutation sites observed for the Omicron variant, while increased protection against HDX was observed for Wuhan and Delta variants. These results indicate a molecular mechanism of ACE2-binding by the Omicron variant, whereby the specific amino acid residues promote receptor binding by maintaining the overall spatial conformation, yet evade immune responses. Furthermore, these HDX findings also explain how ACE2 binding enhances the conformational sampling of VoCs, as observed by their high flexibility and fuzzy densities in cryo-EM maps44,45.
Cross-reactivity with Delta and Omicron VoCs
Using HDXMS, we also mapped and characterized the interactions of LSI-CoVA-014, LSI-CoVA-015, LSI-CoVA-016 antibodies with Spikeδ and Spikeo2 variants, which showed similar deuterium exchange profiles across the S1 (Extended data Fig. 10a) and the S2 (Extended data Fig. 10b) subunits. Peptides flanking the mutation sites of NTD showed no significant change in deuteration as compared to SpikeWuhan, which exhibited lower deuterium exchange upon binding to these three antibodies (Fig. 2, Extended data). Importantly, the linker regions of NTD and RBD showed large-scale protection against deuterium exchange, due to reduced conformational flexibility, indicating that their domain motions were severely restricted. Specifically, peptides spanning residues 399-420 of RBD showed protection from deuterium exchange (Extended data Fig. 10a) in the antibody-bound states of Spikeδ and Spikeo2, indicating that these three HuMabs bind at the same epitopes, akin to SpikeWuhan. Upon closer examination, the magnitude of HDX changes across SpikeWuhan, Spikeδ and Spikeo2 were different, owing to the differences in their binding affinities (Fig. 3e, f, Extended data Fig. 10). Furthermore, HDX kinetics observed across the epitope sites of individual RBD variants (RBDδ, RBDo1, and RBDo2) in the presence of LSI-CoVA-015, LSI-CoVA-016, and LSI-CoVA-017, were stronger than their corresponding Spike trimers. This indicate that variant-specific mutations on Spike induce subtle changes in the conformational dynamics, which alter the binding strengths of antibodies even though the epitope sites are conserved. This is further supported by the varying HDX effects observed across peptides spanning RBM in antibody-bound Spikeδ and Spikeo2 (Extended data Fig. 10a). This effect was more prominent for RBDδ, RBDo1 and RBDo2, where binding of LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 antibodies led to significant protection against deuterium exchange at the ACE2-binding sites (Fig. 3d, Extended data). Our results show allosteric destabilization of the RBD of Spikeδ and Spikeo2, and explain why an up-position of RBD is favoured in the VoCs49.
Within the S2 subunit of Spikeδ, LSI-CoVA-014/015/016 deuterium exchange profiles (Fig. 3e, Extended data Fig. 10b) were similar to that of SpikeWuhan, but were significantly different from Spikeo2 (Fig. 3f, Extended data Fig. 10). Peptides spanning the S1/S2 cleavage site showed minor differences in deuterium exchange and were lower than those observed for SpikeWuhan. The reduced deuterium exchange indicates that the S1/S2 cleavage site is more concealed in the Spikeδ and Spikeo2 variants, and is further occluded by antibody binding. These results also explain the reduced propensity of cleavage of VoCs and the inaccessibility to host proteases was observed for the Omicron VoC29,50. Most notable changes in conformation upon antibody binding were observed across peptide clusters spanning the FP1, FP2 and HR1 (Extended data Fig. 10b). Large-magnitude increases in deuterium exchange were observed across these sites for Spikeδ and Spikeo2, as compared to the effects observed for SpikeWuhan. Higher deuterium uptake correlates with increased dynamics and/or solvent accessibility across regions which span peptide clusters including 539-565 (subdomain 1), 757-779, and 936-974 (Fig. 3eii, 3fii). Increased deuterium exchange at these sites reflects greater solvent accessibility accompanied by the loss of intermonomer contacts. This translates to an allosteric effect by LSI-CoVA-014, LSI-CoVA-015, and LSI-CoVA-016 across the Spikeδ and Spikeo2 trimers. Overall, these results suggest that these antibodies likely induce destabilization of the Spike trimers into antibody-bound monomers. This destabilization of trimer into individual monomers also reflects to altered interaction of ACE2 with Spikeo2 by LSI-CoVA-015/016, accompanied by weak neutralization efficacy.
5A6, CoVA2-39 and CoVA2-04 bind Spikeδ but not Spikeo2
Our binding assays indicated that 5A6, CoVA2-39, and CoVA2-04 bound to the Delta variant, but not the Omicron variant (Fig. 3). We then compared the effects of binding of 5A6 and CoVA2-04 to Spikeδ and RBDδ, using HDXMS. Upon binding of these HuMAbs, significant protection against deuterium exchange was observed across the RBDs, both in isolated constructs (Extended data Fig. 11a) and in Spikeδ (Extended data Fig. 11b, c). While the HDX changes observed across the RBM-sites of SpikeWuhan was about ~4 Da, the changes observed for the Delta variant were lower in magnitude (~2.5Da) with a relatively smaller antibody-binding footprint. As the Delta variant has two key mutations at the RBM-site26, antibody footprint at the epitope site is reduced, affecting the overall conformational dynamics of RBDs of Spikeδ. Although 5A6, CoVA2-39 and CoVA2-04 directly compete with ACE2 binding, various studies have reported that both CoVA2-39 and CoVA2-04 are unable to neutralize Delta or Omicron variants51,52.
We further determined the effects of 5A6 and CoVA2-04 on the S2 subunit of Spikeδ. No significant change was observed for peptides covering the S1/S2 cleavage site. Decreased deuterium exchange was observed for peptides spanning FP2, HR1, connector domain (CD) and HR2. Importantly, peptides spanning the central helix (990-1010), showed increased deuterium exchange in the presence of these two antibodies, suggesting higher localized conformational dynamics. This in contrast to the effects observed for SpikeWuhan (Fig. 2). While the sites essential for trimerization of Spike showed increased conformational rigidity induced by 5A6 and CoVA2-04 HuMAbs, increased conformational mobility was observed across the central helix. Collectively, the binding assays and HDX results indicate that the HuMAbs recognizing the RBM antigenic supersite cannot bind Omicron at all, and bind Spikeδ variant with reduced affinity, by allosterically reducing the conformational dynamics of the S1 and the S2 subunit, thereby mediating overall stabilization.