key: cord-0689030-167f4gyg
authors: Zhou, Tongqing; Wang, Lingshu; Misasi, John; Pegu, Amarendra; Zhang, Yi; Harris, Darcy R.; Olia, Adam S.; Talana, Chloe Adrienna; Yang, Eun Sung; Chen, Man; Choe, Misook; Shi, Wei; Teng, I-Ting; Creanga, Adrian; Jenkins, Claudia; Leung, Kwanyee; Liu, Tracy; Stancofski, Erik-Stephane D.; Stephens, Tyler; Zhang, Baoshan; Tsybovsky, Yaroslav; Graham, Barney S.; Mascola, John R.; Sullivan, Nancy J.; Kwong, Peter D.
title: Structural basis for potent antibody neutralization of SARS-CoV-2 variants including B.1.1.529
date: 2021-12-28
journal: bioRxiv
DOI: 10.1101/2021.12.27.474307
sha: 647485bb62cdb48804104669e88b0a119371a8bb
doc_id: 689030
cord_uid: 167f4gyg

With B.1.1.529 SARS-CoV-2 variant’s rapid spread and substantially increased resistance to neutralization by vaccinee and convalescent sera, monoclonal antibodies with potent neutralization are eagerly sought. To provide insight into effective neutralization, we determined cryo-EM structures and evaluated potent receptor-binding domain (RBD) antibodies for their ability to bind and neutralize this new variant. B.1.1.529 RBD mutations altered 16% of the RBD surface, clustering on a ridge of this domain proximal to the ACE2-binding surface and reducing binding of most antibodies. Significant inhibitory activity was retained, however, by select monoclonal antibodies including A19-58.1, B1-182.1, COV2-2196, S2E12, A19-46.1, S309 and LY-CoV1404, which accommodated these changes and neutralized B.1.1.529 with IC50s between 5.1-281 ng/ml, and we identified combinations of antibodies with potent synergistic neutralization. Structure-function analyses delineated the impact of resistance mutations and revealed structural mechanisms for maintenance of potent neutralization against emerging variants. Summary Sentence We show potent B.1.1.529 neutralization by select antibodies and use EM structures to reveal how potency can be retained.

For B.1.1.529, we noted that all but three antibodies showed binding less than 31% of D614G. It is interesting to note that COV2-2196, S2E12, B1-182.1 and A23-58.1 utilize the same VH1-58 gene in their 135 heavy chain and target a similar region on the RBD (i.e., VH1-58 supersite), but show differential binding to the B.1.1.529 (i.e., 4%, 5%, 8% and 11%, respectively) and B.1.617.2 (i.e., 66%, 67%, 77% and 85%, respectively) (Fig 2B) . Even though the absolute differences in binding are minimal, the shared trend may be reflective of how the RBD tip mutation at T478K mutation is accommodated by each of these antibodies.

Finally, LY-CoV1404 revealed 61% binding to B.1.1.529 spike. Taken together, cell surface binding 140

suggests that while both A19-46.1 and LY-CoV1404 are likely to retain potent neutralizing activity against B.1.1.529, the remaining antibodies in our panel might show decreased neutralizing activity.

Using the same panel of monoclonal antibodies, we further assayed for each antibody's capacity to neutralize the B.1.1.529 variant. While VH1-58 supersite antibodies (Class I) show high neutralization activity against other variants, antibodies targeting the supersite were 40 to 126-fold worse (IC 50 38-269 145 ng/ml) against B.1.1.529 viruses than D614G (IC 50 0.9-2.0 ng/ml) (Fig 2C) . In addition, two other antibodies, CB6 (Class I) and ADG2 (Class I/IV) were shown to be severely impacted (IC 50 >10,000 ng/mL CB6 and 2037 ng/mL ADG2 to B.1.1.529 vs 31 and 50.5 ng/mL to D614G, respectively) ( Fig 2C) . We next analyzed Class II antibodies (i.e., LY-CoV555, C144, A19-46.1) and found that amongst these, neutralization by LY-CoV555 and C144 was completely abolished (IC 50 >10,000 ng/mL B.1.1.529 vs 3.6 150 and 5.1 ng/mL D614G, respectively). In contrast, we found that the A19-46.1 IC 50 neutralization was 223 ng/mL for B.1.1.529 vs 19.4 ng/mL for D614G ( Fig 2C) and was <6 fold of the previously reported IC 50 for WA-1 (39.8 ng/mL) (14). We next analyzed the Class III antibodies (i.e., A19-61.1, REGN10987, COV2-2130, C135, LY-CoV1404) and noted neutralization activity of A19-61.1, REGN10987 and C135 was completely abolished (IC 50 >10,000 ng/mL B.1.1.529 vs 19.4, 20.0, 10.8 ng/mL, respectively on 155 D614G), CoV2-2130 decreased 1581-fold (IC 50 5850 ng/mL B.1.1.529 vs 3.7 ng/mL D614G) and that of S309 decreased by ~8-fold (IC 50 281 ng/mL B.1.1.529 vs 36.1 ng/mL D614G) (Fig 2C) . Strikingly, in contrast to all of the other antibodies, we found that the neutralization of LY-CoV1404 against B.1.1.529 was unchanged relative to D614G (IC 50 5.1 ng/mL for B.1.1.529 vs 3 ng/mL for D614G) (Fig 2C) . Taken together, these data demonstrate that the mutations present in B.1.1.529 mediate resistance to antibodies. 160

We sought to determine the functional basis of B.1.1.529 neutralization and escape for Class I antibodies and to understand how potent neutralization might be retained. We analyzed Class I antibodies, CB6, B1-182.1 and S2E12, with differential B.1.1.529 neutralization (Fig 2C) . We first evaluated CB6 165 using virus particles containing single amino acid substitutions representing 13 of 15 single amino acid changes on the RBD of B.1.1.529 (i.e., all but S375F and G496S) or G496R; we noted while several minimally changed their neutralization IC 50 , only Y505H, S371L, Q493R and K417N decreased neutralization >5-fold, with IC 50 of 50, 212, 320, and >10,000 ng/mL, respectively ( Fig 3A) . This suggests that B.1.1.529 evades CB6-like antibodies through multiple mutations. Docking of the RBD-bound CB6 170 onto the B.1.1.529 structure revealed several B.1.1.529 residues may potentially clash with CB6.

Especially, K417N, Q493R and Y505H were positioned to cause severe steric hindrance to the CB6 paratope, consistent with the neutralization data (Fig. 3B) . We next evaluated two VH1-58 supersite antibodies, B1-182.1 and S2E12, which have highly similar amino acid sequences but show ~6-fold difference in B.1.1.529 neutralizing. These two antibodies remained highly potent (<10.6 ng/mL IC 50 ) for 175 all virus particles with single RBD mutations, with the largest change for Q493R, which caused a 7 and 5.4-fold decrease of neutralization for B1-182.1 and S2E12, respectively. These small differences in neutralization from single mutations suggest that multiple mutations of B.1.1.529 are working in concert to mediate escape from VH1-58 supersite antibodies. Docking of the RBD-bound B1-182.1 onto the B.1.1.529 structure indicated that the epitopes of these VH1-58-derived antibodies were confined by Q493R, S477N, 180 T478K and E484A (Fig. 3C) . With R493 pressing on one side of the antibody like a thumb, N477/K478 squeezed onto the other side of the antibody at the heavy-light chain interface like index and middle fingers ( Fig. 3C) . Analysis of the docked RBD-antibody complex showed that N477/K478 positioned at the junction formed by CDR H3, CDR L1 and L2 with slight clashes to a region centered at CDR H3 residue 100C (Kabat numbering) (Fig. 3D) . Sequence alignment of CDR H3 of VH1-58-derived antibodies 185 indicated that residue 100C varied in sidechain sizes, from serine in S2E12 to tyrosine in A23-58.1.

Analysis showed that size of 100C reversely correlated with neutralization potency IC 80 (p=0.046) (Fig.   3D, Fig. 2C) , suggesting VH1-58 antibodies could alleviate escape imposed by the B.1.1.529 mutations through reduced side chain size at position 100C to minimize clashes from N477/K478.

We next sought to determine the functional basis of B.1.1.529 neutralization and escape for two Class II antibodies, LY-CoV555 (26) and A19-46.1 (14) , which have B.1.1.529 IC 50 of >10,000 and 223 ng/mL, respectively ( Fig 2C) . By assessing the impact of each of the single amino acid changes in RBD from B.1.1.529, we found that for LY-CoV555, either E484A or Q493R resulted in complete loss of CoV555 neutralization (IC 50 >10,000 ng/mL) (Fig 4A) , while the same mutations did not affect A19-46.1.

For A19-46.1, no individual mutation reduce neutralization to the level noted in B.1.1.529; S371L had the highest effect, reducing the IC 50 to 72.3 ng/mL relative to 223 ng/mL for B.1.1.529. One potential explanation for this is that the Phe-Phe interaction, between 375 and 486, occurs in the context of three B1.1.529 alterations, S371L/S373P/S375F (Fig. 1B) . 200

To understand the structural basis of A19-46.1 neutralization of B.1.1.529, we obtained cryo-EM structure of the B.1.1.529 spike in complex with Fab A19-46.1 at 3.86 Å resolution (Fig 4B, Fig. S5 and Table S2 ). The structure revealed that two Fabs bound to the RBD in the "up" conformation in each spike with the third RBD in down position. Focused local refinement of the antibody-RBD region resolved the antibody-RBD interface (Fig. 4B , right). Consistent with previous mapping and negative stain EM data, 205 A19-46.1 binds to a region on RBD generally targeted by the Class II antibodies with an angle approximately 45 degrees towards the viral membrane. Binding involves all light chain CDRs and only CDR H3 of the heavy chain and buried a total of 805 Å 2 interface area from the antibody (Fig. 4C, left) .

With the light chain latching to the outer rim of the RBD and providing about 70 % of the binding surface, 9 A19-46.1 uses its 17-residue-long CDR H3 to form parallel strand interactions with RBD residues 345-350 210 ( Fig. 4B, right) like a sway brace. Docking RBD-bound ACE2 to the A19-46.1-RBD complex indicated that the bound antibody sterically clashes with ACE2 ( Fig. 4D) , providing the structural basis for its neutralization of B.1.1.529.

The 686 Å 2 epitope of A19-46.1 is located within an RBD region that lacks amino acid changes found in B.1.1.529. Of the 15 amino acid changes on RBD, three of residues, S446, A484 and R493, 215 positioned at the edge of epitope with their side chains contributing 8% of the binding surface. LY-CoV555, which targets the same region as a class II antibody, completely lost activity against B.1.1529. To gain structural insights on the viral escape of LY-CoV555, we superimposed the LY-CoV555-RBD complex onto the B.1.1.529 RBD. Even though LY-CoV555 approached the RBD with similar orientation to that of A19-46.1 (Fig. 4D ), its epitope shifted up to the ridge of the RBD and embraced B.1.1.529 alterations A484 220 and R493 within the boundary (Fig. 4D) 

To evaluate the functional basis of B.1.1.529 neutralization and escape for Class III antibodies and to understand how potent neutralization might be retained, we investigated a panel of Class III antibodies with differential potency, including A19-61.1, COV2-2130, S309 and LY-CoV1404 (Fig 5A) . Assessment of the impact of each of the 15 mutations in RBD revealed the G446S amino acid change results in a 230 complete loss in activity for A19-61.1; consistent with the complete loss of function of this antibody against B.1.1.529. For S309, we observed S373P and G496R to result in small changes to neutralization. Surprisingly, while S309 retains moderate neutralizing activity against B.1.1.529, we found the S371L amino acid change to abolish S309 neutralization. This suggests that combinations of S371L with other B.1.1.529 mutations can result in structural changes in spike that allows S309 to partially overcome the 235 S371L change. The evaluation of COV2-2130 did not identify significant differences in neutralization, suggesting a role for an untested mutation or combinations of amino acid changes for the decrease in neutralization potency observed against the full virus. Finally, consistent with the overall high potency of LY-CoV1404 against all tested VOCs, we did not identify an amino acid change that impacted its function.

To understand the structural basis of Class III antibody neutralization and viral escape, we 240 determined the cryo-EM structure of WA-1 S2P in complex with Fab A19-61.1 (and Fab B1-182.1 to aid EM resolution of local refinement) at 2.83 Å resolution (Fig. 5B, Fig. S6 and Table S1 ). The structure revealed that two RBDs were in the up-conformation with both antibodies bound, and the third RBD was in the down-position with only A19-61.1 bound, indicating A19-61.1 could recognize RBD in both up and down conformation (Fig. 5B ). Local refinement of the RBD-Fab A19-61.1 region showed that A19-61.1 245 targeted the Class III epitope with interactions provided by the 18-residue-long CDR H3 from the heavy chain, and all CDRs from the light chain (Fig. 5B) . Docking the A19-61.1 structure to the B.1.1.529 spike structure indicated B.1.1.529 mutations S446, R493 and S496 might interfere with A19-61.1. Analysis of the side chain interaction identified Y111 in CDR H3 posed severe clash with S446 in RBD that could not be resolved by loop flexibility (Fig. 5C) , explaining the loss of A19-61.1 neutralization against G446S-250 containding SARS-CoV-2 variants.

Neutralization assays indicated that COV2-2130, S309 and LY-CoV1404 retained neutralization potency against B.1.1.529. Docking of CoV2-2130 indicated that CoV2-2130 targeted a very similar epitope to that of A19-61.1 with interactions mainly mediated by its CDR L1 and L2 and avoiding close contact with R493 and S496. However, the OH group at the tip of Y50 in CDR L2 posed a minor clash with 255 S446 in RBD, explaining the structural basis for the partial conservation of neutralization by CoV2-2130 ( Fig. 5D) . Antibody S309 showed higher potency against B.1.1.529 than CoV2-2130. Docked complex of S309 and RBD showed the G339D mutation is located inside the epitope and clashes with CDR H3 Y100, however, the void space between S309 and RBD might accommodate an alternate tyrosine rotamer. The S371L/S373P/S375F mutations changed the conformation of their residing loop and may push the glycan 260 on N343 towards S309 to reduce binding (Fig. 5E) . LY-CoV1404 was not affected by B.1.1.529 mutations.

Docking of the LY-CoV1404 onto the B.1.1.529 RBD identified four amino acid substitutions located at the edge of its epitope. Three of the residues, K440, R498 and Y501, only make limited side chain interactions with LY-CoV1404. The 4 th residue, G446S, appeared to cause a potential clash with CDR H2 R60. However, comparison of both LY-CoV1404-bound and non-bound RBD indicated the loop where 265 S446 resided had conformational flexibility to allow LY-CoV1404 binding (Fig. 5F) . Overall, the epitopes to Class III antibodies were mainly located on mutation-free RDB surfaces with edges contacting a few B.1.1.529 alterations (Fig. 5G) . LY-CoV1404 retained high potency by accommodating all four B.1.1.529 alteration at edge of its epitope by exploiting loop mobility or by minimizing side chain interactions. with an appreciably improved potency (i.e., IC 50 of 50.8, 28.3 and 58.1 ng/mL) over the individual component antibodies (Fig 6A,B) . Each of these utilized a VH-158 supersite antibody and showed a 5 to 280 115-fold improvement over the component antibodies (Fig 6B) , suggesting an effect that is more than an additive for the specific combination against B.1.1.529.

To understand the structural basis of the improved neutralization by the cocktail of B1-182.1 and A19-46.1 we determined the cryo-EM structure of the B.1.1.529 S2P spike in complex with Fabs of B1-182.1 and A19-46.1 at 3.86 Å resolution (Fig. 6C, Fig. S7 and Table S1 ). The prevalent 3D reconstruction 285 revealed that the spike recognized by the combination of these two antibodies was the 3-RBD-up conformation with both Fabs bound to each RBD (Fabs on one of the RBDs were lower in occupancy). The spike had a 1.6 Å RMSD relative to the 3-RBD-up WA-1 structure (PDB ID: 7KMS). Overall, the structure showed that both Class I and II antibodies were capable of simultaneously recognizing the same RBD, and the combination increased the overall stoichiometry compared to two Fabs per trimer observed in the S2P-290 A19-46.1 structure described above. Of all the antibodies tested, we note that all VH1-58-derived antibodies retained reasonable level of neutralization against B.1.1.529 while members of other antibody classes suffered complete loss of activity. VH1-58 antibodies have minimal numbers of impacting B.1.1.529 alterations in their epitopes and can evolve means to alleviate the impact. We speculate that binding of the first antibody induced the spike into RBD-up-conformation and facilitated binding of the second RBD-up-295 conformation preferring antibody, thereby synergistically increasing the neutralization potency of the cocktail compared to the individual antibodies.

SARS-CoV2 variants of concern provide a window into the co-evolution of key host-pathogen 300 interactions between the viral spike, human ACE2 receptor and the human immune system. The RBD is a major target for neutralizing antibodies in both survivors and vaccinees. Since 15 of the 37 mutations in the B.1.1.529 variant spike reside within the RBD, there is a great need to understand the mechanisms by which RBD variations evolve, what constraints exist on the evolution and whether there are approaches that can be taken to exploit this understanding to develop and maintain effective antibody therapeutics and vaccines. 305

The B.1.1.529 cryo-EM structure showed a cluster of RBD mutations proximal to the ACE2binding surface, which alters the electrostatic potential of the interface. However, in contrast to studies that showed increased human ACE2 affinity to mutated and variant RBD subdomains (20, 34, 35) , our studies found that in the context of trimeric spike proteins, variant amino acid changes did not provide a biologically meaningful alteration in affinity. This suggests that there is either no further fitness benefit to be gained by 310 improving affinity, that affinity improving changes are being used to compensate for mutations that are deleterious for ACE2 binding but allow immune escape, or both.

We used a series of functional and structural studies to define the mechanisms by which B.1.1.529 is either neutralized by or mediates escape from host immunity. To functionally frame our analyses, we utilized the Barnes classification, which categorizes antibodies based on their binding to the ACE2 binding 315 site and the position of RBD. Our findings for Class I VH1-58 supersite showed that B.1.1.529 requires a series of mutation that are not individually deleterious to bracket the antibody and reduce its potency.

Notably our data suggests that VH1-58 antibodies can alleviate the deleterious impact of this pinching effect by reducing the size of CDR H3 residue 100C to avoid clashes from B.1.1.529 mutations. Since VH1-58 supersite are amongst the most potent and broadly neutralizing anti-SARS-CoV-2 antibodies (14, 25, 29, 320 36) , our findings point the way toward structure-based designs of existing antibodies to mitigate against amino acid changes at these positions.

For the Class II antibody A19-46.1, the angle of approach and a long-CDRH3 combine allow it to target the mutation-free face on RBD and minimize contacting the mutations on the ridge of B.1.1.529 RBD. We observed that A19-46.1 binding requires the RBD-up conformation, and that the S371L 325 substitution, which is located away from the A19-46.1 epitope and near the RBD hinge, partially reduces the neutralization of A19-46.1. Comparing the effect of S371L on neutralization by A19-46.1 and LY-CoV555 (Fig. 4A) , which recognizes both RBD-up and -down conformation, suggested that L371 (and potentially P373/F375) is critical for controlling the RBD-up or -down conformation in B.1.1.529. This concept is supported by the finding that combination with a Class I antibody (such as B1-182.1) 330 synergistically enhances A19-46.1 neutralization (Fig. 6A) .

For Class III antibodies, only one prototype antibody showed complete loss of B.1.1.529 neutralization. Using structural and functional approaches we determined that viral escape was mediated by the G446S amino acid change. This result indicates that potent Class III antibodies might be induced through structure-based vaccine designs that mask residue 446 in RBD. Additionally, the existence of 335 G446S sensitive and resistant antibodies with significant epitope overlap suggest the use of spikes with G446S substitution can be utilized to evaluate the quality of Class III immune response in serum-based epitope mapping assays (37, 38) .

Our analysis of antibodies of clinical importance is consistent with previous reports (32, (39) (40) (41) and showed that S309 and COV2-2196 neutralized to similar degrees. Importantly, we report that unlike 340 other antibodies, the highly potent LY-CoV1404 does not lose neutralization potency against B.1.1.529.

We identified combinations of antibodies that show more than additive increases in neutralization against B.1.1.529, including COV2-2196/COV2-2130, B1-182.1/A19-46.1 and B1-182.1/S309. Each pair contains a VH1-58 supersite antibody that binds RBD in the up position. We speculated that pairing antibodies that neutralize better in the up-RBD conformation with these VH1-58 antibodies may provide a mechanism for 345 better neutralization by the former. The S371L/S373P/S375F alterations in the RBD-up protomer form interprotomer interactions to RBD in the RBD-down protomer and stabilize the B.1.1.529 spike into a single-RBD-up conformation. RBD-up-preferring antibody like the VH1-58-derived B1-182.1, which is not affected by S371L substitution, can effectively break up the interaction to induce the 3-RBD-up conformation and therefore, enhance binding of other antibodies (such as A19-46.1) that require the RBD 350 up-conformation. The identification of SARS-CoV-2 monoclonal antibodies that cooperatively function is similar to that seen previously for other viruses (42) , and supports the concept of using combinations to both enhance potency and mitigate the risk of escape.

an-emergent-sars-cov-2-lineage-in-manaus-preliminary-findings/586). ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). *n.d.= not determined due to incomplete neutralization that plateaued at <80% (See Fig S4B) . 735 S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was 31 available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 800 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). *n.d.= not determined due to incomplete neutralization that plateaued at <80% (See Fig S4B) . 

A. Lentiviruses pseudotyped with SARS-CoV-2 spike proteins from D614G or D614G plus the indicated point mutations found within the B.1.1.529 spike were incubated with serial dilutions of the indicated antibodies, transduced 293T-ACE2 cells and IC 50 and IC 80 values determined. S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml 

A. Lentiviruses pseudotyped with SARS-CoV-2 spike proteins from D614G or D614G plus the indicated point mutations found within the B.1.1.529 spike were incubated with serial dilutions of the indicated antibodies, transduced 293T-ACE2 cells and IC 50 and IC 80 values determined. S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). A. Lentiviruses pseudotyped with SARS-CoV-2 spike proteins from D614G or D614G plus the indicated point mutations found within the B.1.1.529 spike were incubated with serial dilutions of the indicated antibodies and IC 50 and IC 80 values determined. A19-61.1 and LY-COV1404 were assayed on 293T-ACE2 cells while S309 and CoV2-2130 were tested on 293 flpin-TMPRSS2-ACE2 cells. S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). B. Cryo-EM structure of SARS-CoV-2 WA-1 spike in complex with class I antibody B1-182.1 and class III antibody A19-61.1 at 2.83 Å resolution. Overall density map is shown with protomers colored light green, gray and wheat. Two RBDs were in the up conformation with each binding both Fabs, and one RBD was in the down position with A19-61.1 bound. RBD, B1-182.1 and A19-61.1 are colored olive and cyan, respectively (left). Structure of the RBD with both Fabs bound after local focused refinement was shown to the right in cartoon representation. RBD is shown in green cartoon and antibody light chains are colored light blue (middle). Epitope of A19-61.1 is shown as cyan colored surface on RBD with interacting CDRs labeled (right). The contour level of cryo-EM map is 5. A. Lentiviruses pseudotyped with SARS-CoV-2 spike proteins from D614G or D614G plus the indicated point mutations found within the B.1.1.529 spike were incubated with serial dilutions of the indicated combination of antibodies and IC 50 and IC 80 values determined. S375F and G496S viruses were not available and are shown as "not tested" (n.t.). G496R was available and substituted for G496S. Ranges are indicated with white (>10,000 ng/ml), light blue (>1000 to ≤10,000 ng/ml), yellow (>100 to ≤1000 ng/ml), orange (>50 to ≤100 ng/ml), red (>10 to ≤50 ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). B. Neutralization IC 50 (ng/mL) values for each of the indicated cocktail (x-axis) or its component antibodies. The IC 50 for first antibody is listed as mAb1 (black), the second antibody as mAb2 (grey) or cocktail (red). 

Figs. S1 to S7 Tables S1 to S2

Soluble 2P-stabilized SARS-CoV-2 spike proteins were expressed by transient transfection (6, 43) .

Briefly, plasmid was transfected using Expifectamine into Expi293 cells (Life Technology) and the cultures enhanced 16-24 hours post-transfection. Following 4-5 days incubations at 120 rpm, 37 °C, 9% CO2, supernatant was harvested, clarified via centrifugation, and buffer exchanged into 1X PBS. Protein of interests were then isolated by affinity chromatography using Ni-NTA resin (Roche) followed by size 885 exclusion chromatography on a Superose 6 increase 10/300 column (GE healthcare).

Expression and purification of biotinylated S2P used in binding studies were produced by an incolumn biotinylation method as previously described (43) . Using full-length SARS-Cov2 S and human ACE2 cDNA ORF clone vector (Sino Biological, Inc) as the template to the ACE2 dimer proteins. The ACE2 PCR fragment (1~740aa) was digested with Xbal and BamHI and cloned into the VRC8400 with 890

Avi-HRV3C-single chain-human Fc-his (6x) tag on the C-terminal. All constructs were confirmed by sequencing. Proteins were expressed in Expi293 cells by transfection with expression vectors encoding corresponding genes. The transfected cells were cultured in shaker incubator at 120 rpm, 37 °C, 9% CO2 for 4~5 days. Culture supernatants were harvested and filtered, and proteins were purified through a Hispur Ni-NTA resin (Thermo Scientific) and following a Hiload 16/600 Superdex 200 column (GE healthcare, 895

Piscataway NJ) according to manufacturer's instructions. The protein purity was confirmed by SDS-PAGE.

Sequences were selected for synthesis to sample expanded clonal lineages within our dataset and convergent rearrangements both among donors in our cohort and compared to the public literature. In C. The gold-standard Fourier shell correlation resulted in a resolution of 3.85 Å for the overall map using non-uniform refinement with C1 symmetry (left panel); the orientations of all particles used in the final refinement are shown as a heatmap (right panel). D. The gold-standard Fourier shell correlation resulted in a resolution of 5.08 Å for the masked local refinement of the RBD:A19-46.1 interface (left panel) obtained using particle subtraction followed by local refinement; the orientations of all particles used in the local refinement are shown as a heatmap (right panel). E. The local resolution of the final overall map and locally refined map is shown contoured at 0.068 (4.5σ) and 0.311 (13.9σ), respectively. Resolution estimation was generated through cryoSPARC using an FSC cutoff of 0.5. F. Representative density is shown for the spike residues 977-993 region. The contour level is 1.5σ. A. Representative micrograph. B. Representative 2D class averages are shown. C. The gold-standard Fourier shell correlation resulted in a resolution of 2.83 Å for the overall map using non-uniform refinement with C1 symmetry (left panel); the orientations of all particles used in the final refinement are shown as a heatmap (right panel). D. The gold-standard Fourier shell correlation resulted in a resolution of 3.1 Å for the masked local refinement of the RBD:A19-61.1-B1-182.1 ternary interface (left panel) obtained using particle subtraction followed by local refinement; the orientations of all particles used in the local refinement are shown as a heatmap (right panel). E. The local resolution of the final overall map and locally refined map is shown contoured at 0.154 (5.7σ) and 0.246 (22.2σ), respectively. Resolution estimation was generated through cryoSPARC using an FSC cutoff of 0.5. F. Representative density is shown for portion of the CDR H3 of A19-61.1 after local refinement. It is evident that the side chains are well defined. The contour level is 10σ. C. The gold-standard Fourier shell correlation resulted in a resolution of 3.85 Å for the overall map using non-uniform refinement with C1 symmetry (left panel); the orientations of all particles used in the final refinement are shown as a heatmap (right panel). D. The local resolution of the final overall map is shown contoured at 0. 25 (8.5σ 

A new coronavirus associated with human respiratory disease in China

Johns Hopkins University COVID-19 Dashboard

SARS-CoV-2 Omicron has extensive but incomplete escape of Pfizer BNT162b2 elicited neutralization and requires ACE2 365 for infection

Enhancing Readiness for Omicron (B.1.1.529): Technical Brief and Priority Actions for Member States

Broad betacoronavirus 375 neutralization by a stem helix-specific human antibody

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Structural and Functional Analysis of the D614G SARS-CoV-2 Spike Protein Variant

Potent SARS-CoV-2 neutralizing antibodies directed against spike N-terminal domain target a single supersite

SARS-CoV-2 can recruit a heme metabolite to evade antibody 395 immunity

Neutralizing antibody 5-7 defines a distinct site of vulnerability in SARS-CoV-2 spike N-terminal domain

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor

SARS-CoV-2 neutralizing antibody structures inform therapeutic strategies

Convergent antibody responses to SARS-CoV-2 in convalescent individuals

Ultrapotent antibodies against diverse and highly transmissible SARS-CoV-2 variants

Preliminary genomic characterisation of an emergent SARS-CoV-2 lineage in the UK defined by a novel set of spike mutations

Detection of a SARS-CoV-2 variant of concern in South Africa

Genomic characterisation of an emergent SARS-CoV-2 lineage in Manaus: preliminary findings. virological.org (2021)

Broad and potent activity against SARS-like viruses by an engineered human monoclonal antibody

LY-CoV1404 potently neutralizes SARS-CoV-2 variants

Ultrapotent human antibodies protect against SARS-CoV-2 challenge via multiple mechanisms

The neutralizing antibody, LY-CoV555, protects against SARS-CoV-2 infection in non-human primates

A human neutralizing antibody targets the receptor-binding site of SARS-CoV-2

Studies in humanized mice and convalescent humans yield a SARS-CoV-2 antibody cocktail

Potently neutralizing and protective human antibodies against SARS-CoV-2

Cross-neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV antibody

Mapping Neutralizing and Immunodominant Sites on the SARS-CoV-2 Spike Receptor-Binding Domain by Structure-Guided High-Resolution Serology

Mapping mutations to the SARS-CoV-2 RBD that escape binding by different classes of antibodies

Omicron : A heavily mutated SARS-CoV-2 variant exhibits stronger binding to ACE2 and potently escape approved COVID-19 therapeutic antibodies

Genetic and structural basis for recognition of SARS-CoV-2 spike protein by a twoantibody cocktail

High-throughput, single-copy sequencing reveals SARS-CoV-2 spike variants coincident with mounting humoral immunity during acute COVID-19

Protection against SARS-CoV-2 Beta variant in mRNA-1273 vaccine-boosted nonhuman primates. Science (80-. )

Cooperativity Enables Non-neutralizing Antibodies to Neutralize Ebolavirus

Structure-Based Design with Tag-Based Purification and In-Process Biotinylation Enable Streamlined Development of SARS-CoV-2 Spike Molecular Probes

Structure of HIV-1 gp120 V1/V2 domain with broadly neutralizing antibody PG9

Cross-Reactive Neutralizing Antibody Responses Elicited by SARS-CoV-2 501Y

Similarly, variable lambda and kappa light chain 905 sequences were human codon optimized, synthesized and cloned into CMV/R-based lambda or kappa chain expression vectors, as appropriate (Genscript)

For antibody expression, equal amounts of heavy and light chain plasmid DNA were transfected into Expi293 cells (Life Technology) by using Expi293 transfection reagent (Life Technology). The transfected cells were cultured in shaker incubator at 120 rpm, 37 °C, 9% 915 CO2 for 4~5 days. Culture supernatants were harvested and filtered, mAbs were purified over Protein A (GE Health Science) columns. Each antibody was eluted with IgG elution buffer (Pierce) and immediately neutralized with one tenth volume of 1M Tris-HCL

Full-length S constructs Codon optimized cDNAs encoding full-length S from SARS CoV-2 (GenBank ID: QHD43416.1) were synthesized, cloned into the mammalian expression vector VRC8400 (46, 47) and confirmed by sequencing. S containing D614G amino acid change was generated using the wt S sequence. Other variants containing single or multiple aa changes in the S gene from the S wt or D614G were made by mutagenesis 925

Agilent) or via synthesis and cloning (Genscript)

The S genes containing single RBD amino acid changes from the B.1.1.529 variant were generated based on D614G construct by mutagenesis. These full-length S plasmids 935 were used for pseudovirus production and for cell surface binding assays. Generation of 293 Flpin-TMPRSS2-ACE2 cell line 293 Flpin-TMPRSS2-ACE2 isogenic cell line was prepared by co-transfecting pCDNA5/FRT plasmid encoding TMPRSS2-T2A-ACE2 and pOG44 plasmid encoding Flp recombinase in 293 Flpin 940 parental cell line (Thermo Fisher, Cat R75007). Cells expressing TMPRSS2-ACE2 were selected using Hygromycin at 100 micrograms/ml. TMPRSS2 and ACE2 expression profiles in 293 Flpin-TMPSS2-ACE2 were characterized by flow cytometry using a mouse monoclonal antibody against TMPRSS2 (MillliporeSigma) followed by an anti-mouse IgG1 APC conjugate (Jackson laboratories) and a molecular probe containing the SARS-CoV-2 receptor binding domain tagged with biotin

Pseudovirus neutralization assay S-containing lentiviral pseudovirions were produced by co-transfection of packaging plasmid pCMVdR8.2, transducing plasmid pHR' CMV-Luc, a TMPRSS2 plasmid and S plasmids from SARS 950

293T-ACE2 cells (provided by Dr. Michael Farzan) or 293 flpin-TMPRSS2-ACE2 cells were plated into 96-well white/black Isoplates (PerkinElmer, Waltham, MA) at 75,00 cells per well the day before infection of SARS CoV-2 pseudovirus. Serial dilutions of mAbs were Determination of binding kinetics of ACE2

Fc-reactive anti-human IgG antibody (Cytiva) wa coupled to the CM5 chip to approximately 10,000 RU, and dimeric, Fc-tagged ACE2 (ACRO Biosystems) at 35 g/mL was captured for 60 seconds at 10 L

Serially diluted SARS-CoV-2 S2P variants starting at 100 nM were flowed through the sample and 985 reference channels for 180 seconds at 30 L/min, followed by a 300 second dissociation phase at 30 uL/min. The chip was regenerated using 3 M MgCl 2 for 30 seconds at 50 L/min. Blank sensorgrams were obtained with HBS-EP+ buffer. Blank-corrected sensorgrams of the S2P concentration series were fitted globally with Biacore S200 evaluation software using a 1:1 model of binding

Cryo-EM specimen preparation and data collection

For the spike-Fab complexes, the stabilized SARS-CoV-2 spikes of B.1.1.529 or WA-1were1. were mixed with Fab or Fab combinations at 995 a molar ratio of 1.2 Fab per protomer in PBS with final spike protein concentration at 0.5 mg/ml. n-Dodecyl β-D-maltoside (DDM) detergent was added to the protein complex mixtures shortly before vitrification to a concentration of 0.005%. Quantifoil R 2/2 gold grids were subjected to glow discharging in a PELCO easiGlow device (air pressure: 0.39 mBar, current: 20 mA, duration: 30 s) immediately before specimen preparation. Cryo-EM grids were prepared using an FEI Vitrobot Mark IV plunger with the following 1000 settings: chamber temperature of 4°C, chamber humidity of 95%, blotting force of -5, blotting time of 2 to 3.5 s, and drop volume of 2.7 µl. Datasets were collected at the National CryoEM Facility (NCEF), National Cancer Institute, on a Thermo Scientific Titan Krios G3 electron microscope equipped with a Gatan Quantum GIF energy filter (slit width: 20 eV) and a Gatan K3 direct electron detector (Table S2). Four movies per hole were recorded in the counting mode using Latitude software. The dose rate was 14.65 e-1005 /s/pixel. Cryo-EM data processing and model fitting Data process workflow

For local refinement to resolve the RBD-antibody interface, a mask for the entire spike-antibody complex without the RBD-antibody region was used to extract the particles and a mask encompassing the RBD-antibody region was used for refinement. The overall resolution was 3.29 Å for the map of B.1.1.529 spike alone structure, 3.85 Å for the map of B.1.1.529 spike in complex with A19-1015 46.1, 2.83 Å for the map of WA-1 spike in complex with A19-61.1 and B1-182.1, and 3..86Å for the map of B.1.1.529 spike in complex with A19-46.1 and B1-182.1 . The coordinates for the SARS-CoV-2 spike with B1-182.1 molecules bound at pH 7.4 (PDB ID: 7MM0) were used as initial models for fitting the cryo-EM map

Molprobity (52) was used to validate geometry and check structure quality at 1020 each iteration step. UCSF Chimera and ChimeraX were used for map fitting and manipulation (53).Data process workflow, including motion correction, CTF estimation, particle picking and extraction

For local refinement to resolve the RBD-antibody interface, a mask for the entire 1025 spike-antibody complex without the RBD-antibody region was used to extract the particles and a mask encompassing the RBD-antibody region was used for refinement. The overall resolution was 3.29 Å for the map of B.1.1.529 spike alone structure, 3.85 Å for the map of B.1.1.529 spike in complex with A19-46.1, 2.83 Å for the map of WA-1 spike in complex with A19-61

Iterative manual model building and real-space refinement were carried out in Coot (48) and in Phenix 1.19.2 (51), respectively. Molprobity (52) was used to validate geometry and check structure quality at each iteration step. Map fitting and manipulation and display were performed with UCSF Chimera and ChimeraX were used for map fitting and manipulation

Differential Scanning Calorimetry (DSC)

DSC measurements were performed using a VP-ITC (Microcal) instrument

All steps of pre-soaking, binding and dissociation were performed in 1045 PBS with 1% BSA at pH 7.4. IgGs and dACE2-Fc were loaded onto Anti-Human Fc Sensor Tips (FortéBio) at a concentration of 1-4g/mL, resulting in a load response of 0.85-1.5 nm. The plates were agitated at 1,000 rpm and the experiment run at 30°C. Antibodies and ACE2 were loaded onto the tips for 2 minutes, bound to 100nM S2P protein for 5 minutes and dissociated in buffer for 5 minutes

ng/ml), maroon (>1 to ≤10 ng/ml), and purple (≤1 ng/ml). 840 B. Neutralization IC 50 (ng/mL) values for each of the indicated cocktail (x-axis) or its component antibodies. The IC 50 for first antibody is listed as mAb1 (black), the second antibody as mAb2 (grey) or cocktail (red).