key: cord-1018074-n1yd4iae
authors: Gobeil, Sophie M-C.; Henderson, Rory; Stalls, Victoria; Janowska, Katarzyna; Huang, Xiao; May, Aaron; Speakman, Micah; Beaudoin, Esther; Manne, Kartik; Li, Dapeng; Parks, Rob; Barr, Maggie; Deyton, Margaret; Martin, Mitchell; Mansouri, Katayoun; Edwards, Robert J.; Eaton, Amanda; Montefiori, David C.; Sempowski, Gregory D.; Saunders, Kevin O.; Wiehe, Kevin; Williams, Wilton; Korber, Bette; Haynes, Barton F.; Acharya, Priyamvada
title: Structural diversity of the SARS-CoV-2 Omicron spike
date: 2022-03-25
journal: Mol Cell
DOI: 10.1016/j.molcel.2022.03.028
sha: 1d2651093a2e769c488a390ac1929920a75c4fa7
doc_id: 1018074
cord_uid: n1yd4iae

Aided by extensive spike protein mutation, the SARS-CoV-2 Omicron variant overtook the previously dominant Delta variant. Spike conformation plays an essential role in SARS-CoV-2 evolution via changes in receptor binding domain (RBD) and neutralizing antibody epitope presentation affecting virus transmissibility and immune evasion. Here, we determine cryo-EM structures of the Omicron and Delta spikes to understand the conformational impacts of mutations in each. The Omicron spike structure revealed an unusually tightly packed RBD organization with long range impacts that were not observed in the Delta spike. Binding and crystallography revealed increased flexibility at the functionally critical fusion peptide site in the Omicron spike. These results reveal a highly evolved Omicron spike architecture with possible impacts on its high levels of immune evasion and transmissibility.

The SARS-CoV-2 Omicron (B.1.1.529) variant was identified November 24 th , 2021 in South Africa, declared a Variant of Concern (VOC) by the World Health Organization on November 26 th , and has rapidly replaced Delta (B.1.617.2) as the dominant form of SARS-CoV-2 circulating globally. The Omicron spike (S) protein harbors 30 mutations and is the most immune evasive VOC identified thus far, surpassing Beta (B.1.351) in its ability to resist neutralization by antibodies (Abs) (Figure S1 ) , Hoffmann et al., 2021 , Planas et al., 2021 . Structural studies have been instrumental in revealing changes in VOC S protein conformations, and in understanding the atomic level mechanisms that drive higher transmissibility and immune evasion (Gobeil et al., 2021b , McCallum et al., 2021c , McCallum et al., 2021a .

The pre-fusion SARS-CoV-2 S protein is composed of the S1 and S2 subunits that undergo structural changes to facilitate receptor binding and fusion with the host-cell membrane . While the S2 subunit is conformationally stable prior to receptor engagement, the S1 subunit, with its mobile N-terminal domain (NTD) and receptorbinding domain (RBD), is inherently dynamic. The RBD transitions between a "closed" ("down") state where the binding site for the ACE2 receptor is occluded, and an "open" ("up") state that exposes the ACE2 binding site . Following receptor binding and proteolytic processing of the S protein, the S2 subunit undergoes large conformational changes that result in release of the fusion peptide (FP) to mediate fusion of the virus and host-cell membranes. RBD dynamics are impacted by interprotomer RBD-to-RBD and RBD-to-NTD contacts, as well as by other S protein structural units, including the SD1 and SD2 subdomains, and the "N2R (NTD-to-RBD) linker" that connects the NTD and RBD within a protomer. We previously described how the S1 domain interactions are modulated by VOCs to alter the S protein RBD presentation and how these can be exploited for immunogen design (Gobeil et al., 2021b , Gobeil et al., 2021a , Henderson et al., 2020 .

Here, we determine structures of native, unstabilized Omicron and Delta S protein ectodomains to understand how the acquired mutations alter their conformational states, and influence receptor binding site and Ab epitope presentation. The S ectodomains were prepared in J o u r n a l P r e -p r o o f GSAS-2P-Omicron, suggesting that the 2P mutations are limiting the conformational diversity of the Omicron S protein ectodomain. Reduced binding for the 2P vs non-2P Omicron spike is also seen for the 3-RBD-up conformation binding DH1047 and S2X259 Abs , Tortorici et al., 2021 . We therefore determined structures of the native Omicron S protein ectodomain in the S-GSAS form to examine the impacts mutations had on its conformation (Figure 1, Data S2, Table 1 ). In the cryo-EM dataset, we identified 3-RBD-down, 1-RBD-up and 2-RBD-up populations of the S protein ectodomain ( Figure 1C) . The Omicron spike harbors 16 amino acid substitutions in the RBD (Figure 1D) , of which several have been shown to mediate ACE2 recognition and/or immune escape (Mannar et al., 2022 .

The 3-RBD-down populations classified into two asymmetric reconstructions with each displaying close RBD-RBD pairing. Of the two 3-RBD-down structures, named O1 and O2, the interprotomer domain arrangement appeared more symmetric in O2 than in O1. This asymmetry was visualized in difference distance matrices (DDMs) that provide superposition-free comparisons between a pair of structures by calculating the differences between the distances of each pair of Ca atoms in a structure and the corresponding pair of Ca atoms in the second structure ( Figure S2) . In each reconstruction, we assigned the tag Protomer 1 to the protomer with the weakest RBD map density, which is suggestive of enhanced mobility ( Figure 1C ).

Interprotomer interactions between the "down" RBDs were mediated by a loop containing 3 amino acid substitutions, S371L, S373P and S375F. A rearrangement of the loop caused by the S373P mutation facilitated closer packing of the RBD-RBD interface via interactions of S373P and S375F with the N501Y and Y505H substitutions in the adjacent RBD ( Figure 1E ). In the 1-RBD-up structure the S375F substitution created an interprotomer interaction with residue F486 of the adjacent RBD-down protomer ( Figure 1F ). In the 2-RBD-up state, both up RBDs were disordered ( Figure 1C) . Overall, these results show that the mutations in the Omicron S protein induced coupling of the RBDs causing unique S protein closed and open structural states.

We previously showed that the RBD up/down transitions are accompanied by movements of the SD1/SD2 subdomains, and of the N2R linker (residues 306-334) that connects the NTD and RBD within a single protomer unit (Figure 2A ) (Gobeil et al., 2021a) . In the RBD "down" state the N2R region stacks against and contributes a β strand to each SD1 and SD2 subdomain ( Figure 2B) . Notably, in the Omicron 3-RBD-down O1 structure (PDB ID: 7TF8) ( Figures 1C   and 2B ), we found that the N2R region secondary structures in the mobile Protomer 1 was J o u r n a l P r e -p r o o f modified, with a break in the SD2-associated β strand stabilized by a new intraprotomer salt bridge formed between R319 and NTD residue E298 (Figures 2B and 2C ). This was accompanied by packing of residue F318 against the NTD 295-303 helix. This rearrangement permits interaction with a typically disordered segment of SD2, named "SD2 flex" here (residues 619-642). Stabilization of SD2 flex was facilitated by Van der Waals interactions of residue Ile624 (SD2 flex) with SD2 residues V595 and Y612 ( Figure 2C ) and W633 packing between NTD residue P295 and N2R linker residue R319. The spatial positioning of the NTD helix 295-303, of the SD2 loop and of the SD1 loop 554-565 were similar between the "up" protomer (Protomer 1 ) of the 1-RBD-up structure and the mobile protomer (Protomer 1 ) of the 3-RBD-down structure, suggesting that the mobile protomer may be poised to transition to the "up" position. The mobile SD1 region was stabilized in both the 3-RBD-down and 1-RBD-up Omicron S proteins via interprotomer hydrogen bonding with the S2 subunit mediated by residues D568 and T572 of SD1 with the N856K amino acid substitution near the Omicron S fusion peptide (Figures 2E and 2F) . A hydrogen bond also formed between the carboxyl oxygen of residue T315 of the N2R linker in the "up" protomer with the N764K substitution in the Omicron S protein. Thus, strategically placed residue substitutions in the Omicron S protein stabilize the highly mobile regions in the S1 subunit. Together, the close packing of the RBDs in the 3-RBD-down Omicron S protein and the N2R rearrangement in the "down" protomers define the wide range of conformational impacts that emanate from the extensive network of its acquired mutations.

We next studied the Delta S protein to understand the differences in its structural properties that may underlie the differences in its pathobiology with Omicron. We determined cryo-EM structures of the S-GSAS-Delta S protein ectodomain (Figure 3 ; Data S2; Table 2 ).

The Delta variant S protein includes two substitutions and a deletion in the NTD, two RBD substitutions, a P681R substitution proximal to the furin cleavage site, and a D950N substitution in the HR1 region of the S2 subunit (Figures 3A and S1 ). We identified 3-RBD-down (D1), 1-RBD-up (D2) and 2-RBD-up (D3) S protein ectodomain populations, as in the Omicron dataset, in addition to a population, named D4, which exhibited very high disorder in one of its S1 subunits such that the entire S1 subunit, including the NTD, RBD, SD1, N2R linker, as well as J o u r n a l P r e -p r o o f part of the SD2 subdomain, were not visible in the cryo-EM reconstruction ( Figure 3A) . This was similar to a state found in a mink-associated spike (Gobeil et al., 2021b) . The 3-RBD-down population classified into seven distinct subclasses (D5-D10) (Figure 3B ), of which one (D6) displayed an N2R configuration observed in the Omicron O1 Protomer 1 (Figure 3C) . A similar dislocation of the N2R region from its -strand arrangement with the SD2 subdomain was also found in a "down" protomer of a 1-RBD-up subclass (D12) (Figure 3C) , possibly representing an intermediate to the 2-RBD-up state. The array of distinct populations of the Delta S ectodomain that differ in their S1 subunit conformation are reminiscent of our observations with other naturally-occurring variants (Gobeil et al., 2021b , Gobeil et al., 2021a , while the appearance of the single S1-protomer disordered state (Figures 3A and 3D ) as we had observed in a mink-associated spike hints of S protein instability originating in the mobile S1 region encompassing the NTD, N2R linker, RBD and SD1.

We next investigated the source of the N2R linker rearrangements in the Delta and Omicron 3-RBD-down structures. The S1 subunit domain arrangements are responsive to one another across protomers through communication between adjacent, contacting RBDs, NTDs and subdomains. We have demonstrated through engineering (Henderson et al., 2020) and by examination of previous variants (Gobeil et al., 2021a , Gobeil et al., 2021b ) that this communication plays an essential role in RBD up/down state presentation. The close interaction between RBDs in the Omicron spike led us to ask whether a design engineered to lock the 3-RBD-down state, termed rS2d for the introduced RBD to S2 disulfide staple, would show a similar N2R linker rearrangement due to restrictions on RBD movement. We obtained cryo-EM reconstructions of an S2 stabilized HexaPro version of the rS2d design (rS2d-HexaPro) and for the unstabilized rS2d (Figure 4 ; Data S2; Table 3 ) (Henderson et al., 2020 . Three-dimensional classification of both datasets led to two prominent structural states (Figure 4) . One reconstruction in both datasets, each referred to as State 1, displayed a similar SD2 rearrangement as that observed in the Delta D6 and the Omicron O1 3-RBD-down structures ( Figure 4A ). Examination of a structural morph between both states (Video S1) identified a marked shift in the SD1 toward SD2 in the linker displaced protomer ( Figures 4B and 4C) . Because of the proximity between the two subdomains, the -sheet J o u r n a l P r e -p r o o f secondary structure linking the two is broken which in turn breaks the paired N2R -sheet structure ( Figure 4B ). The loss of this secondary structure permits the observed rearrangement in the linkers and the disordered segment.

We next asked whether the features in the rS2d spike domain arrangements that led to N2R linker rearrangement occurred in the Delta and Omicron 3-RBD-down states. Alignment of the SD2 subdomains of the N2R rearranged protomers indicated the overall subdomain and NTD domain arrangements were similar, except for the State 2 protomers ( Figure 4D) . Alignment of the disordered RBD (Protomer 1 ) revealed the RBD positions differed markedly among the trimers ( Figure 4D ). Using a vector-based quantification of S protein domain arrangement (Henderson et al., 2020 , Gobeil et al., 2021b , we previously found that absolute positions in spike domains may differ, but that trends in their overall architecture are correlated with important structural features such as the propensity to occupy the RBD-up state. The most important feature observed in the SD2-rearranged rS2d state was forcing of the SD1 toward its adjacent SD2.

We examined a series of vectors connecting the subdomains and NTDs of each protomer ( Figure 5A ) finding that, while the angle and distance values differ slightly, the trend in relative positions are the same for each of the S protein trimers where the N2R rearrangement was observed ( Figure 4E) . Specifically, the shift in SD1 position toward SD2 relative to the rS2d and rS2d-HexaPro State 2 is retained, consistent with a role of these subdomain shifts in causing the N2R rearranged state. These local effects were accompanied by a reduced distance between the RBD-NTD pair across from the disordered RBD protomer (RBD2 to NTD3) suggesting, despite variability in RBD positions, this interaction plays a role in the N2R rearrangement ( Figure 4E ).

Though the N2R rearranged trimers were similar, the non-rearranged State 2 trimers of the Omicron and rS2d trimers differed markedly. Together, comparison of the rS2d constructs and the variant structures indicates local and global rearrangements in S1 lead to the rearranged state.

We next examined clustering of the Omicron and Delta variant 3-RBD-down structures with previous variant structures utilizing sets of interprotomer and intraprotomer vectors using principal components analysis (PCA) ( Figure 5A Figure 5B ). The Delta structures were spread throughout the variant structure set along principal component one, consistent with the structural variability recovered through sub classification of the cryo-EM dataset (Figure 3) . These results demonstrate the considerable structural rearrangements that the S protein has acquired through SARS-CoV-2 evolution ( Figure 5B) . The structural conservation of the rearranged N2R state between the engineered and variant S proteins suggests this plays a specific role in S1 dynamics. As a specific domain organization accompanies the appearance of this state, we asked whether previous variant structures may present this state, albeit at lower proportions of the total population and with less conspicuous map density. We therefore examined the intraprotomer vector set by PCA to identify candidate structures ( Figure 5A and 5C). Consistent with the interprotomer vectors, N2R rearranged protomers of the rS2d constructs and the Omicron and Delta variants occupied a distinct position along principal component one ( Figure 5C ).

Protomers from the D614G, Mink, and Beta variants clustered with these structures. Densities in the N2R region for these structures were consistent with the rearranged state as determined by fitting of the rS2d State 1 coordinates, where this N2R rearranged state was particularly wellresolved, to the respective cryo-EM densities ( Figure 5D ). As expected, the densities were less clear than in the rS2d constructs, suggestive of multistate or dynamic behavior. As was observed in the rS2d, Omicron and Delta structures, the rearranged N2R protomer contained the disordered RBD and corresponded to the most distant RBD-NTD pair. These results highlight the differences in the Omicron 3-RBD-down structures from other variant S proteins, characterized by the stabilization of a N2R rearranged state.

We next studied the antigenicity and receptor binding properties of the Omicron S protein. Consistent with the extensive immune escape observed with the Omicron variant , Planas et al., 2021 , Mannar et al., 2022 , Hoffmann et al., 2021 , we found that its S protein lost binding to several SARS-CoV-2 neutralizing Abs ( Figures 6A-C and S3) . Ab DH1050.1 that targets a site of vulnerability in the NTD (McCallum et al., 2021b , no longer bound the Delta and Omicron S proteins (Figures 6A   and 6B ). The non-neutralizing, protective Ab DH1052 retained binding to the Delta but not to J o u r n a l P r e -p r o o f the Omicron S protein, its binding likely disrupted by changes in the region spanning residues 211-215 of the Omicron S protein (Figures 6A and 6B ) . The RBD receptor binding site directed Abs DH1041 and DH1042 lost binding to Omicron S protein while DH1042 also lost binding to Delta S protein. Both Omicron and Delta S proteins retained similar binding levels to ACE2 (Figures 6A, 6C, and S3 ). DH1041 and DH1042 bind similar RBD epitopes overlapping the ACE2 binding site (Figures 6C and S4 ; Data S5, Table 4 ) ,

showing that subtle changes in epitope footprint can alter the susceptibility profile of Abs to residue substitutions. For DH1042, insertion of a charged residue within a hydrophobic binding site by the L452R substitution in the Delta S protein ( Figure S4 ) results in reduction of binding and loss in neutralization activity. Indeed, DH1042 showed substantially reduced binding to S-GSAS-L452R and the S-GSAS-Epsilon S protein ectodomain (B.1.429) that harbors the L452R substitution. DH1041 binding, on the other hand, is unaffected by the L452R substitution ( Figure S4 ).

Ab S309 (Pinto et al., 2020) , the parental form of the engineered therapeutic Ab Sotrovimab, retained substantial binding to the Omicron spike ( Figures 6A and S3) . The broad sarbecovirus neutralizing Ab DH1047 lost substantial binding to the Omicron spike, resulting in loss in its neutralization activity against Omicron .

The S2X259 Ab targets a similar epitope (Tortorici et al., 2021) but retained binding and neutralization activity against Omicron . The SARS-CoV-2 cross-reactive Ab CR3022 targets a cryptic, unmutated epitope on the spike and retains binding to the S-GSAS-Omicron and Delta spikes despite considerable differences in their S1 subunit dynamics. Two RBD-directed Abs, DH1044 and DH1193, that do not compete for ACE2 binding ) retained binding to the Omicron S protein . The DH1044 epitope was mapped by NSEM to a region adjacent to the epitope of Ab S309, although shifted towards residue L452 making it susceptible to the L452R substitution in the Delta spike ( Figures 6A and 6C ) .

Mapping of the DH1193 epitope by NSEM revealed an epitope in between the S309 and DH1044 epitopes ( Figure S5 ). Consistent with their binding to the Omicron S protein, Abs DH1044 and DH1193 neutralized SARS-CoV-2 D614G and Omicron in a pseudovirus neutralization assay (Data S6).

We next probed S2 subunit conformation by measuring binding to S2 targeting Abs ( Figures 6D, 7A , S3, S6 and S7). We have previously described the binding of HIV-1 neutralizing, Fab-dimerized glycan-reactive (FDG) Ab 2G12 to a quaternary glycan cluster in the S2 subunit of the SARS-CoV-2 S protein (Williams et al., 2021) , and have demonstrated that 2G12 binding is sensitive to changes in S protein conformation , Gobeil et al., 2021b . 2G12 and a panel of FDG Abs targeting the same glycan cluster (Williams et al., 2021) showed glycan-dependent binding to the Delta and Omicron S proteins (Figures 6D and   S7 ). Binding of 2G12 to the Omicron S protein was weaker than its binding to the G614 and Delta S proteins, suggesting altered presentation of the glycan cluster, either due to a global change in S2 conformation or a local change due to the stabilization of glycan 709 by the D796Y substitution resulting in a change in presentation of the glycan epitope ( Figure 6D ). These results are consistent with considerable Omicron S protein structural shifts and suggest that S2 dynamics and flexibility are impacted by its acquired mutations.

We next tested binding by ELISA of the G614, Delta and Omicron spikes to SARS-CoV-2 fusion peptide (FP)-directed Ab DH1058 ( Figure 7A ) . DH1058 binds a 25residue peptide spanning residues 808-833 that includes the FP (Gobeil et al., 2021b) , and showed ~6-fold increased binding to S-GSAS-Omicron compared to the G614 and Delta S proteins, suggesting greater access of DH1058 to the FP in the Omicron S protein ( Figure 7A ). S-GSAS-2P-Omicron binding to DH1058 was substantially reduced compared to S-GSAS-Omicron suggesting reduced accessibility of the fusion peptide epitope when the 2P mutations were incorporated in the Omicron S protein. Binding rate and equilibrium constants (kon, koff and KD) measured by SPR revealed no differences between S-GSAS-Omicron and the corresponding D614G and Delta S protein constructs ( Figure 7B ). As the ELISA assay measures binding on a timescale slower than captured by the SPR assay, these results suggested time-dependent changes in the accessibility or presentation of the DH1058 epitope as the Ab is incubated with the spike for longer times in an ELISA experiment. The FP residues that are targeted by DH1058 were wellresolved in the cryo-EM reconstruction of the Omicron S protein ( Figure 7C ) with more residues resolved in the cryo-EM reconstruction compared to other variant S proteins. The overall orientation of the FP was conserved between D614G, Alpha, Beta, Delta and Omicron S protein J o u r n a l P r e -p r o o f structures ( Figure 7C ). While several attempts to obtain cryo-EM structures of DH1058 bound to furin-cleaved and uncleaved spikes were unsuccessful, we obtained a crystal structure of DH1058 Fab bound to a peptide comprising FP residues 808 to 833 at a resolution of 2.15 Å (P 1 21 1 space group) ( Figure 7D and Table 5 ). The interaction between DH1058 and the FP is mediated by all heavy chain (HC) complementary determining regions (CDRs). The portion of the fusion peptide between residues 816 and 825 defines the interaction with Ser816 and Asp820 sidechains forming hydrogen bonds (H-bond) with the HCDR2 residues Y53, E54, R56 and N57 side chains. The HCDR1 D31 main chain carbonyl formed an H-bond with the FP R815 sidechain, while the HCDR3 Y115 and Y116 formed sidechain-to-sidechain H-bonds with E819 and K825 respectively ( Figure 7D) . Aligning the structure of the trimeric pre-fusion Omicron S protein using the FP fragment from the crystal structure for superposition revealed clashes between the SD2 subdomain and bound DH1058 HC loops 13-17, 61-68 and 84-88, and the S2 subunit HR1 subdomain and DH1058 HCDR2 ( Figure 7E) . These data show that the binding of DH1058 to the SARS-CoV-2 FP, as revealed by the crystal structure, is incompatible with the structures of the pre-fusion SARS-CoV-2 S proteins. Taken together, these data suggest that a weak initial contact of the DH1058 with the FP is made in the pre-fusion S protein ectodomains as captured in the SPR assay, followed by a conformational change in the spike leading to greater FP exposure and stable binding of the DH1058 Fab. DH1058 Fab cannot bind stably to the pre-fusion conformation of the SARS-CoV-2 S protein, consistent with its lack of SARS-CoV-2 neutralization activity and would require greater exposure of the FP to make a stable interaction avoiding clashes with adjacent regions of the pre-fusion S protein. Our data here suggest that the conformational changes leading to greater FP exposure occur more readily in the Omicron S protein. Taken together, these results show altered flexibility and ease of exposure and release around the fusion peptide region in the S2 subunit of the Omicron spike relative to other variants.

As SARS-CoV-2 continues to evolve, the emergence of the Omicron variant is poised to change the course of the COVID-19 pandemic with its unprecedented transmissibility and immune evasion. The Omicron spike protein, that is central to defining these properties, is riddled with mutations in both its receptor binding S1 subunit, and its fusion subunit S2. As the exposure of Ab and receptor binding sites can be affected by both direct substitutions at the binding interface, J o u r n a l P r e -p r o o f and conformational masking of key sites, we have sought here to understand the conformational changes of the Omicron spike resulting from its altered primary sequence. Our structural studies were performed in our previously established platform S-GSAS that did not contain any extraneous stabilizing mutations in the S2 subunit, thus aiming to visualize S protein conformations in a more native format (Gobeil et al., 2021a , Gobeil et al., 2021b . Indeed, we were able to resolve a more varied repertoire of structural states of the Omicron S protein than were revealed by several studies that have used constructs with stabilizing Proline mutations in the S2 subunit , Cerutti, 2021 , Ni, 2021 . The Omicron S protein presented a substantially different domain organization compared to other variants (Figure 1 ), differences that we were able to visualize in our structures and quantify using sets of intra-and inter-protomer vectors (Figures 1, 4 and 5) . We found a tightly packed RBD-RBD interface in by binding to a fusion peptide directed Ab. Thus, the increased transmissibility of the Omicron spike may be facilitated by a combined effect of the ease of accessing the RBD-up state despite stabilization of the down-state RBD-RBD interface, retained affinity for ACE2 interactions despite the large number of RBD mutations, as well as by more ready release of the FP. Further functional studies will be needed to determine the extent to which the structural observations reported here impact the biological properties of the Omicron variant. Close monitoring of the continued evolution of the structure of future variants on the Omicron template will be required to achieve a deep understanding of Omicron pathobiology, and to anticipate the immune escape potential of the further evolved variants.

The data reported in this study have been obtained using an engineered SARS-CoV-2 spike ectodomain construct where the furin site has been mutated rendering this construct resistant to protease cleavage at this site. A recent preprint reporting cryo-EM structures of a fulllength Omicron spike construct show similar RBD-RBD interactions in the 3-RBD-down state as we have observed here, thus further supporting our spike ectodomain construct as being representative of the full-length spike (Zhang, 2022) . Structural and spectroscopic studies performed on the SARS-CoV-2 spike in the context of the native virion were also found consistent with spike populations observed by single particle cryo-EM , Ke et al., 2020 . Similar studies performed on the Omicron spike will be valuable to bridge the conformational states observed in our structural studies with the spike conformations adopted in the virion context. Figure S1 , S2, Data S1, S2, and Table 1 . Figure S1 , S2, Data S1, S2, and Table 1. The N2R linker in one of the Delta 3-RBD-down classes (D6) showed a N2R region that was dislodged from its beta strand arrangement with the SD2 subdomain, and is shown in red. Right.

Overlay of 1-RBD-up subclasses. Dashed rectangle indicates the N2R region zoomed in on the right image. The N2R linker in one of the two "down" protomers of a 1-RBD-up structure (D12) J o u r n a l P r e -p r o o f showed a N2R region that was dislodged from its beta strand arrangement with the SD2 subdomain, and is shown in red. Mutations in the Delta variant are shown as red spheres. D. M1 state with the S1 subunit and SD2 subdomain of one of the three protomers disordered. The cryo-EM density is shown as a blue mesh with the underlying fitted model colored by protomer. The protomer colored orange shows disorder in the S1 subunit and SD2 subdomain, therefore, these regions could not be built in this protomer. See also Figure S1 , Data S1, S3 and Table 2 . between the rS2d-HexaPro states 1 and 2, rS2d states 1 and 2, Omicron, and Delta D6 trimers.

Phi angles 3, 6, and 9 and Theta angles 2, 4, and 6 correspond to dihedrals/angles in protomers 3, 1, and 2, respectively. See also Data S4 and Table 3 . 

CoV-2 S protein (PDB 7KDK). The S protein is shown as surface with 2 protomers colored gray and the protomer used for alignment with the crystal structure of DH1058-fusion peptide complex has its S1 subunits colored pale green (NTD), cyan (N2R linker), red (RBD) and dark blue and orange (SD1 and SD2 respectively). The inset panel shows a clash between DH1058 HC and the SARS-CoV-2 SD2 subdomain. A clash is also observed for between the HC and the 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Priyamvada Acharya (priyamvada.acharya@duke.edu).

Materials Availability Further information and requests for resources and reagents should be directed to Priyamvada Acharya (priyamvada.acharya@duke.edu). Plasmids generated in this study have been deposited to Addgene under the codes 180423, 180593, 182575, 183515, 183516. Data and Code Availability • Cryo-EM reconstructions and atomic models generated during this study are available at wwPDB and EMBD ( 7THT and 7THE, and EMDB IDs 25865, 25983, 25846, 25984, 25880, 26038, 26040, 26041, 26042, 26043, 26045, 26046, 26039, 26047, 26048, 26049, 26050, 26051, 26052, 26053, 26055, 26059, 25985, 25986, 25987, 25988, 25904 and 25893 . The crystal structure of DH1058 Fab bound to the SARS-CoV-2 fusion peptide is deposited at wwPDB with PDB ID 7TOW. Additional Supplemental Items are available from Mendeley Data at https://data.mendeley.com/datasets/kxjf56yy6d/1 • This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Cell culture Gibco FreeStyle 293-F cells (embryonal, human kidney) were incubated at 37°C and 9% CO2 in a humidified atmosphere. Cells were incubated in FreeStyle 293 Expression Medium (Gibco) with agitation at 120 rpm. Plasmids were transiently transfected into cells using Turbo293 (SpeedBiosystems) and incubated at 37 °C, 9% CO2, 120 rpm for 6 days. On the day following transfection, HyClone CDM4HEK293 media (Cytiva, MA) was added to the cells. Antibodies were produced in Expi293 cells (embryonal, human kidney). Cells were incubated in Expi293 Expression Medium at 37°C, 120 rpm and 8% CO2 in a humidified atmosphere. Plasmids were transiently transfected into cells using the ExpiFectamine 293 Transfection Kit and protocol (Gibco).

Mutagenesis of all plasmids generated by this study were performed and the sequence confirmed by GeneImmune Biotechnology (Rockville, MD). The SARS-CoV-2 spike protein ectodomain constructs comprised the S protein residues 1 to 1208 (GenBank: MN908947) with the D614G mutation, the furin cleavage site (RRAR; residue 682-685) mutated to GSAS, a C-terminal T4 fibritin trimerization motif, a C-terminal HRV3C protease cleavage site, a TwinStrepTag and an 8XHisTag. All spike ectodomains were cloned into the mammalian expression vector pαH and have been deposited to Addgene (https://www.addgene.org).

Protein purification Spike ectodomains were harvested from the concentrated supernatant on the 6 th day post transfection. The spike ectodomains were purified using StrepTactin resin (IBA LifeSciences) and size exclusion chromatography (SEC) using a Superose 6 10/300 GL Increase column (Cytiva, MA) equilibrated in 2mM Tris, pH 8.0, 200 mM NaCl, 0.02% NaN3. All steps of the purification were performed at room temperature within a single day. Protein quality was assessed by SDS-PAGE using NuPage 4-12% (Invitrogen, CA). The purified proteins were flash frozen and stored at -80°C in single-use aliquots. Each aliquot was thawed by a 20-minute incubation at 37 °C before use. Antibodies were produced in Expi293F cells and purified by Protein A affinity and digested using LysC to generate Fab fragments. ACE2 with human Fc tag was purified by Protein A affinity chromatography and SEC.

Negative-stain electron microscopy Fab-spike complex of DH1044 was generated by mixing 5.9 µg of spike with 6.2 µg of Fab in ~25 µl of HEPES-buffered saline (HBS) containing 20 mM HEPES and 150 mM NaCl, pH 7.4, and incubating at 37 °C for 1 hr and using sample directly for negative stain without further purification. Fab-spike complex of DH1193 was generated by mixing 20 µg of spike with 28 µg of Fab in ~200 µl of phosphate-buffered saline and incubating 1 hr at 37 °C. Sample was then brought to room temperature and diluted with 800 µl HBS, mixed, and then diluted with HBS augmented with 16 mM glutaraldehyde (Electron Microscopy Sciences, PA), fixed for 5 min, quenched by addition of 40 µl of 1 M Tris buffer, pH 7.4, and concentrated in a 2-ml 100-kDa MCWO Amicon centrifugal concentrator by spinning 10 min at 4000 rpm in a Sorvall benchtop centrifuge with a swinging-bucket rotor, yielding a final volume of ~75 µl. Protein concentration was measured using a Nanodrop which reported a nominal spike concentration of 0.6 mg/ml. For negative stain, samples were diluted to 0.1 mg/ml with HBS augmented with 5 g/dl glycerol and 8 mM glutaraldehyde. After 5 min incubation, excess glutaraldehyde was quenched by adding sufficient 1 M Tris stock, pH 7.4, to give 75 mM final Tris concentration and incubated for 5 min. Quenched sample was applied to a glow-discharged carbon-coated EM grid (Electron Microscopy Sciences, PA, CF300-Cu) for 10-12 second, then blotted, and stained with 2 g/dL uranyl formate (Electron Microscopy Sciences, PA), for 1 min, blotted and air-dried. Grids were examined on a Philips EM420 electron microscope operating at 120 kV and images were collected at a nominal magnification of 82,000x on a 4 Mpix CCD camera at 4.02 Å/pixel for DH1044 data, or at 49,000x on a 76 Mpix CCD camera at 2.4 Å/pixel for DH1193 data. Images were analyzed and 3D reconstructions generated using standard protocols with Relion 3.0 (Zivanov et al., 2018) .

Differential scanning fluorimetry DSF assay was performed using Tycho NT. 6 (NanoTemper Technologies). Spike ectodomains were diluted to approximatively 0.15 mg/ml. Intrinsic fluorescence was measured at 330 nm and 350 nm while the sample was heated from 35 to 95 °C at a rate of 30°C/min. The ratio of fluorescence (350/330 nm) and inflection temperatures (Ti) were calculated by the Tycho NT. 6 apparatus.

ELISA assays Spike ectodomains tested for antibody-or ACE2-binding in ELISA assays as previously described . Assays were run in two formats i.e., antibodies/ACE2 coated, or spike coated. For the first format, the assay was performed on 384-well plates coated at 2 µg/ml overnight at 4°C, washed, blocked and followed by two-fold serially diluted spike protein starting at 25 µg/mL. Binding was detected with polyclonal anti-SARS-CoV-2 spike rabbit serum (developed in our lab), followed by goat anti-rabbit-HRP (Abcam, Ab97080) and TMB substrate (Sera Care Life Sciences, MA). Absorbance was read at 450 nm. In the second format, serially diluted spike protein was bound in wells of a 384-well plates, which were previously coated with streptavidin (Thermo Fisher Scientific, MA) at 2 µg/mL and blocked. Proteins were incubated at room temperature for 1 hour, washed, then human mAbs were added at 10 µg/ml. Antibodies were incubated at room temperature for 1 hour, washed and binding detected with goat anti-human-HRP (Jackson ImmunoResearch Laboratories, PA) and TMB substrate. Recombinant FDG mAbs were tested for binding to the SARS-CoV-2 Omicron spike (S-GSAS-Omicron, Lot: 486KJ), Delta spike (S-GSAS/B.1.617.2.v1, Lot: 076XH), SARS-CoV-2 spike (nCoV-1_2 ProRev+D614G, Lot: 001AM), in ELISA in the absence or presence of single monomer D-mannose as previously described (PMID: 34019795). Briefly, spike proteins (20ng) were captured by streptavidin (30ng per well) to individual wells of a 384-well Nunc-absorb ELISA plates using PBS-based buffers and assay conditions as previously described (PMID: 34019795; PMID: 28298421; PMID: 28298420). Commercially obtained D-mannose (Sigma, St. Louis, MO) was used to outcompete mAb binding to glycans on the spike proteins; D-mannose solutions were also produced in ELISA PBS-based glycan buffers at a concentration of [1M] Dmannose as described (PMID: 34019795). Mouse anti-monkey IgG-HRP (Southern Biotech, CAT# 4700-05) and Goat anti-human IgG-HRP (Jackson ImmunoResearch Laboratories, CAT# 109-035-098) secondary antibodies were used to detect antibody bound to the spike proteins. HRP detection was subsequently quantified with 3,30,5,50-tetramethylbenzidine (TMB) by measuring binding levels at an absorbance of 450nm, and binding titers were also reported as Log area under the curve (AUC).

Surface Plasmon Resonance Antibody binding to SARS-CoV-2 spike was assessed using SPR on a Biacore T-200 (Cytiva, MA, formerly GE Healthcare) with HBS buffer supplemented with 3 mM EDTA and 0.05% surfactant P-20 (HBS-EP+, Cytiva, MA). All binding assays were performed at 25 °C. Spike variants were captured on a Series S Strepavidin (SA) chip (Cytiva, MA) coated at 100 nM (60s at 10µL/min). Fabs were injected at concentrations ranging from 0.625 nM to 800 nM (prepared in a 2-fold serial dilution manner) over the S proteins using the single cycle kinetics mode with 5 concentrations per cycle. The surface was regenerated after the last injection with 3 pulses of a 50mM NaoH + 1M NaCl solution for 10 seconds at 100µL/min. Sensogram data were analyzed using the BiaEvaluation software (Cytiva, MA) Cryo-EM Purified SARS-CoV-2 spike ectodomains were diluted to a concentration of ~1.5 mg/mL in 2 mM Tris pH 8.0, 200 mM NaCl and 0.02% NaN3 and 0.5% glycerol was added. A 2.3-µL drop of protein was deposited on a Quantifoil-1.2/1.3 grid (Electron Microscopy Sciences, PA) that had been glow discharged for 10 seconds using a PELCO easiGlow™ Glow Discharge Cleaning System. After a 30-second incubation in >95% humidity, excess protein was blotted away for 2.5 seconds before being plunge frozen into liquid ethane using a Leica EM GP2 plunge freezer (Leica Microsystems). Frozen grids were imaged using a Titan Krios (Thermo Fisher) equipped with a K3 detector (Gatan). The cryoSPARC (Punjani et al., 2017) software was used for data processing. Phenix (Liebschner et al., 2019 , Afonine et al., 2018 , Coot (Emsley et al., 2010) , Pymol (Schrodinger, 2015) , Chimera (Pettersen et al., 2004) , ChimeraX (Goddard et al., 2018) and Isolde (Croll, 2018) were used for model building and refinement.

Vector Based Structure Analysis.

Vector analysis of intraprotomer domain positions was performed as described previously using the Visual Molecular Dynamics (VMD) (Humphrey et al., 1996) software package Tcl interface. For each protomer of each structure, Cα centroids were determined for the NTD (residues 27 to 69, 80 to 130, 168 to 172, 187 to 209, 216 to 242, and 263 to 271), NTD′ (residues 44 to 53 and 272 to 293), RBD (residues 334 to 378, 389 to 443, and 503 to 521), SD1 (residues 323 to 329 and 529 to 590), SD2 (residues 294 to 322, 591 to 620, 641 to 691, and 692 to 696), CD (residues 711 to 716 1072 to 1121), and a S2 sheet motif (S2s; residues 717 to 727 and 1047 to 1071). Additional centroids for the NTD (NTDc; residues 116 to 129 and 169 to 172) and RBD (RBDc; residues 403 to 410) were determined for use as reference points for monitoring the relative NTD and RBD orientations to the NTD′ and SD1, respectively. Vectors were calculated between the following within protomer centroids: NTD to NTD′, NTD′ to SD2, SD2 to SD1, SD2 to CD, SD1 to RBD, CD to S2s, NTDc to NTD, RBD to RBDc. Vector magnitudes, angles, and dihedrals were determined from these vectors and centroids. Inter-protomer domain vector calculations for the SD2, SD1, and NTD′ used these centroids in addition to anchor residue Cα positions for each domain including SD2 residue 671 (SD2a), SD1 residue 575 (SD1a), and NTD′ residue 276 (NTD′a). These were selected based upon visualization of position variation in all protomers used in this analysis via alignment of all of each domain in PyMol (Schrodinger, 2015) . Vectors were calculated for the following: NTD′ to NTD′r, NTD′ to SD2, SD2 to SD2r, SD2 to SD1, SD1 to SD1r, and SD1 to NTD′. Angles and dihedrals were determined from these vectors and centroids. Vectors for the RBD to adjacent RBD and RBD to adjacent NTD were calculated using the above RBD, NTD, and RBDc centroids. Vectors were calculated for the following: RBD2 to RBD1, RBD3 to RBD2, and RBD3 to RBD1. Angles and dihedrals were determined from these vectors and centroids. Principal components analysis, K-means clustering, and Pearson correlation (confidence interval 0.95, p<0.05) analysis of vectors sets was performed in R(Team, 2017). Data were centered and scaled for the PCA analyses. Principal components analysis, K-means clustering, and Pearson correlation (confidence interval 0.95, p < 0.05) analysis of vectors sets was performed in R. Data were centered and scaled for the PCA analyses. Difference distance matrices (DDM). DDM were generated using the Bio3D package (Grant et al., 2020) implemented in R (R Core Team (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/)

Pseudovirus Neutralization Assay. The pseudovirus neutralization assay performed at Duke has been described in detail (Gilbert et al., 2022) and is a formally validated adaptation of the assay utilized by the Vaccine Research Center; the Duke assay is FDA approved for D614G. For measurements of neutralization, pseudovirus was incubated with 8 serial 5-fold dilutions of antibody samples (1:20 starting dilution using antibodies diluted to 1.0 mg/ml) in duplicate in a total volume of 150 µl for 1 hr at 37°C in 96-well flat-bottom culture plates. 293T/ACE2-MF cells were detached from T75 culture flasks using TrypLE Select Enzyme solution, suspended in growth medium (100,000 cells/ml) and immediately added to all wells (10,000 cells in 100 µL of growth medium per well). One set of 8 wells received cells + virus (virus control) and another set of 8 wells received cells only (background control). After 71-73 hrs of incubation, medium was removed by gentle aspiration and 30 µl of Promega 1X lysis buffer was added to all wells. After a 10-minute incubation at room temperature, 100 µl of Bright-Glo luciferase reagent was added to all wells. After 1-2 minutes, 110 µl of the cell lysate was transferred to a black/white plate. Luminescence was measured using a GloMax Navigator luminometer (Promega). Neutralization titers are the inhibitory dilution (ID) of serum samples at which RLUs were reduced by 50% (ID50) compared to virus control wells after subtraction of background RLUs. Serum samples were heatinactivated for 30 minutes at 56°C prior to assay.

X-ray crystallography After size exclusion purification, DH1058 fab was concentrated to 26 mg/mL. The fusion peptide fragment was solubilized in PBS + 10% DMSO at a concentration of 10 mg/mL. The protein and the peptide were mixed at a Fab:peptide molar ratio of 1:2. Crystals were grown in 20% PEG3000, 100mM Tris base/HCl pH 7.0, 200mM calcium acetate at 22°C in a sitting drop vapor diffusion setting using a drop ratio of 0.4 µL protein : 0.2 µL reservoir solution. Large UV-active plate shaped crystals were observed after 24 hours. A single crystal was cryopreserved directly from the drop. Diffraction data was collected at the Advanced Photon Source using sector 22ID beamline. The collected diffraction images were indexed, integrated, and scaled using HKL2000 (Otwinowski and Minor, 1997 ) Initial phases were calculated by molecular replacement using Phenix.PHASER (Adams et al., 2010 , McCoy et al., 2007 and the PDB 5GGU (Crystal structure of tremelimumab Fab) as a search model. Iterative rounds of manual model building using Coot (Emsley et al., 2010) and automatic refinement in PHENIX (Liebschner et al., 2019 , Afonine et al., 2018 were performed. Data collection and refinement statistics are summarized in Table 5 . The refined structure has been deposited to the Protein Data Bank (http://www.pdb.org) under the accession code 7TOW.

No statistical analyses were performed in this study.

Video S1: Morph between rS2d-HexaPro States 1 and 2.

Structure S2 subunits were aligned for comparison of S1 domain movements. Promoters 1-3 are colored are green, purple, and cyan respectively. Protomer 1 contains the N2R rearranged site. The initial frame corresponds to State 2. Related to Figure 5 .

PHENIX: a comprehensive Python-based system for macromolecular structure solution

Real-space refinement inPHENIXfor cryo-EM and crystallography

Structural basis for enhanced infectivity and immune evasion of SARS-CoV-2 variants

Structural basis for antibody resistance to SARS-CoV-2 omicron variant. bioRxiv

ISOLDE: a physically realistic environment for model building into lowresolution electron-density maps

Cold sensitivity of the SARS-CoV-2 spike ectodomain

Features and development ofCoot

D614G Mutation Alters SARS-CoV-2 Spike Conformation and Enhances Protease Cleavage at the S1/S2 Junction

Effect of natural mutations of SARS-CoV-2 on spike structure, conformation, and antigenicity. Science

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

The Bio3D packages for structural bioinformatics

Controlling the SARS-CoV-2 spike glycoprotein conformation

The Omicron variant is highly resistant against antibody-mediated neutralization

Structure-based design of prefusion-stabilized SARS-CoV-2 spikes

VMD: Visual molecular dynamics

Structures and distributions of SARS-CoV-2 spike proteins on intact virions

vitro and in vivo functions of SARS-CoV-2 infection-enhancing and neutralizing antibodies

Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix

Real-Time Conformational Dynamics of SARS-CoV-2 Spikes on Virus Particles

SARS-CoV-2 Omicron variant: Antibody evasion and cryo-EM structure of spike protein-ACE2 complex

A broadly cross-reactive antibody neutralizes and protects against sarbecovirus challenge in mice

N-terminal domain antigenic mapping reveals a site of vulnerability for SARS-CoV-2. Cell

Molecular basis of immune evasion by the Delta and Kappa SARS-CoV-2 variants

Phaser crystallographic software

Structural analysis of the Spike of the Omicron SARS-COV-2 variant by cryo-EM and implications for immune evasion

Processing of X-ray diffraction data collected in oscillation mode

UCSF Chimera?A visualization system for exploratory research and analysis

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

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

Plasma Neutralization of the SARS-CoV-2 Omicron Variant

The PyMOL Molecular Graphics System. TEAM, R. C. 2017. R: A Language and Environment for Statistical Computing

Broad sarbecovirus neutralization by a human monoclonal antibody

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Fabdimerized glycan-reactive antibodies are a structural category of natural antibodies

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

SARS-CoV-2 Variants Increase Kinetic Stability of Open Spike Conformations as an Evolutionary Strategy. mBio

A highly conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

Membrane fusion and immune evasion by the spike protein of SARS-CoV-2 Delta variant

Structural and functional impact by SARS-CoV-2 Omicron spike mutations bioRxiv

New tools for automated high-resolution cryo-EM structure determination in RELION-3. Elife

• Omicron S architecture differs from Delta and other variants. • Tight packing of Omicron S RBDs results in unique up and down state arrangements

• 3-RBD-down Omicron S stabilizes a rearrangement of the NTD-to-RBD (N2R) linker

• S2 subunit conformational changes lead to altered fusion peptide dynamics. eTOC Blurb Gobeil, Henderson, Stalls, et al. identify diverse Omicron S ectodomain conformations demonstrating altered architecture exhibiting tight packing of the 3-RBD-down state, NTD-to-RBD (N2R) linker rearrangements, and changes in fusion peptide conformational dynamics. These distinct conformational features of its S protein may underlie Omicron's higher transmissibility and immune evasion

J o u r n a l P r e -p r o o f REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies ACE2 (Henderson et al., 2020) N/A CR3022 (Henderson et al., 2020) N/A 2G12 (Williams et al., 2021) N/A Goat anti-rabbit-HRP Abcam ab97080 Goat anti-human-HRP Jackson ImmunoResearch Laboratories 109-035-098Mouse anti-monkey IgG HRP Southern Biotech Cat #:4700-05; RRID:AB_2796069 DH1050.1 N/A DH1052 N/A S309 (Pinto et al., 2020) N/A DH1041 N/A DH1042 N/A DH1047 N/A S2X259 (Totorici et al., 2021) N/A DH1044 N/A DH1193 N/A DH1058 N/A