Structures of the Omicron Spike trimer with ACE2 and an anti-Omicron antibody

The SARS-CoV-2 Omicron variant has become the dominant infective strain. We report the structures of the Omicron spike trimer on its own or in complex with ACE2 or an anti-Omicron antibody. Most Omicron mutations are located on the surface of the spike protein, which change binding epitopes to many current antibodies. In the ACE2 binding site, compensating mutations strengthen RBD binding to ACE2. Both the RBD and the apo form of the Omicron spike trimer are thermodynamically unstable. An unusual RBD-RBD interaction in the ACE2-spike complex supports the open conformation and further reinforces ACE2 binding to the spike trimer. A broad-spectrum therapeutic antibody, JMB2002, which has completed a Phase 1 clinical trial, maintains neutralizing activity against Omicron. JMB2002 binds to RBD differently from other characterized antibodies and inhibits ACE2 binding.

T he Omicron variant of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative virus of COVID-19, was initially reported in South Africa in November 2021 and quickly became the dominant strain worldwide (1). Phylogenetic tree analyses revealed that Omicron evolved independently from previous variants of concerns (VOCs), including the predominant Alpha, Beta, Gamma, and Delta variants (Fig. 1A) (2)(3)(4)(5). Compared with the original wild-type (WT) strain of SARS-CoV-2, Omicron has 60 amino acid mutations, of which 37 are in the spike protein, the target of most COVID-19 vaccines and therapeutic antibodies (Fig. 1B). This high variation is reflected in different behavior, with the Omicron variant showing enhanced transmission, antibody evasion, and vaccine resistance (6)(7)(8).
To study the mechanism for Omicron's enhanced transmission, we first biochemically characterized the interactions of the SARS-CoV-2 receptor angiotensin-converting enzyme 2 (ACE2) with the trimer of the spike extracellular domain from Omicron and the original WT strain, both of which contain proline substitutions (2P or 6P) and a mutated furin cleavage site to stabilize the prefusion conformation (9,10). Monomeric human ACE2 bound to Omicron trimeric spike protein with approximately sixfold higher affinity [dissociation constant (K D ) = 2.5 ± 0.6 nM] than the WT spike trimer (K D = 14.7 ± 4.9 nM). The dimeric human ACE2 bound to Omicron spike trimer (K D = 0.3 ± 0.2 nM) with approximately ninefold higher avidity than WT (K D = 2.7 ± 1.4 nM) (Fig. 1, C and D). We then studied the interactions of ACE2 with the monomeric receptor binding domain (RBD) from the Omicron and WT strains. Monomeric human ACE2 bound to immobilized Omicron RBD (K D = 38.9 ± 10.5 nM) with approximately twofold higher affinity than WT RBD (K D = 75.5 ± 2.1 nM) (Fig. 1, C and D). The enhanced interaction of Omicron spike and RBD proteins with human ACE2 is consistent with previously published data (11) and may contribute to the increased infectivity of the Omicron variant.
To determine the structural basis of the higher affinity of the Omicron spike trimer for ACE2, we solved the structure of the ACE2-Omicron spike trimer complex at a global resolution of 2.77 Å (table S1). Despite an excess of ACE2 (molar ratio of 3.2:1 ACE2 to spike trimer; fig. S1A), we only observed strong density for one ACE2 bound to one RBD from the spike trimer in the open "up" conformation ( Fig. 2A and fig. S2). The other two RBDs, with clear density, are in the closed "down" conformation. Particle classification revealed that most of the picked particles (~70%) do not have ACE2 bound. We also determined the structure of this apo Omicron spike trimer at a global resolution of 2.56 Å ( fig. S2 and table S1). All three RBDs are in the closed-down conformation and are less visible in the high-resolution map (2.56 Å; fig. S3A) yet become more visible in lower-resolution maps (4.5 and 6.5 Å; fig. S3, B and C). This contrasts with the clear visibility of the three RBDs in the ACE2-Omicron spike complex in a high-resolution map (2.56 Å; Fig. 2A), indicating that the RBD in the apo form is more dynamic, and ACE2 binding likely stabilizes the conformation of the three RBDs. Thermal shift assays at pH 7.4 revealed that the Omicron and WT RBD have single melting temperatures (T m s) of 45.7°a nd 51.0°C, respectively (fig. S1C), indicating that the Omicron RBD is less stable than the WT RBD. By contrast, both the Omicron and WT spike trimer displayed two T m s ( fig. S1D), with the high T m corresponding to the dissociation of the spike trimer and the low T m corresponding to unfolding of the RBD. The T m profile of the WT spike trimer is similar to previous reports (10,12). T m 1 of both the Omicron and the WT spike trimer is similar to the respective T m for the isolated RBD (fig. S1, C and D), indicating that the Omicron RBD within the context of the spike trimer remains less stable than the WT RBD. We further confirmed the highly flexible nature of the Omicron RBD by performing hydrogendeuterium exchange mass spectrometry (HDX), which showed that the Omicron spike trimer has an overall higher rate of HDX ( fig. S4), particularly in the RBD region, consistent with its lower thermal stability.
Mapping the 37 mutations onto the up protomer of the ACE2-bound spike trimer revealed that most mutations are located on the surface of the spike protein, with many of them in known epitopes of therapeutic antibodies (Fig. 2B). We grouped the surface mutations into three hotspots ( Fig. 2C and table S2). Eight mutations in the N-terminal domain (hotspot I) would affect the structures of the epitopes for a number of antibodies; for example, D143-145 would remove the epitope for the 4A8 antibody (13). Fifteen mutations are in the RBD, which contains the ACE2binding site as well as the epitopes for 90% of antibodies induced by infection or vaccination. Ten of these mutations are in the RBM (hotspot II) and five are near the core structure domain (hotspot III) (Fig. 2C). Hotspot II encompasses the epitopes for therapeutic antibodies AZD1061, REGN10987, and REGN10933, and hotspot III overlaps the epitope for LY-CoV555 ( Fig. 2B) (14)(15)(16).
Local refinement of the ACE2-RBD region produced a high-quality map at 2.57-Å resolution, which allowed unambiguous building of the ACE2-RBD complex (Fig. 3A, table S1, and fig. S2). Although their RBDs differ at 15 residues, the overall structure of the Omicron ACE2-RBD complex is similar to two high-resolution x-ray structures of the WT ACE2-RBD complex [PDB codes: 6LZG and 6M0J (17, 18)], with the Ca atoms of the whole RBD deviating by <0.4 Å ( Fig. 3B and table S3). We did see local differences at the ACE2-RBD interface: The Omicron RBD forms extra interactions with ACE2, including interactions from RBD mutations S477N, Q493R, Q496S, Q498R, and N501Y to ACE2 (Fig. 3C). In particular, the side chain of S477N forms two extra hydrogen bonds with S19 of ACE2, the Q498R mutation forms two additional hydrogen bonds with Q42 and D38 from ACE2, and the N501Y mutation forms extensive packing interactions with ACE2 residues Y41 and K353. These additional interactions may compensate for the loss of polar interactions between WT RBD and ACE2 caused by RBD mutations K417N and E484A (Fig. 3, C and D), consistent with the enhanced affinity of the Omicron RBD with ACE2 ( Fig. 1, C and D).
We observed RBD-RBD interactions from one of the two down RBDs to the up RBD (Fig.  3A), with the interface composed of residues A475 and F486 from the down RBD and residues L368, F374, and T385 from the up RBD ( Fig. 3A and fig. S5A). Structure comparison revealed that the RBD-RBD interface is not observed within the WT spike trimer because of movement of a loop (residues 368 to 374) caused by Omicron RBD mutations S371L, S373P, G339D, and S375F, which are in hotspot III ( Fig. 2C and fig. S5A). This RBD-RBD interaction may stabilize the up conformation of the RBD that promotes ACE2 binding. In addition, the Omicron mutations S371L, S373P, and S375F are located at the entrance to the fatty acid-binding pocket in the WT RBD (19), and these Omicron mutations distort the pocket ( fig. S5B), thus destabilizing the RBDs in the all closed-down conformation. Consistent with involvement of spike dynamics, ACE2 binds to the Omicron spike trimer with sixfold to ninefold higher avidity than to the WT spike trimer but binds to the Omicron RBD monomer with twofold higher affinity than to WT RBD (Fig. 1, C and D). We suggest that in addition to destabilization of RBDs in the closed conformation, the RBD-RBD interactions that stabilize one RBD in the open up conformation within the spike trimer may act together with the compensating mutations in the ACE2binding site to contribute to the higher affinity of Omicron, and this likely plays a role in its higher infectivity.
We previously discovered an antibody, JMB2002, that showed potent efficacy against   (20). JMB2002 has completed a phase 1 clinical trial in healthy donors and was shown to have excellent safety and pharmacokinetic properties and has been approved for a clinical trial in the United States (IND 154745). We evaluated the binding of JMB2002 to WT and Omicron spike trimers. JMB2002 Fab bound the Omicron spike trimer with approximately fourfold increased affinity (K D = 3.2 ± 3.0 nM) compared with the WT spike trimer (K D = 12.2 ± 11.6 nM), whereas JMB2002 IgG showed similar avidity for the Omicron spike trimer (K D = 0.4 ± 0.1 nM) and WT (K D = 0.5 ± 0.3 nM) (Fig. 4, A to D). Furthermore, JMB2002 was able to directly inhibit the binding of ACE2 to the Omicron spike trimer with a median inhibitory concentration of 1.8 nM (Fig. 4E). In pseudovirus neutralization assays, JMB2002 effectively blocked the entry of the Omicron pseudovirus into human ACE2-expressing cells in addition to blocking the WT pseudovirus ( Fig. 4F and  fig. S6A). JMB2002 was also able to neutralize a number of VOCs, including variants of Alpha, Beta, and Gamma, but not Delta ( fig.  S6, B to E).
To reveal the basis of JMB2002 inhibition of Omicron, we solved the structure of the Omicron spike trimer bound to a Fab from JMB2002 at a global resolution of 2.69 Å (Fig. 5A, table S1, and figs. S7 and S8). To stabilize the constant regions of Fab, we used a nanobody that binds to the interface between the variable and constant regions of the light chain (21). The cryo-electron microscopy (cryo-EM) density map reveals the binding of two Fab molecules to two RBDs (one up and one down) of the trimeric spike (Fig. 5,  A and B). The overall structure of the spike trimer in the Fab-bound complex is very similar to that of the ACE2-bound complex, with a root mean square difference of 1.0 Å over all Ca atoms of the spike trimer, including the unusual RBD-RBD dimer configuration ( fig. S9, A and B).
Within the Fab-spike trimer structure, both Fabs bind to the same region in their respective RBD (Fig. 5, C and D). Local refinement of the Fab-bound RBD structure generated a density map to a resolution of 2.47 Å (figs. S7G and S8D), which provides detailed interactions between Fab and RBD. The Fab-binding site does not overlap with the ACE2-binding site ( fig. S9C). However, in the context of the trimer, Fab binding to the down RBD would clash with ACE2 binding to the up RBD (Fig.  5E), consistent with direct inhibition of ACE2 binding to the Omicron spike trimer by JMB2002 (Fig. 4E).
Particle classification also revealed two additional antibody-bound complexes at a global resolution of 2.92 and 3.18 Å, respectively (figs. S7 and S8). One of the two complexes has the spike trimer with one up RBD bound to one Fab and two down RBDs in the apo state ( fig.  S8A). The other complex contains the spike trimer with two up RBDs and one down RBD, with each RBD bound to one Fab ( fig. S8C). The up-down RBD-RBD interactions within the spike trimer are conserved in these complexes. The diverse configuration and stoichiometry ratio of the spike trimer bound to the antibody further highlight the conformation flexibility of the Omicron spike RBD. The ability of the spike trimer to bind to three Fab molecules provides additional evidence for the potency of JMB2002 against Omicron.
The L452R mutation in the Delta variant is at the center of the binding epitope of JMB2002, and this mutation would clash with Y102 from the heavy chain of the Fab (Fig. 5F), thus pro-viding an explanation for its loss of potency against the Delta variant. In addition, the binding site of the JMB2002 Fab on the RBD is distinct from the epitopes for previously defined class I to class IV antibodies (Fig. 5G) (22), so JMB2002 represents a new class of antibody against the spike trimer.
In this study, we report biochemical characterization of the Omicron spike trimer and its binding to ACE2. Our data reveal that the Omicron RBD is less stable and more dynamic than the WT RBD, and the Omicron spike trimer has sixfold to ninefold increased affinity for binding to ACE2. We further solved the structures of the Omicron spike trimer in the apo state or bound to ACE2 or an anti-Omicron antibody. The ACE2-bound structure reveals that the Omicron spike trimer contains an unusual RBD-RBD interaction and extra interactions in the ACE2-RBD interface, both of which contribute to the higher affinity of ACE2 to the Omicron spike trimer. Structural analysis of the Omicron spike trimer also provides a mechanistic basis for the ability of Omicron to escape most therapeutic antibodies and reduce the efficacy of vaccinations. In addition, our structures of antibody-bound Omicron spiker trimer uncover a distinct mode of antibody binding to the spike trimer, in which the unusual RBD-RBD configuration is preserved. The binding epitope of this broad-spectrum antibody is different from previous anti-SARS-CoV-2 antibodies, thus opening a new venue for antibody drug discovery targeting various strains of SARS-CoV-2, including Omicron.