A delicate balance between antibody evasion and ACE2 affinity for Omicron BA.2.75

Variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have caused successive global waves of infection. These variants, with multiple mutations in the spike protein, are thought to facilitate escape from natural and vaccine-induced immunity and often increase in affinity for ACE2. The latest variant to cause concern is BA.2.75, identified in India where it is now the dominant strain, with evidence of wider dissemination. BA.2.75 is derived from BA.2 and contains four additional mutations in the receptor-binding domain (RBD). Here, we perform an antigenic and biophysical characterization of BA.2.75, revealing an interesting balance between humoral evasion and ACE2 receptor affinity. ACE2 affinity for BA.2.75 is increased 9-fold compared with BA.2; there is also evidence of escape of BA.2.75 from immune serum, particularly that induced by Delta infection, which may explain the rapid spread in India, where where there is a high background of Delta infection. ACE2 affinity appears to be prioritized over greater escape.

In brief Huo et al. characterize the SARS-CoV-2 variant BA.2.75 (originally identified in India). Its affinity for ACE2 is increased 9-fold over BA.2, and there is evidence of escape of BA.2.75 from immune serum, particularly from Delta infection. ACE2 affinity appears to be prioritized over greater escape via the R493Q reversion mutation.

INTRODUCTION
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of coronavirus disease 2019 (COVID-19), has caused a devastating global pandemic, resulting in more than half a billion reported cases (probably greatly underestimating the number of infections) and over 6.4 million deaths as of August 2022 (https://covid19.who.int/). As a positive-strand RNA virus, although its replication machinery contains a proofreading exonuclease, SARS-CoV-2 has a high viral replication error rate. 1 This, combined with the massive scale of the pandemic and chronic infection in immunocompromised individuals, 2 has generated mutational changes that endow viral fitness. The spike (S) gene, in particular, is the site of intense mutational change and selection, 3 and the encoded S protein, the major viral surface glycoprotein, is the principal antigenic target of all SARS-CoV-2 vaccines 4 and monoclonal antibody therapeutics 5 in current use.
S is presented as elongated trimeric spikes protruding from the virion surface. S is subdivided into an N-terminal S1 domain, responsible for host cell adhesion, and a C-terminal S2 domain anchored in the viral membrane, responsible for membrane fusion and cell entry after cleavage from S1, allowing the viral RNA to enter the host cell cytoplasm and initiate viral replication. 6 S1 consists of an N-terminal domain (NTD) and the receptor-binding domain (RBD), which mediates interaction with the ACE2 receptor on the host cell surface. Although a number of neutralizing monoclonal antibodies (nmAbs) have been found to target the NTD, especially the NTD supersite, 7 the majority of the nmAbs, particularly the most potent broadly reactive, target the RBD, 8,9 including all those in clinical use. 10 The RBD is thus under intense selective pressure, and mutational changes may endow the virus a fitness advantage by enhancing viral transmissibility via an increased binding affinity for ACE2 11 or to evade the humoral response by impairing binding of the nmAbs to the RBD. 12 The rapid genetic evolution of SARS-CoV-2 raises an immediate need to monitor and characterize the transmissibility of new variants and their capacity for immune evasion.
A large number of variants have emerged, several of which have been designated variants of concern (VoCs) (https://www.cdc. gov/coronavirus/2019-ncov/variants/variant-classifications.html). Some VoCs have caused successive waves of infection worldwide: Alpha, 13 then Delta, 14 and recently Omicron, 15 while Beta 16 in Southern Africa and Gamma in South America 17 have caused regional outbreaks without wide global spread.
In early May 2022, a new Omicron BA.2 sublineage designated BA.2.75 was reported in India (https://www.who.int/en/activities/ tracking-SARS-CoV-2-variants/) and has spread to multiple countries, including the UK, the USA, Australia, Germany, and Canada. Here, we report the antigenic characterization of BA.2.75 compared with other Omicron sublineages. In India, confirmed cases of BA.2.75 have outcompeted BA.5 and increased steeply from less than 20% of the total in early July to nearly 70% in mid-August (https://cov-spectrum.org/explore/ India/AllSamples/from=2022-07-01&to=2022-08-21/variants? variantQuery=nextcladePangoLineage%3ABA.2.75*&). We find that neutralization of BA.2.75 is reduced compared with BA.2 using a number of vaccine and immune sera, but reductions are not as great as those found with BA.4/5. However, sera from Delta-infected cases showed no neutralization of BA.2.75, which may underlie the evolution and emergence of BA.2.75 in India, which suffered a major Delta wave in 2021. Finally, perhaps the most striking change found in BA.2.75 is the affinity of ACE2/ RBD interaction. BA.2.75 affinity is increased 9-fold compared with BA.2. BA.2.75 has the highest affinity of all the SARS-CoV-2 variants measured to date and the only subnanomolar affinity we have determined. The N460K mutation probably increases affinity for ACE2 and also reduces the binding of some potent neutralizing antibodies. However, affinity to ACE2 appears to be prioritized over neutralization escape as evidenced by the acquisition of the RBD reversion mutation R493Q, which increases ACE2 affinity but makes the virus more sensitive to neutralization by vaccine sera. The very high affinity of BA.2.75 for ACE2 may increase the transmissibility of BA.2.75.

RESULTS
The Omicron lineage BA.2.75 BA.2.75 contains multiple mutational changes in the S protein compared with BA.2, including four substitutions in the NTD (W152R, F157L, I210V, and G257S) and four in the RBD (D339H, G446S, N460K, and R493Q) ( Figure 1). The RBD mutations impinge on major epitopes for neutralizing antibodies and are likely to modulate ACE2 binding. D339H represents a further evolution of the G339D mutation found in all previous Omicron variants that has been found to impair the binding of certain ''right-flank'' antibodies belonging to the IGHV1-69 family (e.g., Beta-49 and -50) and falls in the binding footprint of certain class 3 antibodies such as S309/sotrovimab. 15 G446S was found in BA.1, BA.1.1, and BA.3, but not in BA.2 and other BA.2 subvariants, and is also able to impair binding of certain class 3 antibodies binding the right shoulder such as REGN10987/imdevimab. 15 The R493Q reversion was also found in BA.4/5 and may make the virus more sensitive to neutralization by a number of class 1 and 2 antibodies binding the neck/left shoulder. This reversion may also increase the affinity for ACE2 (see below).
N460K is a mutation not seen in previous VoCs or Omicron sublineages, but it was found after in vitro (yeast display) evolution in RBD-62, which has an ultra-high ACE2 affinity (K D = 16-18 pM). 11,15 N460K was found repeatedly in these screens and is presumed to increase affinity for ACE2. 11 2    Figure 2E, which had essentially no neutralization of BA.1.1 (<50% neutralization at 1:20 serum dilution) and a low titer to BA.2.75 (7.7fold reduced compared with BA.4/5), was from the unvaccinated case in this series; if this was representative of the response in the unvaccinated, it would suggest that unvaccinated individuals may be more susceptible to BA.2.75 infection following BA.4/5 infection.
Individual BA.2.75 mutations have differential effects on neutralization To understand the effects of the individual mutations in the BA.2.75 RBD, we introduced them individually into the pseudovirus BA.2 background and assayed their neutralization using triple-vaccinated Pfizer BNT162b2 serum ( Figure 2F). Neutralization titers for BA.2 were reduced for 3 of the 4 single-mutation variants of BA.2, with the greatest decrease for N460K (2.9-fold, p < 0.0001), followed by D339H (1.3-fold, p = 0.0006), and then by G446S (1.2-fold, p = 0.2312); however, neutralization titers were increased 1.5-fold by the R493Q reversion mutation (p < 0.0001). Q493 is present in all vaccines, thus explaining the increase in activity of vaccine serum to this reversion mutation.  3B). This represents a considerable increase in affinity compared with BA.2 (9-fold) ( Figure 3C) and is even tighter than BA.4/5 (5-fold) ( Figure 3D), which was previously shown to bind ACE2 with higher affinity than BA.2. 12 Indeed, BA.2.75 is the strongest ACE2 binder among all SARS-CoV-2 VoCs, including Alpha (Alpha/ACE2 K D = 1.5 nM; (Figure 3E), and is the only subnanomolar affinity we have measured. We were unable to express the BA.2 + N460K RBD, which is expected to contribute to the increased affinity, but we measured the binding affinity of the BA.2 + R493Q RBD to ACE2 (K D = 0.55 nM) ( Figure 3F), confirming that the reversion mutation contributes to the high affinity of the BA.2.75 RBD.

ACE2/BA.2.75 RBD structure
To elucidate the molecular mechanism for high affinity, we determined the structure of the BA.2.75 RBD with ACE2 by crystallography (see STAR Methods). As expected, the binding mode was essentially indistinguishable from that observed before (Figure 4A), although there were significant rearrangements outside of the ACE2 footprint, with the flexible RBD 371-375 loop rearranging and part of the C-terminal 63His tag becoming ordered. Figure 4B shows a close up of the binding interface compared with the ACE2/BA.2 RBD complex. We note that in other complexes (with either R or Q at RBD 493), K31 of ACE2 tends to be disordered, whereas it is well ordered in the BA.2.75 complex, allowing K31 to form a potential hydrogen bond with the glutamine 493 side chain of the RBD, possibly increasing the affinity of ACE2. Although N460K is outside of the footprint of ACE2 on the RBD ( Figure 4A), evidence from in vitro evolution suggests that it probably increases the affinity for ACE2. 11 This is probably due to the improved electrostatic match, 11 although we also note that the density map for RBD-62 with ACE2 11 (EMDB: 12187) suggests that the glycan attached to N90 of ACE2 makes a direct interaction with the RBD close to residue 460.
The Omi mAbs were also tested against the pseudoviruses encoding single point mutations in the BA.2 RBD described above ( Figure S1; Table S2). As expected, the IGHV3-53/66 mAbs that lost neutralization to BA.2.75 were also impacted by the N460K Article ll OPEN ACCESS mutation, confirming the prediction that this residue was critical for the binding of a number of this public gene family. Interestingly, The BA.2 + N460K mutation in isolation shows a larger impact than the full BA.2.75 complement of S mutations on the activity of several mAbs: the neutralization titer of Omi-3 (IGHV3-53) was reduced 50-fold for BA.2 + N460K but only 2-fold for BA.2.75; Omi-17 (IGHV3-66) was completely knocked out on BA.2 + N460K but only reduced 4-fold for BA.2.75; and Omi-33 (IGHV3-33) was reduced 7-fold for BA.2 + N460K, but there was no change observed for BA.2.75. Thus, other mutations in BA.2.75 might have mitigated the effect of the N460K mutation, particularly the R493Q mutation, which has a different impact on various IGHV gene families and even differs within the IGHV3-53/66 family ( Figure 6C). However, we cannot fully explain the marked differences of effect observed for the impact of the 460 mutations between Omi-3 and Omi-18 ( Figure S1; Table S2), since the contacting GGS/T CDR-H2 motif is structurally almost identical between these two mAbs ( Figure 6B). Interestingly, BA.2.75 is more sensitive to Omi-32 (IGHV3-33) than is BA.2, with an 8-fold increase in neutralization titer ( Figure 5A; Table S1).
To confirm that the changes in neutralizing activities observed are associated with alterations in RBD interaction, we performed binding analyses of selected antibodies to BA.2.75 and BA.2 RBDs by SPR ( Figure S2). Binding of Omi-29 (IGHV3-53) and Omi-36 (IGHV3-66) to BA.2.75 was severely impaired, and Omi-18 and Omi-20 showed 8-fold reductions compared with BA.2. On the other hand, a 2-fold increase in binding affinity of Omi-32 was seen for BA.2.75 compared with BA.2, in line with the enhanced neutralization titer observed (above).

Effect of commercial monoclonals against BA.2.75
We evaluated the sensitivity of a panel of mAbs that have been developed as therapeutics against BA.2.75 ( Figure 5B; Table S1B). The neutralization profiles are, in general, similar between BA.2.75 and BA.2; however, further to the 6/12 mAbs (REGN10933, ADG10, ADG20, ADG30, LY-CoV555, LY-CoV16), which have already suffered complete loss of   15 For AZD1061, activity against BA.2.75 was similar to that against BA.2 (<3-fold reduction), while the AZD8895 titer was restored to 8 ng/mL for BA.2.75 from 1,333 ng/mL for BA.2, a 167-fold increase in activity. As a result, AZD7442 (a combination of AZD8895 and AZD1061) 26 showed similar activity against BA.2.75 and BA.2 (2-fold reduction). The results can be explained by the structure of the ternary complex of the ancestral SARS-CoV-2 RBD/AZD1061/AZD8895. 26 G446 has contacts with CDR-L2 Y55 and W56 of AZD1061, thus the G446S mutation will induce steric clashes ( Figures 6D and 6E), and while the CDR-H2 of AZD8895 sits above and makes a hydrogen bond to Q493 of the RBD, an arginine at 493 will severely clash with the CDR-H2 ( Figures 6F and 6G). The activity of S309 27 is increased 3-fold for BA.2.75 compared with BA.2, suggesting that the D339H mutation in BA.2.75 reduces the impact of the preceding G339D mutation in BA.2 on the activity of S309. LY-CoV1404 (bebtelovimab) 28 is the only mAb where neutralization is fully retained on all Omicron sublineages.

Antigenic mapping
We tested the neutralization of BA.2.75 using serum from previously infected individuals. This included serum obtained early in the pandemic (before the emergence of Alpha) together with sera obtained following Alpha, Beta, Gamma, Delta, BA.1, and BA.2 infections ( Figure S3). As expected, BA.2.75 neutralization titers were lower than the homologous infecting strain (e.g., Alpha serum on the Alpha strain). Most striking, however, was the complete loss of BA.2.75 neutralization using Delta serum (zero samples achieved 50% neutralization at 1/20 dilution). However, titers to BA.2.75 were much higher in cases who had been vaccinated before or after Delta infection.
We used these data to place BA.2.75 onto a 3D antigenic map using the method we have previously reported 12 ( Figures 7A and 7B; Videos S1 and S2). Initially, all VoCs  Figure 7A; Video S1); BA.2.75 forms part of the constellation of Omicron viruses, which segregate into one hemisphere of the 3D plot. BA.2.75 is well separated from other Omicron sublineages and especially from BA.4/5. It is notable that BA.2.75 and Delta are diametrically opposed in the diagram, emphasizing the antigenic distance between these two viruses. Since the data are higher dimensional, this 3D projection is likely to distort the true distances, and so we recalculated the map only for the Omicron lineage and early pandemic viruses (but retain the fully serology information for these). The results are shown in Figure 7B and Video S2 and recapitulate the major features of the plot containing the other VoCs but allow the Omicron sublineages to distribute more broadly in 3D space. It is remarkable that if we consider the two early pandemic viruses as a single point, and likewise merge BA.2 and BA.3 pairs, then the points are distributed as a trigonal bipyramid, maximizing their separation, consistent with antigenic escape being a significant factor in their evolution.

DISCUSSION
Following the designation of Omicron as a VoC in November 2021, a succession of sublineages emerged, including BA.1.1, BA.2, BA.2.12.1, and BA.4/5, which have outcompeted preceding strains to become regionally or globally dominant. Since June 2022, BA.4/5, which has both higher receptor binding affinity and a markedly enhanced escape from antibody responses, 12 quickly spread from South Africa across the world and has now become the new globally dominant strain, with BA.5 in the ascendency in many regions.
Very recently, a new Omicron sublineage designated as BA.2.75 has emerged in India and spread to many countries. The true prevalence of BA.2.75 is difficult to determine as sequencing in many countries is patchy and has, in recent months, been greatly scaled back. However, in India, BA. Overall, the constellation of mutations in BA.2.75 compared with BA.2 have opposing effects on neutralization, the reversion mutation R493Q makes the virus easier to neutralize using vaccine serum (the vaccine contains Q493), while N460K reduces neutralization titers to a greater extent when expressed in isolation compared with the combination of mutations seen in BA.2.75. N460K is a substitution that has not appeared in preceding variants of SARS-CoV-2. When we introduced N460K into the BA.2 backbone, BA.2 + N460K titers were reduced 2.9-fold compared with BA.2, which is greater than the reduction seen with BA.2.75 and on a par with the reduction seen for BA.4/ 5, using BNT162b2 triple-vaccinated serum.
Dissecting these effects using a panel of potent mAbs derived from vaccinated individuals who suffered BA.1 vaccine breakthrough infection, we show that those belonging to the IGHV3-53/66 family are reduced or knocked out against BA.2.75. IGHV3-53/66 are the most frequently isolated mAbs in SARS-CoV-2 and bind an epitope on the ''neck.'' 17 IGHV3-53/66 thus forms a major public antibody response, and it is no surprise that the virus has evolved to escape this response. Mutations found in previous VoCs lead to loss of function of many IGHV53/66 mAbs, but this antibody class has proved to be very adaptable to accommodate change, 20 and it would seem likely that somatic mutation will allow the response to adapt to the N460K mutation following BA.2.75 infection.
Interestingly, BA.2.75 has also acquired the R493Q reversion (Q493R was acquired in BA.1 and present in all other Omicron sublineages except BA.4/5). Here, we show that the BA.2.75 RBD is able to bind ACE2 with 9-fold higher affinity than BA.2 and more tightly than BA.4/5. 12,15 BA.2.75 has the highest ACE2 affinity among all SARS-CoV-2 variants we have measured to date, and we show that this is partly attributable to the R493Q mutation. Although we have been unable to express the BA.2 + N460K RBD, previous studies show that N460K can enhance RBD binding for ACE2, an effect similar in magnitude to that seen with the N501Y mutation described initially in Alpha; 11 thus N460K probably both enhances antibody escape and increases receptor binding affinity.
There is likely a fine interplay between antibody escape and ACE2 receptor affinity; Alpha (N501Y) evolved early during the pandemic, when the background population SARS-CoV-2 exposure was relatively low. Although neutralization titers against Alpha were modestly reduced compared with ancestral strains, 29 it is likely that the major driver for the evolution of Alpha N501Y was an increase in ACE2 affinity, giving the virus a transmission advantage. 30 Currently, population exposure to SARS-CoV-2 by either natural infection or vaccination is high, leading to the dual pressure of increased ACE2 affinity and antibody evasion. For the R493Q reversion, the balance between a reduction in antibody escape but increased ACE2 affinity may have tipped to allow BA.2.75 to more effectively transmit in certain populations. Other factors such as S stability, replication time, and reduced TMPRSS2 dependence also influence the success of SARS-CoV-2 variants. 30 In summary, we show the mutations in BA.2.75 lead to a reduction in neutralization titers of vaccine serum compared with BA.2. Individual BA.2.75 mutations can cause greater reduction in neutralization titers compared with the full BA.2.75 S sequence, but these are balanced by the R493Q reversion mutation, which may have been selected to increase affinity to ACE2 and increase the transmissibility of BA.2.75. It seems inevitable that further evolution of the Omicron lineage will occur, and there are likely many possible trade-offs between antibody escape and ACE2 affinity that can and will be made, leading to successive waves of infection.

Limitations of the study
Limitations of this study are that the in vitro neutralization assays we used do not probe the full function of the antibody response, as they do not measure the effects of complement-or antibodydependent cell-mediated cytotoxicity, which both operate in vivo. In addition, as live BA.2.75 virus was not available in our laboratory, we relied on lentiviral pseudoneutralization assays for characterization. Furthermore, they do not take account of T cell responses, which have been shown to be more resilient to the mutations expressed by VoCs.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Materials availability
Reagents generated in this study are available from the lead contact with a completed Materials Transfer Agreement.
Data and code availability d Data availability. The coordinates and structure factors of the crystallographic complex are available from the PDB with accession code 8ASY. d Code availability. This paper does not report original code. d Reagents generated in this study are available from the lead contact with a completed Materials Transfer Agreement. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Plasma from early pandemic and Alpha cases
Participants from the first wave of SARS-CoV2 in the U.K. and those sequence confirmed with B. Organisation) and times between symptom onset and sampling and age of participant was captured for all individuals at the time of sampling. Following heat inactivation of plasma/serum samples they were aliquoted so that no more than 3 freeze thaw cycles were performed for data generation. For subject details see Table S3. Diagnosis was confirmed through reporting of symptoms consistent with COVID-19, hospital presentation, and a test positive for SARS-CoV-2 using reverse transcriptase polymerase chain reaction (RT-PCR) from an upper respiratory tract (nose/ throat) swab tested in accredited laboratories and lineage sequence confirmed through national reference laboratories in the United Kingdom. A blood sample was taken following consent at least 14 days after PCR test confirmation. Clinical information including severity of disease (mild, severe or critical infection according to recommendations from the World Health Organisation) and times between symptom onset and sampling and age of participant was captured for all individuals at the time of sampling. For subject details see Table S3.

Sera from Beta, Gamma and Delta infected cases
Beta and Delta samples from UK infected cases were collected under the ''Innate and adaptive immunity against SARS-CoV-2 in healthcare worker family and household members'' protocol affiliated to the Gastro-intestinal illness in Oxford: COVID sub study discussed above and approved by the University of Oxford Central University Research Ethics Committee. All individuals had sequence confirmed Beta/Delta infection or PCR-confirmed symptomatic disease occurring whilst in isolation and in direct contact with Beta/ Delta sequence-confirmed cases. Additional Beta infected serum (sequence confirmed) was obtained from South Africa. At the time of swab collection patients signed an informed consent to consent for the collection of data and serial blood samples. Diagnosis was confirmed through reporting of symptoms consistent with COVID-19 or a positive contact of a known Omicron case, and a test positive for SARS-CoV-2 using reverse transcriptase polymerase chain reaction (RT-PCR) from an upper respiratory tract (nose/ throat) swab tested in accredited laboratories and lineage sequence confirmed through national reference laboratories. A blood sample was taken following consent at least 10 days after PCR test confirmation. Clinical information including severity of disease (mild, severe or critical infection according to recommendations from the World Health Organisation) and times between symptom onset and sampling and age of participant was captured for all individuals at the time of sampling. For subject details see Table S3.
Sera from BA.2 infected cases, study subjects Following informed consent, healthcare workers with BA.2 infection were co-enrolled under the Sheffield Biobank study (STHObs) (18/YH/0441). All individuals had PCR-confirmed symptomatic disease and sequence confirmed BA.2 infection through national UKHSA sequencing data. A blood sample was taken following consent at least 12 days after PCR test confirmation. Clinical information including vaccination history, times between symptom onset and sampling and age of participant was captured for all individuals at the time of sampling. For subject details see Table S3. Article ll OPEN ACCESS 20/SC/0179) and a regulatory agency in the United Kingdom (the Medicines and Healthcare Products Regulatory Agency). An independent DSMB reviewed all interim safety reports. A copy of the protocols was included in previous publications. 41 Data from vaccinated volunteers who received three vaccinations are included in this study. Blood samples were collected and serum separated approximately 28 days (range 26-34 days) following the third dose. For subject details see column 'AZ V3+28' in Table S3.

METHOD DETAILS
Pseudovirus plasmid construction and lentiviral particles production Pseudotyped lentivirus expressing SARS-CoV-2 S proteins from ancestral strain (Victoria, S247R) supernatants were removed and 50 mL of 1:2 Bright-Glo TM Luciferase assay system (Promega, USA) in 1 3 PBS was added to each well. The reaction was incubated at room temperature for 5 min and firefly luciferase activity was measured using CLARIOstar (BMG Labtech, Ortenberg, Germany). The percentage neutralization was calculated relative to the control. Probit analysis was used to estimate the dilution that inhibited half maximum pseudotyped lentivirus infection (PVNT50).
To determine the neutralizing activity of convalescent plasma/serum samples or vaccine sera, 3-fold serial dilutions of each samples were incubated with pseudoviral particles for 1 h and the same strategy as mAb was applied.

Cloning of RBDs
To generate the BA.2.75 RBD construct, site-directed PCR mutagenesis was performed using the BA.2 Spike construct as the template, 20 with the introduction of D339H, G446S, N460K and R493Q mutations using primers listed in Table S4; the gene fragment was amplified with D339H_pNeoF and RBD333_BAP_R (Table S4), and cloned into the pOPINTTGneo-BAP vector. 43 To generate the BA.2 + R493Q RBD construct, site-directed PCR mutagenesis was performed using the BA.2 Spike construct as the template, with the introduction of R493Q mutation suing primers listed in Table S4; the gene fragment was amplified with pNeoRBD333Omi_F and RBD333_BAP_R, and cloned into the pNeo vector. 13 Cloning was performed using the ClonExpress II One Step Cloning Kit (Vazyme). The Constructs were verified by Sanger sequencing after plasmid isolation using QIAGEN Miniprep kit (QIAGEN).

Production of RBDs
Plasmids encoding RBDs were transfected into Expi293F Cells (ThermoFisher) by PEI, cultured in FreeStyle 293 Expression Medium (ThermoFisher) at 37 C for 1 day followed by 30 C for 3 days with 8% CO 2 . To express biotinylated RBDs, the RBD-BAP plasmid was co-transfected with pDisplay-BirA-ER (Addgene plasmid 20,856; coding for an ER-localized biotin ligase), in the presence of 0.8 mM D-biotin (Sigma-Aldrich). The conditioned medium was diluted 1:2 into binding buffer (50 mM sodium phosphate, 500 mM sodium chloride, pH 8.0). RBDs were purified with a 5 mL HisTrap nickel column (GE Healthcare) through His-tag binding, followed by a Superdex 75 10/300 GL gel filtration column (GE Healthcare) in 10 mM HEPES and 150 mM sodium chloride.

Surface plasmon resonance
Surface plasmon resonance experiments were performed using a Biacore T200 (GE Healthcare). All assays were performed with running buffer of HBS-EP (Cytiva) at 25 C.
To determine the binding kinetics between BA.2.75 or BA.2 + R493Q RBD and ACE2, a Protein A sensor chip (Cytiva) was used. ACE2-Fc was immobilised onto the sample flow cell of the sensor chip. The reference flow cell was left blank. RBD was injected over the two flow cells at a range of five concentrations prepared by serial two-fold dilutions, at a flow rate of 30 mL min À1 using a singlecycle kinetics program. Running buffer was also injected using the same program for background subtraction. All data were fitted to a 1:1 binding model using Biacore T200 Evaluation Software 3.1.
To confirm the binding kinetics between the BA.2.75 RBD and ACE2, a Biotin CAPture Kit (Cytiva) was used. Biotinylated ACE2 (bio-ACE2) was immobilised onto the sample flow cell of the sensor chip. The reference flow cell was left blank. The BA.2.75 RBD was injected over the two flow cells at a range of five concentrations prepared by serial two-fold dilutions, at a flow rate of 30 mL min À1 using a single-cycle kinetics program. Running buffer was also injected using the same program for background subtraction. All data were fitted to a 1:1 binding model using Biacore T200 Evaluation Software 3.1. To determine the binding kinetics between the BA.2.75 or BA.2 RBD and mAbs, a Biotin CAPture Kit (Cytiva) was used. Biotinylated RBD was immobilised onto the sample flow cell of the sensor chip. The reference flow cell was left blank. The Fab of Omi-18 or Omi-32 was injected over the two flow cells at a range of five concentrations prepared by serial two-fold dilutions, at a flow rate of 30 mL min À1 using a single-cycle kinetics program. For the binding of Omi-20 for bio-BA.2 RBD, the Fab of Omi-20 was injected over the two flow cells at a range of five concentrations prepared by serial two-fold dilutions, at a flow rate of 30 mL min À1 using a single-cycle kinetics program. For the binding of Omi-20 for bio-BA.2.75 RBD, the Fab of Omi-20 was injected over the two flow cells at a range of eight concentrations prepared by serial twofold dilutions, at a flow rate of 30 mL min À1 . Running buffer was also injected using the same program for background subtraction. All data were fitted to a 1:1 binding model using Biacore T200 Evaluation Software 3.1.
To compare the binding profiles between BA.2 and BA.2.75 RBD for mAb Omi-29, a Biotin CAPture Kit (Cytiva) was used. Biotinylated BA.2 and BA.2.75 RBD was immobilised onto the sample flow cell of the sensor chip to a similar level ($110 RU). The reference flow cell was left blank. A single injection of mAb Fab was performed over the two flow cells at 1 mM, at a flow rate of 30 mL min À1 . Running buffer was also injected using the same program for background subtraction. The sensorgrams were plotted using Prism9 (GraphPad).
To compare the binding profiles between BA.2 and BA.2.75 RBD for mAb Omi-36, a sensor chip Protein A (Cytiva) was used. mAb Omi-36 in the IgG form was immobilised onto the sample flow cell of the sensor chip. The reference flow cell was left blank. A single injection of RBD was performed over the two flow cells at 200 nM, at a flow rate of 30 mL min À1 . Running buffer was also injected using the same program for background subtraction. The sensorgrams were plotted using Prism9 (GraphPad).

IgG mAbs and Fabs production
AstraZeneca and Regeneron antibodies were provided by AstraZeneca, Vir, Lilly and Adagio antibodies were provided by Adagio, LY-CoV1404 was provided by LifeArc. For the in-house antibodies, heavy and light chains of the indicated antibodies were transiently transfected into 293T cells and antibody purified from supernatant on protein A as previously described. 20 Fabs were digested from purified IgGs with papain using a Pierce Fab Preparation Kit (Thermo Fisher), following the manufacturer's protocol.
Crystallization, X-Ray data collection and structure determination Purified BA.2.75 RBD was deglycosylated with Endoglycosidase H1 and mixed with ACE2 in a 1:1 M ratio, with a final concentration of 13.0 mg mL À1 . Initial screening of crystals was set up in Crystalquick 96-well X plates (Greiner Bio-One) with a Cartesian Robot using the nanoliter sitting-drop vapor-diffusion method, with 100 nL of protein plus 100 nL of reservoir in each drop, as previously described. 44 Crystals of BA.2.75 RBD-ACE2 complex were formed in Hampton Research PEGRx condition 2-25, containing 0.1% (w/v) n-Octyl-b-D-glucoside, 0.1 M Sodium citrate tribasic dihydrate pH 5.5 and 22% (w/v) PEG 3350. Crystals were mounted in loops and dipped in solution containing 25% glycerol and 75% mother liquor for a second before frozen in liquid nitrogen. Diffraction data were collected at 100 K at beamline I03 of Diamond Light Source, UK, using the automated queue system that allows unattended automated data collection (https://www.diamond.ac.uk/Instruments/Mx/I03/I03-Manual/Unattended-Data-Collections. html). The best crystal diffracted to 2.85 Å resolution. 3600 diffraction images of 0.1 each were collected and automatically processed with Xia2-dials. 38,45 The structure was determined by rigid body refinement using the model of BA.2 RBD/ACE2 complex (PDB, 7ZF7) 20 of which the unit cell is isomorphous to the current crystal. Model rebuilding is done with COOT 37 and refinement with Phenix. 39 Data collection and structure refinement statistics are given in Table S5. Structural comparisons used SHP 46 and figures were prepared with PyMOL (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrö dinger, LLC).

Antigenic mapping
Antigenic mapping of omicron was carried out using a previously described. 12 In short, coronavirus variants were assigned threedimensional coordinates whereby the distance between two points indicates the base drop in neutralization titer. Each serum was assigned a strength parameter which provided a scalar offset to the logarithm of the neutralization titer. These parameters were refined to match predicted neutralization titers to observed values by taking an average of superimposed positions from 30 separate runs. The three-dimensional positions of the variants of concern: Victoria, Alpha, Beta, Gamma, Delta and Omicron were plotted for display.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analyses are reported in the results and figure legends. Neutralization was measured on pseudovirus. The percentage reduction was calculated and IC 50 determined using the probit program from the SPSS package. The Wilcoxon matched-pairs signed rank test was used for the analysis and two-tailed p values were calculated on geometric mean values.   s e u d o v i r a l n e u t r a l i z a t i o n a s s a y s a g a i n s t mo n o c l o n a l a n t Figure S1, related to Figure 5)      Table S5. X-ray data collection and structure refinement statistics (related to Figure 4)