Antibody evasion by SARS-CoV-2 Omicron subvariants BA.2.12.1, BA.4 and BA.5

SARS-CoV-2 Omicron subvariants BA.2.12.1 and BA.4/5 have surged notably to become dominant in the United States and South Africa, respectively1,2. These new subvariants carrying further mutations in their spike proteins raise concerns that they may further evade neutralizing antibodies, thereby further compromising the efficacy of COVID-19 vaccines and therapeutic monoclonals. We now report findings from a systematic antigenic analysis of these surging Omicron subvariants. BA.2.12.1 is only modestly (1.8-fold) more resistant to sera from vaccinated and boosted individuals than BA.2. However, BA.4/5 is substantially (4.2-fold) more resistant and thus more likely to lead to vaccine breakthrough infections. Mutation at spike residue L452 found in both BA.2.12.1 and BA.4/5 facilitates escape from some antibodies directed to the so-called class 2 and 3 regions of the receptor-binding domain3. The F486V mutation found in BA.4/5 facilitates escape from certain class 1 and 2 antibodies but compromises the spike affinity for the viral receptor. The R493Q reversion mutation, however, restores receptor affinity and consequently the fitness of BA.4/5. Among therapeutic antibodies authorized for clinical use, only bebtelovimab retains full potency against both BA.2.12.1 and BA.4/5. The Omicron lineage of SARS-CoV-2 continues to evolve, successively yielding subvariants that are not only more transmissible but also more evasive to antibodies.

A subset of the pseudovirus neutralization data was confirmed for four monoclonal antibodies (COV2-2196, ZCB11, REGN10987 and LY-CoV1404) in neutralization experiments using authentic viruses BA.2 and BA.4 (Extended Data Fig. 1b and Extended Data Table 1). Similar neutralization patterns were observed in the two assays, although the precise 50% neutralizing titres were different.
To identify the mutations in BA.2.12.1 and BA.4/5 that confer antibody resistance, we assessed the neutralization sensitivity of pseudoviruses carrying each of the point mutations in the background of D614G or BA.2 to the aforementioned panel of mAbs and combinations. Detailed findings are presented in Extended Data Figs. 2 and 3 and Extended Data Table 2, and the most salient results are highlighted in Fig. 2b and discussed here. Substitutions (M, R and Q) at residue L452, previously found in the Delta and Lambda variants 21,22 , conferred resistance largely to classes 2 and 3 RBD mAbs, with L452R being the more detrimental mutation. F486V broadly impaired the neutralizing activity of several class 1 and 2 RBD mAbs. Notably, this mutation decreased the potency of ZCB11 2,000-fold. By contrast, the reversion mutation R493Q sensitized BA.2 to neutralization by several class 1 and 2 RBD mAbs. This finding is consistent with our previous study 23 showing that Q493R found in the earlier Omicron subvariants mediated resistance to the same set of mAbs. L452, F486 and Q493, situated at the top of RBD, are among the residues most commonly targeted by SARS-CoV-2 neutralizing mAbs whose epitopes have been defined (Fig. 2c). In silico structural analysis showed that both L452R and L452Q caused steric hindrance to the binding by class 2 RBD mAbs. One such example is shown for LY-CoV555 (Fig. 2d), demonstrating the greater clash because of the arginine substitution and explaining why this particular mutation led to a larger loss of virus-neutralizing activity (Fig. 2b). Structural modelling of the F486V again showed steric hindrance to binding by class 2 RBD mAbs such as REGN10933, LY-CoV555 and 2-15 caused by the valine substitution (Fig. 2e).

Receptor affinity
Epidemiological data clearly indicate that both BA.2.12.1 and BA.4/5 are very transmissible (Fig. 1a); however, the further mutations at the top of RBD (Fig. 2c) of these subvariants raises the possibility of a significant loss of affinity for the viral receptor, human angiotensin-converting enzyme 2 (hACE2). We therefore measured the binding affinity of purified spike proteins of D614G and main Omicron subvariants to dimeric hACE2 using surface plasmon resonance (SPR  including some that mediate antibody escape, BA.2.12.1 and BA.4/5 also evolved concurrently to gain a slightly higher affinity for the receptor than an ancestral SARS-CoV-2, D614G (K D 5.20 nM).
To support the findings by SPR and to probe the role of point mutations in hACE2 binding, we tested BA.2, BA.2.12.1, and BA.4/5 pseudoviruses, as well as pseudoviruses containing key mutations, for their neutralization by dimeric hACE2 in vitro. The 50% inhibitory concentration (IC 50 ) values were lower for BA.4/5 and BA.2.12.1 than that for BA.2 (Fig. 3b), again indicating that these two emerging Omicron subvariants have not lost receptor affinity. Our results also showed that the F486V mutation compromised receptor affinity, as previously reported 24 , while the R493Q reversion mutation improved receptor affinity. To structurally interpret these results, we modelled F486V and R493Q mutations on the basis of the crystal structure of BA.1-RBD-hACE2 complex 25 overlaid with ligand-free BA.2 RBD (Protein Data Bank (PDB) 7U0N and 7UB0). This analysis found that both R493 and F486 are conformationally similar between BA.1 and BA.2, and F486V led to a loss of interaction with a hydrophobic pocket in hACE2 (Fig. 3c). On the other hand, the R493Q reversion mutation restored a hydrogen bond with H34 and avoided the charge repulsion by K31, seemingly having the opposite effect of F486V. The mutation frequency at F486 had been exceedingly low (<10 × 10 −5 ) until the emergence of BA.4/5 (Extended Data Table 3), probably because of a compromised receptor affinity. Taken together, our findings in Figs. 2 and 3 suggest that F486V allowed BA.4 and BA.5 to extend antibody evasion while R493Q compensated to regain fitness in receptor binding.

Neutralization by polyclonal sera
We next assessed the extent of BA.2.12.1 and BA.4/5 resistance to neutralization by sera from four different clinical cohorts. Sera from people immunized with only two doses of COVID-19 messenger RNA vaccines

Article
were not examined because most of them could not neutralize earlier Omicron subvariants 23,26 . Instead, we measured serum neutralizing activity for people who received three shots of mRNA vaccines (boosted), individuals who received mRNA vaccines before or after non-Omicron infection and patients with either BA.1 or BA.2 breakthrough infection after vaccination. Their clinical information is described in Extended Data Table 4, and the serum neutralization profiles are presented in Extended Data Fig. 4 and the 50% inhibitory dose (ID 50 ) titres are summarized in Fig. 4a. For the 'boosted' cohort, neutralization titres were noticeably lower (4.6-to 6.2-fold) for BA.1, BA.1.1 and BA.2 compared to D614G (Fig. 4b), as previously reported 23 We also conducted serum neutralization assays on pseudoviruses containing point mutations found in BA.2.12.1 or BA.4/5 in the background of BA.2. Del69-70, L452M/R/Q and F486V each modestly (1.1-to 2.4-fold) decreased the neutralizing activity of sera from all cohorts, while the R493Q reversion mutation modestly (roughly 1.5-fold) enhanced the neutralization (Fig. 4c and Extended Data Fig. 5). S704L, a mutation close to the S1/S2 cleavage site, did not appreciably alter the serum neutralization titres against BA.2. For boosted serum samples, the impact of each point mutant on neutralization resistance was quantified and summarized in Fig. 4b.
Using these serum neutralization results, we then constructed a graphic display to map antigenic distances among D614G, various Omicron subvariants, and individual point mutants using only results from the boosted serum samples to avoid confounding effects from differences in clinical histories in the other cohorts. Using methods well established in influenza research 27 , all virus and serum positions on the antigenic map were optimized so that the distances between them correspond to the fold drop in neutralizing ID 50 titre relative to the maximum titre for each serum. Each unit of distance in any direction on the antigenic map corresponds to a two-fold change in ID 50 titre. The resultant antigenic cartography (Fig. 4d)

Discussion
We have systematically evaluated the antigenic properties of SARS-CoV-2 Omicron subvariants BA.2.12.1 and BA.4/5, which are rapidly expanding globally (Fig. 1a)  modestly (1.8-fold) more resistant to sera from vaccinated and boosted individuals than the BA.2 subvariant that currently dominates the global pandemic (Fig. 4b). On the other hand, BA.4/5 is substantially (4.2-fold) more resistant, a finding consistent with results recently posted by other groups 1,28 . This antigenic distance is similar to that between the Delta variant and the ancestral virus 29 and thus is likely to lead to more breakthrough infections in the coming months. A key question now is whether BA.4/5 would out-compete BA.2.12.1, which poses less of an antigenic threat. This competition is now playing out in the United Kingdom. These new Omicron subvariants were first detected there almost simultaneously in late March of 2022. However, BA.2.12.1 now accounts for 13% of new infections in the United Kingdom, whereas the frequency is over 50% for BA.4/5 (Fig. 1a), suggesting a transmission advantage for the latter. Epidemiologically, as both of these two Omicron subvariants have a clear advantage in transmission, it is therefore not surprising that their abilities to bind the hACE2 receptor remain robust (Fig. 3a)

Article
Our studies on the specific mutations found in BA.2.12.1 and BA.4/5 show that Del69-70, L452M/R/Q and F486V could individually contribute to antibody resistance, whereas R493Q confers antibody sensitivity (Fig. 4b). Moreover, the data generated using SARS-CoV-2-neutralizing mAbs indicate that a mutation at L452 allows escape from class 2 and class 3 RBD antibodies and that the F486V mutation mediates escape from class 1 and class 2 RBD antibodies (Fig. 2b). It is not clear how Del69-70, a mutation that might increase infectivity 32 and previously seen in the Alpha variant 33 , contributes to antibody resistance except for the possible evasion from certain neutralizing antibodies directed to the NTD. As for the use of clinically authorized mAbs to treat or block infection by BA.2.12.1 or BA.4/5, only bebtelovimab (LY-COV1404) 11 retains potency, whereas the combination of tixagevimab and cilgavimab (COV2-2196 and COV2-2130) 6 shows a modest loss of activity (Fig. 2a).
As the Omicron lineage has evolved over the past few months (Fig. 1b), each successive subvariant has seemingly become better and better at human transmission (Fig. 1a) as well as in antibody evasion 23,34 . It is only natural that scientific attention remains intently focused on each new subvariant of Omicron. However, we must be mindful that each of the globally dominant variants of SARS-CoV-2 (Alpha, Delta and Omicron) emerged stochastically and unexpectedly. Vigilance in our collective surveillance effort must be sustained.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-05053-w.