The COVID-19 pandemic continues as the causative agent SARS-CoV-2 continues to evolve. Many diverse viral variants have emerged (Fig. 1a), each characterized by mutations in the spike protein that raise concerns of both antibody evasion and enhanced transmission. The B.1.351/Beta variant was found to be highly refractory to antibody neutralization4 and therefore compromised the efficacy of vaccines5,6,7 and therapeutic antibodies. The B.1.1.7/Alpha variant became dominant globally in early 2021 due to having an edge in transmission8, only to be replaced by the B.1.617.2/Delta variant, which exhibited an even greater propensity to spread coupled with a moderate level of antibody resistance9. Next came the B.1.1.529/Omicron variant, which was first detected in southern Africa in November 2021 (refs. 3,10,11) (Fig. 1a). Omicron has since spread rapidly in the region, as well as to over 60 countries, gaining traction even in regions in which the Delta variant is prevalent. The short doubling time (2–3 days) of Omicron cases suggests that it could soon become dominant2. Moreover, its spike protein contains an alarming number of mutations (>30) (Fig. 1b and Extended Data Fig. 1), including at least 15 in the receptor-binding domain (RBD), the principal target of neutralizing antibodies. These extensive spike mutations raise the possibility that current vaccines and therapeutic antibodies would be greatly compromised. This concern is amplified by the findings that we now report.

Fig. 1: Resistance of B.1.1.529 to neutralization by sera.
figure 1

a, Unrooted phylogenetic tree of B.1.1.529 with other major SARS-CoV-2 variants. b, Key spike mutations found in the viruses isolated in the major lineage of B.1.1.529. Del, deletion; ins, insertion. c, Neutralization of D614G and B.1.1.529 pseudoviruses by convalescent patient sera. d, Neutralization of D614G and B.1.1.529 pseudoviruses by vaccinee sera. Within the four standard vaccination groups, individuals who were vaccinated without documented infection are denoted as circles and individuals who were both vaccinated and infected are denoted as triangles. Within the boosted group, Moderna vaccinees are denoted as squares and Pfizer vaccinees are denoted as diamonds. e, Neutralization of authentic D614G and B.1.1.529 viruses by vaccinee sera. Moderna vaccinees are denoted as squares and Pfizer vaccinees are denoted as diamonds. Data represent one out of two independent experiments. For all of the panels, the values above the symbols denote the geometric mean titre and the numbers in parentheses denote the number of samples above the limit of detection. P values were determined by using two-tailed Wilcoxon matched-pairs signed-rank tests.

Serum neutralization of B.1.1.529

We first examined the neutralizing activity of serum collected in spring 2020 from patients with COVID-19, who were presumably infected with the wild-type (WT) SARS-CoV-2 (9–120 days after symptoms) (Methods and Extended Data Table 1). Samples from ten individuals were tested for neutralization against both D614G (WT) and B.1.1.529 pseudoviruses. Although robust titres were observed against D614G, a significant drop (>32-fold) in 50% infectious dose (ID50) titres was observed against B.1.1.529, with only two samples showing titres above the limit of detection (LOD) (Fig. 1c and Extended Data Fig. 2a). We next assessed the neutralizing activity of sera from individuals who received one of the four widely used COVID-19 vaccines: BNT162b2 (Pfizer, 15–213 days after vaccination), mRNA-1273 (Moderna, 6–177 days after vaccination), Ad26.COV2.S (Johnson & Johnson, 50–186 days after vaccination) and ChAdOx1 nCoV-19 (AstraZeneca, 91–159 days after vaccination) (Methods and Extended Data Table 2). In all cases, a substantial loss in neutralizing potency was observed against B.1.1.529 (Fig. 1d and Extended Data Fig. 2b–f). For the two mRNA-based vaccines, BNT162b2 and mRNA-1273, a >21-fold and >8.6-fold decrease in ID50 was observed, respectively. Note that, for these two groups, we specifically chose samples with high titres so that the fold change in titre could be better quantified; the difference in the number of samples with titres above the LOD (6 out of 13 for BNT162b2 versus 11 out of 12 for mRNA-1273) may therefore be favourably biased. Within the Ad26.COV2.S and ChAdOx1 nCOV-19 groups, all of the samples were below the LOD against B.1.1.529, except for two Ad26.COV2.S samples from patients who had a previous history of SARS-CoV-2 infection (Fig. 1d). Collectively, these results suggest that individuals who were previously infected or fully vaccinated remain at risk for B.1.1.529 infection.

Booster shots are now routinely administered in many countries 6 months after full vaccination. We therefore also examined the serum neutralizing activity of individuals who had received three homologous mRNA vaccinations (13 with BNT162b2 and 2 with mRNA-1273, 14–90 days after vaccination). Every sample showed lower activity in neutralizing B.1.1.529, with a mean decrease of 6.5-fold compared with the WT (Fig. 1d). Although all of the samples had titres above the LOD, the substantial loss in activity may still pose a risk for B.1.1.529 infection despite the booster vaccination.

We next confirmed the above findings by testing a subset of the BNT162b2 and mRNA-1273 vaccinee serum samples using authentic SARS-CoV-2 isolates: WT and B.1.1.529. Again, a substantial decrease in the neutralization of B.1.1.529 was observed, with mean decreases of >6.0-fold and >4.1-fold for the fully vaccinated group and the boosted group, respectively (Fig. 1e).

Antibody neutralization of B.1.1.529

To understand the types of antibodies in the serum that lost neutralizing activity against B.1.1.529, we assessed the neutralization profile of 19 well-characterized monoclonal antibodies against the spike protein, including 17 directed to the RBD and 2 directed to the N-terminal domain (NTD). We included the following monoclonal antibodies that have been authorized or approved for clinical use, either individually or in combination: REGN10987 (imdevimab)12, REGN10933 (casirivimab)12, COV2-2196 (tixagevimab)13, COV2-2130 (cilgavimab)13, LY-CoV555 (bamlanivimab)14, CB6 (etesevimab)15, Brii-196 (amubarvimab)16, Brii-198 (romlusevimab)16 and S309 (sotrovimab)17. We also included other monoclonal antibodies of interest: 910-30 (ref. 18), ADG-2 (ref. 19), DH1047 (ref. 20), S2X259 (ref. 21) and our antibodies 1-20, 2-15, 2-7, 4-18, 5-7 and 10-40 (refs. 22,23,24). The footprints of monoclonal antibodies with structures available were drawn in relation to the mutations found in the B.1.1.529 RBD (Fig. 2a) and NTD (Fig. 2b). The risk to each of the four classes25 of RBD monoclonal antibodies, as well as to the NTD monoclonal antibodies, was immediately apparent. Indeed, neutralization studies on B.1.1.529 pseudovirus showed that 17 out of the 19 monoclonal antibodies tested lost neutralizing activity completely or partially (Fig. 2c and Extended Data Fig. 3). The potency of class 1 and class 2 RBD monoclonal antibodies all decreased by more than 100-fold, as did the more potent monoclonal antibodies in RBD class 3 (REGN10987, COV2-2130 and 2-7). The activities of S309 and Brii-198 were spared. All of the monoclonal antibodies in RBD class 4 lost neutralization potency against B.1.1.529 by at least tenfold, as did monoclonal antibodies directed to the antigenic supersite26 (4-18) or the alternative site23 (5-7) on the NTD. Strikingly, all four combination monoclonal antibody drugs in clinical use lost substantial activity against B.1.1.529, probably abolishing or impairing their efficacy in patients.

Fig. 2: Resistance of B.1.1.529 to neutralization by monoclonal antibodies.
figure 2

a, Footprints of RBD-directed antibodies. Mutations within B.1.1.529 are highlighted in cyan. Approved or authorized antibodies are shown in bold. The receptor-binding motif (RBM) residues are highlighted in yellow. b, Footprints of NTD-directed antibodies. Mutations within B.1.1.529 are highlighted in cyan. The NTD supersite residues are highlighted in light pink. c, Neutralization of D614G and B.1.1.529 pseudoviruses by RBD-directed and NTD-directed monoclonal antibodies (mAbs). d, Neutralization of D614G and B.1.1.529+R346K pseudoviruses by RBD-directed and NTD-directed monoclonal antibodies. Data represent one out of two independent experiments.

Approximately 10% of the B.1.1.529 viruses in the Global Initiative on Sharing All Influenza Data (GISAID)1 also contain an additional RBD mutation, R346K, which is the defining mutation for the B.1.621/Mu variant27. We therefore constructed another pseudovirus (B.1.1.529+R346K) containing this mutation for additional testing using the same panel of monoclonal antibodies (Fig. 2d). The overall findings resembled those already shown in Fig. 2c, with the exception that the neutralizing activity of Brii-198 was abolished. In fact, nearly the entire panel of antibodies was essentially rendered inactive against this minor form of the Omicron variant.

The fold changes in IC50 of the monoclonal antibodies against B.1.1.529 and B.1.1.529+R346K relative to D614G are summarized in the first two rows of Fig. 3a. The considerable loss of activity observed for all classes of monoclonal antibodies against B.1.1.529 suggests that perhaps the same is occurring in the serum of convalescent patients and vaccinated individuals.

Fig. 3: Impact of individual mutations within B.1.1.529 against monoclonal antibodies.
figure 3

a, Neutralization of pseudoviruses containing single mutations found within B.1.1.529 by a panel of 19 monoclonal antibodies. The fold change relative to neutralization of D614G is denoted, with resistance coloured red and sensitization coloured green. b, Modelling of critical mutations in B.1.1.529 that affect antibody neutralization.

Mutations conferring antibody resistance

To understand the specific B.1.1.529 mutations that confer antibody resistance, we next individually tested the same panel of 19 monoclonal antibodies against pseudoviruses for each of the 34 mutations (excluding D614G) found in B.1.1.529 or B.1.1.529+R346K. Our findings not only confirmed the role of known mutations at spike residues 142–145, 417, 484 and 501 in conferring resistance to NTD or RBD (class 1 or class 2) antibodies4 but also revealed several mutations that were not previously known to have functional importance to neutralization (Fig. 3a and Extended Data Fig. 4). Q493R, which was previously shown to affect binding of CB6 and LY-CoV555 (ref. 28) as well as polyclonal sera29, mediated resistance to CB6 (class 1) as well as to LY-CoV555 and 2-15 (class 2), findings that could be explained by the abolishment of hydrogen bonds due to the long side chain of arginine and induced steric clashes with CDRH3 in these antibodies (Fig. 3b (left)). Both N440K and G446S mediated resistance to REGN10987 and 2-7 (class 3), observations that could also be explained by steric hindrance (Fig. 3b (middle)). The most striking and perhaps unexpected finding was that S371L broadly affected neutralization by monoclonal antibodies in all four RBD classes (Fig. 3a and Extended Data Fig. 4). Although the precise mechanism of this resistance is unknown, in silico modelling suggested two possibilities (Fig. 3b (right)). First, in the RBD-down state, mutating Ser to Leu results in an interference with the N343 glycan, thereby possibly alteringits conformation and affecting class 3 antibodies that typically bind to this region. Second, in the RBD-up state, S371L may alter the local conformation of the loop consisting of Ser371-Ser373-Ser375, thereby affecting the binding of class 4 antibodies that generally target a portion of this loop24. It is not clear how class 1 and class 2 RBD monoclonal antibodies are affected by this mutation.

Evolution of SARS-CoV-2 to antibodies

To gain insights into the antibody resistance of B.1.1.529 relative to previous SARS-CoV-2 variants, we evaluated the neutralizing activity of the same panel of neutralizing monoclonal antibodies against pseudoviruses for B.1.1.7 (ref. 8), B.1.526 (ref. 30), B.1.429 (ref. 31), B.1.617.2 (ref. 9), P.1 (ref. 32) and B.1.351 (ref. 33). It is evident from these results (Fig. 4 and Extended Data Fig. 5) that previous variants developed resistance to only NTD antibodies and class 1 and class 2 RBD antibodies. Here B.1.1.529, with or without R346K, has made a big mutational leap by becoming not only nearly completely resistant to class 1 and class 2 RBD antibodies, but also substantially resistant to both class 3 and class 4 RBD antibodies. B.1.1.529 is now the most complete ‘escapee’ from neutralization by currently available antibodies.

Fig. 4: Evolution of antibody resistance across SARS-CoV-2 variants.
figure 4

Neutralization of SARS-CoV-2 variant pseudoviruses by a panel of 19 monoclonal antibodies. The fold change relative to neutralization of D614G is denoted.


The Omicron variant caused fear almost as soon as it was detected to be spreading in South Africa. The suggestion that this new variant would transmit more readily has come true in the ensuing weeks2. The extensive mutations found in its spike protein raised concerns that the efficacy of current COVID-19 vaccines and antibody therapies might be compromised. Indeed, in this study, sera from convalescent patients (Fig. 1c) and vaccinees (Fig. 1d, e) showed markedly reduced neutralizing activity against B.1.1.529. Other studies have found similar losses34,35,36,37,38. These findings are consistent with emerging clinical data on the Omicron variant demonstrating higher rates of reinfection11 and vaccine breakthroughs. In fact, recent reports showed that the efficacy of two doses of BNT162b2 vaccine has dropped from over 90% against the original SARS-CoV-2 strain to approximately 40% and 33% against B.1.1.529 in the United Kingdom39 and South Africa40, respectively. Even a third booster shot may not adequately protect against Omicron infection39,41, although the protection against disease still makes it advisable to administer booster vaccinations. Vaccines that elicited lower neutralizing titres35,42 are expected to fare worse against B.1.1.529.

The nature of the loss in serum neutralizing activity against B.1.1.529 could be discerned from our findings on a panel of monoclonal antibodies directed to the viral spike. The neutralizing activities of all four major classes of RBD monoclonal antibodies and two distinct classes of NTD monoclonal antibodies are either abolished or impaired (Fig. 2c, d). In addition to previously identified mutations that confer antibody resistance4, we have uncovered four previously undescribed spike mutations with functional consequences. Q493R confers resistance to some class 1 and class 2 RBD monoclonal antibodies; N440K and G446S confer resistance to some class 3 RBD monoclonal antibodies; and S371L confers global resistance to many RBD monoclonal antibodies through mechanisms that are not yet apparent. While performing these monoclonal antibody studies, we also observed that nearly all of the currently authorized or approved monoclonal antibody drugs are rendered weak or inactive by B.1.1.529 (Figs. 2c and 3a). In fact, the Omicron variant that contains R346K renders almost all current antibody therapy for COVID-19 ineffective (Figs. 2d and 3a).

The scientific community has chased after SARS-CoV-2 variants for a year. As more and more of them appeared, our interventions directed to the spike became increasingly ineffective. The Omicron variant has now put an exclamation mark on this point. It is not too farfetched to think that this SARS-CoV-2 is now only a mutation or two away from being pan-resistant to current antibodies, either monoclonal or polyclonal. We must devise strategies that anticipate the evolutional direction of the virus and develop agents that target better-conserved viral elements.


Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

Serum samples

Convalescent plasma samples were obtained from patients with documented SARS-CoV-2 infection. These samples were collected at the beginning of the pandemic in early 2020 at Columbia University Irving Medical Center and are therefore assumed to be infection by the WT strain of SARS-CoV-2 (ref. 4). Sera from individuals who received two or three doses of mRNA-1273 or BNT162b2 vaccine were collected at Columbia University Irving Medical Center at least two weeks after the final dose. Sera from individuals who received one dose of Ad26.COV2.S or two doses of ChAdOx1 nCov-19 were obtained from BEI Resources. Some individuals were also infected by SARS-CoV-2 in addition to the vaccinations they received. Note that, whenever possible, we specifically chose samples with high titres against the WT strain of SARS-CoV-2 such that the loss in activity against B.1.1.529 could be better quantified and the titres observed here should therefore be considered in that context. All collections were conducted under protocols reviewed and approved by the Institutional Review Board of Columbia University. All of the participants provided written informed consent. Additional information for the convalescent samples and vaccinee samples is provided in Extended Data Tables 1 and 2, respectively.

Monoclonal antibodies

Antibodies were expressed as previously described22 by synthesis of heavy chain variable (VH) and light chain variable (VL) genes (GenScript), transfection of Expi293 cells (Thermo Fisher Scientific) and affinity purification from the supernatant by rProtein A Sepharose (GE). REGN10987, REGN10933, COV2-2196 and COV2-2130 were provided by Regeneron Pharmaceuticals; Brii-196 and Brii-198 were provided by Brii Biosciences; CB6 was provided by B. Zhang and P. Kwong (NIH); and 910-30 was provided by B. DeKosky (MIT).

Cell lines

Expi293 cells were obtained from Thermo Fisher Scientific (A14527); Vero E6 cells were obtained from the ATCC (CRL-1586); HEK293T cells were obtained from the ATCC (CRL-3216); and Vero-E6-TMPRSS2 cells were obtained from JCRB (JCRB1819). Cells were purchased from authenticated vendors and morphology was confirmed visually before use. All cell lines tested mycoplasma negative.

Variant SARS-CoV-2 spike plasmid construction

An in-house high-throughput template-guide gene synthesis approach was used to generate spike genes with single or full mutations of B.1.1.529. In brief, 5′-phosphorylated oligos with designed mutations were annealed to the reverse strand of the WT spike gene construct and extended by DNA polymerase. Extension products (forward-stranded fragments) were then ligated together by Taq DNA ligase and subsequently amplified by PCR to generate variants of interest. To verify the sequences of variants, next-generation sequencing libraries were prepared according to a low-volume Nextera sequencing protocol43 and sequenced on the Illumina Miseq platform (single-end mode with 50 bp R1). Raw reads were processed using Cutadapt v.2.144 with the default settings to remove adapters, and were then aligned to reference variants sequences using Bowtie2 v.2.3.445 with the default settings. The resulting reads alignments were then visualized in the Integrative Genomics Viewer46 and manually inspected to verify the fidelity of variants. The sequences of the oligos used in variants generation are provided in Extended Data Table 3.

Pseudovirus production

Pseudoviruses were produced in the vesicular stomatitis virus (VSV) background, in which the native glycoprotein was replaced by that of SARS-CoV-2 and its variants, as previously described24. In brief, HEK293T cells were transfected with a spike expression construct with polyethylenimine (PEI) (1 mg ml−1) and cultured overnight at 37 °C under 5% CO2, and then infected with VSV-G pseudotyped ΔG-luciferase (G*ΔG-luciferase, Kerafast) 1 day after transfection. After 2 h of infection, cells were washed three times, changed to fresh medium and then cultured for approximately another 24 h before the supernatants were collected, centrifuged and aliquoted to use in assays.

Pseudovirus neutralization assay

All viruses were first titrated to normalize the viral input between assays. Heat-inactivated sera or antibodies were first serially diluted in 96 well-plates in triplicate, starting at 1:100 dilution for sera and 10 µg ml−1 for antibodies. Viruses were then added and the virus–sample mixture was incubated at 37 °C for 1 h. Vero-E6 cells (ATCC) were then added at a density of 3 × 104 cells per well and the plates were incubated at 37 °C for approximately 10 h. Luciferase activity was quantified using the Luciferase Assay System (Promega) according to the manufacturer’s instructions using SoftMax Pro v.7.0.2 (Molecular Devices). Neutralization curves and IC50 values were derived by fitting a nonlinear five-parameter dose–response curve to the data in GraphPad Prism v.9.2.

Authentic virus isolation and propagation

Authentic B.1.1.529 was isolated from a specimen from the respiratory tract of a patient with COVID-19 in Hong Kong by K.-Y. Yuen and colleagues at the Department of Microbiology, The University of Hong Kong. Isolation of WT SARS-CoV-2 was previously described47. Viruses were propagated in Vero-E6-TMPRSS2 cells and the sequence was confirmed by next-generation sequencing before use.

Authentic virus neutralization assay

To measure neutralization of authentic SARS-CoV-2 viruses, Vero-E6-TMPRSS2 cells were first seeded in 96-well plates in cell culture medium (Dulbecco’s Modified Eagle Medium (DMEM) + 10% fetal bovine serum (FBS) + 1% penicillin–streptomycin) overnight at 37 °C under 5% CO2 to establish a monolayer. The next day, sera or antibodies were serially diluted in 96-well plates in triplicate in DMEM + 2% FBS and then incubated with 0.01 multiplicity of infection of WT SARS-CoV-2 or B.1.1.529 at 37 °C for 1 h. Sera were diluted from 1:100 dilution and antibodies were diluted from 10 µg ml−1. Next, the mixture was overlaid onto cells and further incubated at 37 °C under 5% CO2 for approximately 72 h. Cytopathic effects were then scored by plaque assay in a blinded manner. Neutralization curves and IC50 values were derived by fitting a nonlinear five-parameter dose–response curve to the data in GraphPad Prism v.9.2.

Antibody footprint analysis and RBD mutagenesis analysis

The SARS-CoV-2 spike structure used for displaying epitope footprints and mutations within emerging strains was downloaded from the Protein Data Bank (PDB: 6ZGE). The structures of antibody-spike complexes were also obtained from PDB (7L5B (2-15), 6XDG (REGN10933 and REGN10987), 7L2E (4-18), 7RW2 (5–7), 7C01 (CB6), 7KMG (LY-COV555), 7CDI (Brii-196), 7KS9 (910-30), 7LD1 (DH1047), 7RAL (S2X259), 7LSS (2–7) and 6WPT (S309)). Interface residues were identified using PISA48 using the default parameters. The footprint for each antibody was defined by the boundaries of all epitope residues. The border for each footprint was then optimized by ImageMagick v.7.0.10-31 ( PyMOL v.2.3.2 was used to perform mutagenesis and to generate structural plots (Schrödinger).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.