An infectious SARS-CoV-2 B.1.1.529 Omicron virus escapes neutralization by therapeutic monoclonal antibodies.

The emergence of the highly transmissible B.1.1.529 Omicron variant of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is concerning for antibody countermeasure efficacy because of the number of mutations in the spike protein. In this study, we tested a panel of anti-receptor-binding domain monoclonal antibodies (mAbs) corresponding to those in clinical use by Vir Biotechnology (S309, the parent mAb of VIR-7831 (sotrovimab)), AstraZeneca (COV2-2196 and COV2-2130, the parent mAbs of AZD8895 and AZD1061), Regeneron (REGN10933 and REGN10987), Eli Lilly (LY-CoV555 and LY-CoV016) and Celltrion (CT-P59) for their ability to neutralize an infectious B.1.1.529 Omicron isolate. Several mAbs (LY-CoV555, LY-CoV016, REGN10933, REGN10987 and CT-P59) completely lost neutralizing activity against B.1.1.529 virus in both Vero-TMPRSS2 and Vero-hACE2-TMPRSS2 cells, whereas others were reduced (COV2-2196 and COV2-2130 combination, ~12-fold decrease) or minimally affected (S309). Our results suggest that several, but not all, of the antibodies in clinical use might lose efficacy against the B.1.1.529 Omicron variant.

Main

Since December 2019, the global Coronavirus Disease 2019 (COVID-19) pandemic caused by SARS-CoV-2 has resulted in 298 million infections and 5.4 million deaths. The expansion of the COVID-19 pandemic and its accompanying morbidity, mortality and destabilizing socioeconomic effects have made the development and distribution of SARS-CoV-2 therapeutics and vaccines an urgent global health priority[1]. Although the rapid deployment of countermeasures, including mAbs and multiple highly effective vaccines, has provided hope for curtailing disease and ending the pandemic, this has been jeopardized by the emergence of more transmissible variants with mutations in the spike protein that also could evade protective immune responses.

Indeed, over the past year, several variant strains have emerged, including B.1.1.7 (Alpha), B.1.351 (Beta), B.1.1.28 (also called P.1, Gamma) and B.1.617.2 (Delta), among others, each having varying numbers of substitutions in the N-terminal domain (NTD) and the receptor-binding domain (RBD) of the SARS-CoV-2 spike. Cell-based assays with pseudoviruses or authentic SARS-CoV-2 strains suggest that neutralization by many Emergency Use Authorization (EUA) mAbs might be diminished against some of these variants, especially those containing mutations at positions L452, K477 and E484 (refs. [2-6]). Notwithstanding this, in vivo studies in animals showed that, when most EUA mAbs were used in combination, they retained efficacy against different variants[7]. The recent emergence of B.1.1.529, the Omicron variant[8,9], which has a larger number of mutations (>30 substitutions, deletions or insertions) in the spike protein, has raised concerns that this variant will escape from protection conferred by vaccines and therapeutic mAbs.

Results

We obtained an infectious clinical isolate of B.1.1.529 from a symptomatic individual in the United States (hCoV-19/USA/WI-WSLH-221686/2021). We propagated the virus once in Vero cells expressing human transmembrane protease serine 2 (TMPRSS2) to prevent the emergence of adventitious mutations at or near the furin cleavage site in the spike protein[10]. Our B.1.1.529 isolate encodes the following mutations in the spike protein (A67V, Δ69−70, T95I, G142D, Δ143-145, Δ211, L212I, insertion 214EPE, G339D, S371L, S373P, S375F, K417N, N440K, G446S, S477N, T478K, E484A, Q493R, G496S, Q498R, N501Y, Y505H, T547K, D614G, H655Y, N679K, P681H, N764K, D796Y, N856K, Q954H, N969K and L981F; Fig. 1a,b and GISAID: EPI_ISL_7263803), which is similar to strains identified in Africa[11]. Our isolate, however, lacks an R346K mutation, which is present in a minority (~8%) of reported strains.

Fig. 1

Neutralizing mAb epitopes on B.1.1.529.

a, b, SARS-CoV-2 spike trimer (PDB: 7C2L and PDB: 6W41). One spike protomer is highlighted, showing the NTD in orange, RBD in green, RBM in magenta and S2 portion of the molecule in blue (a). Close-up view of the RBD with the RBM outlined in magenta (b). Amino acids that are changed in B.1.1.529 compared to WA1/2020 are indicated in light green (a, b), with the exception of N679K and P681H, which were not modeled in the structures used. c–k, SARS-CoV-2 RBD bound by EUA mAbs COV2-2196 (c, PDB: 7L7D); COV2-2130 (d, PDB: 7L7E); S309 (e, PDB: 6WPS); REGN10987 (f, PDB: 6XDG); REGN10933 (g, PDB: 6XDG); LY-CoV555 (h, PDB: 7KMG); LY-CoV016 (i, PDB: 7C01); CT-P59 (j PDB: 7CM4); and SARS2-38 (k, PDB: 7MKM). Residues mutated in the B.1.1.529 RBD and contained in these mAbsʼ respective epitopes are shaded red, whereas those outside the epitope are shaded green. l, Multiple sequence alignment showing the epitope footprints of each EUA mAb on the SARS-CoV-2 RBD highlighted in cyan. B.1.1.529 RBD is shown in the top row, with sequence changes relative to the wild-type RBD highlighted red. A green diamond indicates the location of the N-linked glycan at residue 343. Stars below the alignment indicate hACE2 contact residues on the SARS-CoV-2 RBD[40].

Source data

Given the number of substitutions in the B.1.1.529 spike protein, including eight amino acid changes (K417N, G446S, S477N, Q493R, G496S, Q498R, N501Y and Y505H) in the ACE2 receptor-binding motif (RBM), we first evaluated possible effects on the structurally defined binding epitopes[12,13] of mAbs corresponding to those with EUA approval or in advanced clinical development (S309 (parent of VIR-7831 (sotrovimab)), RBD group III)[14,15]; COV2-2196 (RBD group I) and COV2-2130 (RBD group III) (parent mAbs of AZD8895 and AZD1061, respectively)[16]; REGN10933 (RBD group I) and REGN10987 (RBD group III)[17]; LY-CoV555 (RBD group I) and LY-CoV016 (RBD group I)[18,19]; and CT-P59 (Celltrion, RBD group I)[20], along with an additional broadly neutralizing mAb (SARS2-38 (RBD group II)) that we recently described[21]. We mapped the B.1.1.529 spike mutations onto the antibody-bound SARS-CoV-2 spike or RBD structures published in the RCSB Protein Data Bank (PDB) (Fig. 1c–k). Although every antibody analyzed had structurally defined recognition sites that were altered in the B.1.1.529 spike, the differences varied among mAbs, with some showing larger numbers of changed residues (Fig. 1l: COV2-2196, n = 5; COV2-2130, n = 4; S309, n = 2; REGN10987, n = 4; REGN10933, n = 8; Ly-CoV555, n = 2; Ly-CoV016, n = 6; CT-P59, n = 8; and SARS2-38, n = 2).

To address the functional significance of the spike sequence variation in B.1.1.529 for antibody neutralization, we used a high-throughput focus reduction neutralization test (FRNT)[22] with WA1/2020 D614G and B.1.1.529 in Vero-TMPRSS2 cells (Fig. 2). We tested individual mAbs and combinations of mAbs that target the RBD in Vero-TMPRSS2 cells, including S309 (Vir Biotechnology); COV2-2130/COV2-2196 (parent mAbs of AZD1061 and AZD8895, provided by Vanderbilt University Medical Center); REGN10933/REGN10987 (synthesized based on casirivimab and imdevimab sequences from Regeneron); LY-CoV555/LY-CoV016 (synthesized based on bamlanivimab and etesevimab sequences from Eli Lilly); CT-P59 (synthesized based on regdanvimab sequences from Celltrion); and SARS2-38. As expected, all individual mAbs or combinations of mAbs tested neutralized the WA1/2020 D614G isolate, with half-maximal inhibitory concentration (EC50) values similar to published data[6,20,23]. However, when tested alone, REGN10933, REGN10987, LY-CoV555, LV-CoV016, CT-P59 and SARS2-38 completely lost neutralizing activity against B.1.1.529, with little inhibitory capacity even at the highest (10,000 ng ml−1) concentration tested. COV2-2130 and COV2-2196 showed an intermediate ~12-fold and 150-fold (P < 0.0001) loss in inhibitory activity, respectively, against the B.1.1.529 strain. In comparison, S309 showed a less than two-fold (P > 0.5) reduction in neutralizing activity against B.1.1.529 (Fig. 2a–h). Analysis of mAb combinations currently in clinical use showed that REGN10933/REGN10987 and LY-CoV555/LV-CoV016 lost all neutralizing activity against B.1.1.529, whereas COV2-2130/COV2-2196 showed a ~12-fold (P < 0.0001) reduction in inhibitory activity from an EC50 of 12 ng ml−1 to 147 ng ml−1.

Fig. 2

Neutralization of SARS-CoV-2 B.1.1.529 Omicron strain by mAbs in Vero-TMPRSS2 cells.

a–f, Neutralization curves in Vero-TMPRSS2 cells comparing the sensitivity of SARS-CoV-2 strains with the indicated mAbs (COV2-2196, COV2-2130, REGN10933, REGN10987, LY-CoV555, LY-CoV016, S309, CT-P59 and SARS2-38) with WA1/2020 D614G and B.1.1.529. Also shown are the neutralization curves for antibody cocktails (COV2-2196/COV2-2130, REGN10933/REGN10987 or LY-CoV555/LY-CoV016). For data with mAb combinations, the x axis represents the total concentration of mAb used. One representative experiment of three performed in technical duplicate is shown. Error bars indicate the range of technical replicates. Data (% relative infection) are normalized to a no-mAb control. g, Summary of EC50 values (ng ml−1) of neutralization of SARS-CoV-2 viruses (WA1/2020 D614G and B.1.1.529) performed in Vero-TMPRSS2 cells. Data are the geometric mean of three experiments. Blue shading: light, EC50 > 5,000 ng ml−1; dark, EC50 > 10,000 ng ml−1. h, Comparison of EC50 values by mAbs against WA1/2020 D614G and B.1.1.529 (three experiments; NS, not significant; ****P < 0.0001; two-way ANOVA with Sidak’s post test). Each symbol represents neutralization data from an individual experiment. Bars indicate mean values. The dotted line indicates the upper limit of dosing of the assay.

Source data

We repeated experiments in Vero-hACE2-TMPRSS2 cells to account for effects of hACE2 expression, which can affect neutralization by some anti-SARS-CoV-2 mAbs[21,24]. Moreover, modeling studies suggest that the mutations in the B.1.1.529 spike might enhance interactions with hACE2 (ref. [25]). All individual mAbs or combinations of mAbs tested neutralized the WA1/2020 D614G isolate as expected. However, REGN10933, REGN10987, LY-CoV555, LV-CoV016, SARS2-38 and CT-P59 completely lost neutralizing activity against B.1.1.529, and the combinations of REGN10933/REGN10987 or LY-CoV555/LV-CoV016 also lacked inhibitory capacity (Fig. 3a–h). In comparison, COV2-2130 and COV2-2196 showed reduced activity (~12-fold and 16-fold, respectively, P < 0.0001) against B.1.1.529, as did the combination of COV2-2130/COV2-2196 mAbs (~11-fold, P < 0.0001). The S309 mAb exhibited less potent neutralizing activity in Vero-hACE2-TMPRSS2 cells against WA1/2020 D614G virus with a flatter dose–response curve (Fig. 3d), as seen previously[6,26], and showed a moderate (~six-fold, P < 0.0001) reduction in neutralizing activity against B.1.1.529. Thus, although the trends in mAb neutralization of B.1.1.529 generally were similar to Vero-TMPRSS2 cells, some minor differences in potency were noted in cells expressing hACE2.

Fig. 3

Neutralization of SARS-CoV-2 B.1.1.529 Omicron strain by mAbs in Vero-hACE2-TMPRSS2 cells.

a–f, Neutralization curves in Vero-hACE2-TMPRSS2 cells comparing the sensitivity of SARS-CoV-2 strains with the indicated mAbs (S309, COV2-2196, COV2-2130, REGN10933, REGN10987, LY-CoV555, LY-CoV016, CT-P59 and SARS2-38) with WA1/2020 D614G and B.1.1.529. Also shown are the neutralization curves for antibody cocktails (COV2-2196/COV2-2130, REGN10933/REGN10987 or LY-CoV555/LY-CoV016). For data with mAb combinations, the x axis represents the total concentration of mAb used. One representative experiment of three performed in technical duplicate is shown. Error bars indicate range of technical replicates. Data (% relative infection) are normalized to a no-mAb control. g, Summary of EC50 values (ng ml−1) of neutralization of SARS-CoV-2 viruses (WA1/2020 D614G and B.1.1.529) performed in Vero-hACE2-TMPRSS2 cells. Data are the geometric mean of three experiments. Blue shading: light, EC50 > 5,000 ng ml−1; dark, EC50 > 10,000 ng ml−1. h, Comparison of EC50 values by mAbs against WA1/2020 D614G and B.1.1.529. i, j, Neutralization curves in Vero-hACE2-TMPRSS2 cells comparing WA1/2020 D614G and B.1.1.529 infection in the presence of AZD1061, AZD8895 and the combination AZD7442. h, j, Three experiments; ****P < 0.0001; two-way ANOVA with Sidak’s post test. Each symbol represents neutralization data from an individual experiment. Bars indicate mean values. The dotted line indicates the upper limit of dosing of the assay.

Discussion

Our experiments show a marked loss of inhibitory activity by several of the most highly neutralizing mAbs that are in advanced clinical development or have EUA approval. We evaluated antibodies that correspond to monotherapy or combination therapy that have shown pre- and post-exposure success in clinical trials and in patients infected with historical SARS-CoV-2 isolates. Our results confirm in silico predictions of how amino acid changes in B.1.1.529 RBD might negatively affect neutralizing antibody interactions[18,27]. Moreover, they agree with preliminary studies showing that several clinically used antibodies lose neutralizing activity against B.1.1.529 spike-expressing recombinant lentiviral or vesicular stomatitis virus (VSV)-based pseudoviruses[28-30]. One difference is that our study with authentic B.1.1.529 showed only moderately reduced neutralization by antibodies corresponding to the AstraZeneca combination (COV2-2196 and COV2-2130); in contrast, another group reported escape of these mAbs using a VSV pseudovirus displaying a B.1.1.529 spike protein in Huh7 hepatoma cells[29]. Additional studies are needed to determine whether this disparity in results is due to the cell type, the virus (authentic versus pseudotype) or preparation and combination of antibody. To begin to address this issue, we obtained AZD1061, AZD8895 and the combination AZD7442 directly from the manufacturer and tested them for neutralization of WA1/2020 D614G and B.1.1529 in Vero-hACE2-TMPRSS2 cells. We observed relatively similar reductions in inhibitory activity compared to the preclinical COV2-2130 and COV2-2196 mAbs with 49-, 92- and 33-fold lower EC50 values against B.1.1.529 by AZD1061, AZD8895 and AZD7442, respectively (Fig. 3g,i–j).

Although the Regeneron (REGN10933 and REGN10987), Eli Lilly (LY-CoV555 and LV-CoV016) and Celltrion (CT-P59) antibodies or combinations showed an almost complete loss of neutralizing activity against B.1.1.529, in our assays with Vero-TMPRSS2 and Vero-hACE2-TMPRSS2 cells the mAbs corresponding to the AstraZeneca combination (COV-2196 and COV-2130) or Vir Biotechnology (S309) products retained substantial inhibitory activity. Although these data suggest that some mAbs in clinical use might retain benefit, validation experiments in vivo[7] are needed to support this conclusion and inform clinical decisions.

Given the loss of inhibitory activity against B.1.1.529 of many highly neutralizing anti-RBD mAbs in our study, it appears likely that serum polyclonal antibody responses generated after vaccination or natural infection also might lose substantial inhibitory activity against B.1.1.529, which could compromise protective immunity and explain a rise in symptomatic infections in vaccinated individuals[31]. Indeed, studies have reported approximately 25-fold to 40-fold reductions in serum neutralizing activity compared to historical D614G-containing strains from individuals immunized with the Pfizer BNT162b2 and AstraZeneca AZD1222 vaccines[28,30,32,33].

We note several limitations of our study. (1) Our experiments focused on the effect of the extensive sequence changes in the B.1.1.529 spike protein on mAb neutralization in cell culture. Despite observing differences in neutralizing activity with certain mAbs, it remains to be determined how this finding translates into effects on clinical protection against B.1.1.529. (2) Although virus neutralization is a correlate of immune protection against SARS-CoV-2 (refs. [7,34,35]), this measurement does not account for Fc effector functions if antibodies residually bind B.1.1.529 spike proteins on the virion or surface of infected cells. Fcγ receptor or complement protein engagement by spike-binding antibodies could confer substantial protection[36-38]. It should be noted that some antibodies have been engineered to have reduced Fc effector binding/function (e.g., the clinical antibodies AZD1061 and AZD8895). (3) We used the prevailing B.1.1.529 Omicron isolate that lacks an R346K mutation. Although only 8.3% of B.1.1.529 sequences in GISAID (accessed on 14 December 2021) have an R346K mutation, this substitution might further affect neutralization of some of the clinically used mAbs given that R346 is a contact residue for COV2-2130, REGN10987 and S309 (Fig. 1l). At least for S309, the R346K mutation did not affect neutralization of pseduoviruses displaying B.1.1.529 spike proteins[30]. Nonetheless, studies with infectious B.1.1.529 isolates with R346K mutations might be warranted if the substitution becomes more prevalent. (4) Our data are derived from experiments with Vero-TMPRSS2 and Vero-hACE2-TMPRSS2 cells. Although these cells standardly are used to measure antibody neutralization of SARS-CoV-2 strains, primary cells targeted by SARS-CoV-2 in vivo can express unique sets of attachment and entry factors[39], which could affect receptor and entry blockade by specific antibodies. Indeed, previous studies have reported that the cell line used can affect the potency of antibody neutralization against different SARS-CoV-2 variants[6].

In summary, our cell-culture-based analysis of neutralizing mAb activity against an authentic infectious B.1.1.529 Omicron SARS-CoV-2 isolate suggests that several, but not all, existing therapeutic antibodies will lose protective benefit. Thus, the continued identification and use of broadly and potently neutralizing mAbs that target the most highly conserved residues on the SARS-CoV-2 spike likely is needed to prevent resistance against B.1.1.529 and future variants with highly mutated spike sequences.

Methods

Cells

Vero-TMPRSS2 (ref. [41]) and Vero-hACE2-TMPRSS2 (ref. [6]) cells were cultured at 37 °C in DMEM supplemented with 10% FBS, 10 mM HEPES pH 7.3 and 100 U ml−1 of penicillin–streptomycin. Vero-TMPRSS2 cells were supplemented with 5 μg ml−1 of blasticidin. Vero-hACE2-TMPRSS2 cells were supplemented with 10 µg ml−1 of puromycin. All cells routinely tested negative for mycoplasma using a PCR-based assay.

Viruses

The WA1/2020 recombinant strain with substitutions (D614G) was described previously[42]. The B.1.1.529 isolate (hCoV-19/USA/WI-WSLH-221686/2021) was obtained from an individual in Wisconsin as a midturbinate nasal swab and passaged once on Vero-TMPRSS2 cells[43]. All viruses were subjected to next-generation sequencing (GISAID: EPI_ISL_7263803) to confirm the stability of substitutions. All virus experiments were performed in an approved Biosafety Level 3 facility.

Monoclonal antibody purification

The mAbs used in this study (COV2-2196, COV2-2130, S309, REGN10933, REGN10987, LY-CoV555, LY-CoV016, CT-P59, SARS2-38, AZD1061, AZD8895 and AZD7442) have been described previously[14,17,21,44-48]. S309 is the parent of VIR-7831 (sotrovimab); the clinically used mAb is engineered for enhanced clinical developability, as reported previously[23]. COV2-2196 and COV2-2130 mAbs were produced after transient transfection using the Gibco ExpiCHO Expression System (Thermo Fisher Scientific) following the manufacturer’s protocol. Culture supernatants were purified using HiTrap MabSelect SuRe columns (Cytiva, formerly GE Healthcare Life Sciences) on an ÄKTA Pure chromatographer (GE Healthcare Life Sciences). Purified mAbs were buffer exchanged into PBS, concentrated using Amicon Ultra-4 50-kDa centrifugal filter units (Millipore Sigma) and stored at −80 °C until use. Purified mAbs were tested for endotoxin levels (found to be less than 30 endotoxin units (EU) per milligram IgG). Endotoxin testing was performed using the PTS201F cartridge (Charles River Laboratories), with a sensitivity range from 10 to 0.1 EU per milliliter, and an Endosafe Nexgen-MCS instrument (Charles River Laboratories). S309, REGN10933, REGN10987, LY-CoV016, LY-CoV555, CT-P59 and SARS2-38 mAb proteins were produced in CHOEXPI or EXPI293F cells and affinity purified using HiTrap Protein A columns (GE Healthcare, HiTrap mAb select Xtra no. 28-4082-61). Purified mAbs were suspended into 20 mM histidine, 8% sucrose pH 6.0 or PBS. The final products were sterilized by filtration through 0.22-μm filters and stored at 4 °C.

FRNT

Serial dilutions of mAbs were incubated with 102 focus-forming units of SARS-CoV-2 (WA1/2020 D614G or B.1.1.529) for 1 h at 37 °C. Antibody–virus complexes were added to Vero-TMPRSS2 or Vero-hACE2-TMPRSS2 cell monolayers in 96-well plates and incubated at 37 °C for 1 h. Subsequently, cells were overlaid with 1% (wt/vol) methylcellulose in MEM. Plates were harvested at 30 h (WA1/2020 D614G on Vero-TMPRSS2 cells), 70 h (B.1.1.529 on Vero-TMPRSS2 cells) or 24 h (both viruses on Vero-hACE2-TMPRSS2 cells) later by removal of overlays and fixation with 4% paraformaldehyde in PBS for 20 min at room temperature. A longer time of incubation was required for B.1.1.529-infected Vero-TMPRSS2 cells because the foci were smaller at the time point and difficult to quantitate. Plates with WA1/2020 D614G were washed and sequentially incubated with an oligoclonal pool (1 μg ml−1 of each) of SARS2-2, SARS2-11, SARS2-16, SARS2-31, SARS2-38, SARS2-57 and SARS2-71 (ref. [49]) anti-S antibodies. Plates with B.1.1.529 were additionally incubated with a pool of mAbs that cross-react with SARS-CoV-1 and bind a CR3022-competing epitope on the RBD[21]. All plates were subsequently stained with HRP-conjugated goat anti-mouse IgG (Sigma-Aldrich, A8924, 1:1,000) in PBS supplemented with 0.1% saponin and 0.1% BSA. SARS-CoV-2-infected cell foci were visualized using KPL TrueBlue peroxidase substrate and quantitated on an ImmunoSpot microanalyzer (Cellular Technologies). Data (% relative infection) are normalized to a no-mAb control. Antibody dose–response curves were analyzed using non-linear regression analysis with a variable slope (GraphPad Software), and the EC50 was calculated.

Model of mAb-B.1.1.529 spike complexes

The spike model is a composite of data from PDB: 7C2L and PDB: 6W41. Models of mAb complexes were generated from their respective PDB files with the following accession codes: COV2-2196 (PDB: 7L7D); COV2-2130 (PDB: 7L7E); S309 (PDB: 6WPS); REGN10987 (PDB: 6XDG); REGN10933 (PDB: 6XDG); LY-CoV555 (PDB: 7KMG); LY-CoV016 (PDB: 7C01); CT-P59 (PDB: 7CM4); and SARS2-38 (PDB: 7MKM). Epitope footprints used in the multiple sequence alignment were determined using PISA interfacial analysis on the various mAb:RBD complexes[50]. Structural figures were generated using UCSF ChimeraX[51].

Reagent availability

All reagents described in this paper are available through material transfer agreements. AZD8895 and AZD1061 may be obtained from AstraZeneca for non-commercial internal research purposes under material transfer agreements upon reasonable request.

Statistical analysis

The number of independent experiments and technical replicates used are indicated in the relevant figure legends. A two-way ANOVA with Sidak’s post test was used for comparisons of antibody potency between WA1/2020 D614G and B.1.1.59.

Reporting Summary

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

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 10.1038/s41591-021-01678-y.

Reporting Summary