Main

Variant strains of SARS-CoV-2 have been detected in the UK (B.1.1.7, also known as Alpha), South Africa (B.1.351, also known as Beta), Brazil (B.1.1.28 (also known as P.1, also known as Gamma)) and elsewhere that contain substitutions in the N-terminal domain and the receptor-binding motif of the receptor-binding domain (RBD). Cell-based assays suggest that neutralization by many antibodies may be diminished against variants that express spike mutations, especially at position E4841,2,3,4,5. However, the in vivo implications of this loss of mAb neutralizing activity remains uncertain, particularly forcombination mAb therapies.

To evaluate the effects of SARS-CoV-2 strain variation on mAb protection, we assembled a panel of infectious SARS-CoV-2 strains with sequence substitutions in the spike gene (Fig. 1a, b) including a B.1.1.7 isolate from the UK, a B.1.429 isolate from California (USA), a B.1.617.1 isolate (of a clade identified in India) and two B.1.526 isolates from New York (USA). We also used SARS-CoV-2 strains from Washington (USA) with a D614G substitution (WA1/2020 D614G) or with both N501Y and D614G substitutions (WA1/2020 N501Y/D614G) as well as chimeric SARS-CoV-2 strains with B.1.351 or B.1.1.28 spike genes in the Washington strain background (denoted Wash-B.1.351 and Wash-B.1.1.28, respectively)1,6. All viruses were propagated in Vero cells expressing transmembrane protease serine 2 (Vero-TMPRSS2 cells) to prevent the emergence of mutations at or near the furin cleavage site in the spike protein that affect virulence7. All viruses were deep-sequenced to confirm the presence of expected mutations before use (Supplementary Table 1).

Fig. 1: Neutralization of SARS-CoV-2 variant strains by clinically relevant mAbs.
figure 1

a, b, Amino acid substitutions in SARS-CoV-2 variants mapped onto the structure of the spike protein. Schematic layout of the spike protein monomer is depicted at the top. Structure of spike monomer (Protein Data Bank code (PDB) 7C2L, with RBD from PDB 6W41) is depicted as a cartoon, with the N-terminal domain (NTD), RBD, receptor-binding motif (RBM) and S2 coloured in orange, green, magenta and light blue, respectively. Substitutions for each variant are shown as spheres and coloured according to the legend. Substitutions shown in black are shared between several variants. The purple triangle, pink square, purple hexagon and black pentagon represent approximate locations of L5, S13, D253 and P681, respectively, which were not modelled in the original structures. CH, central helix; FP, fusion peptide; HR, heptad repeat; TM, transmembrane domain. The structural figure in a was generated using UCSF ChimeraX33. b, Viruses used with indicated coloured mutations in the spike protein. c, Summary of EC50 values (ng ml−1) of neutralization of SARS-CoV-2 viruses performed in Vero-TMPRSS2 cells. Blue shading of cells indicates a partial (EC50 > 1,000 ng ml−1) or complete (EC50 > 10,000 ng ml−1) loss of neutralizing activity. d, Neutralization curves comparing the sensitivity of SARS-CoV-2 strains to the indicated individual or combinations of mAbs. Data are representative of two to five experiments, each performed in technical duplicate.

Source data

We first assessed the effect of SARS-CoV-2 spike variation on antibody neutralization (Fig. 1c, d). We tested individual mAbs and cocktails of mAbs in clinical development that target the RBD, including 2B04 and 47D11 (AbbVie), S309 and S2E12 (Vir Biotechnology), COV2-2130 and COV2-2196 (Vanderbilt University Medical Center, with derivatives being evaluated by AstraZeneca), REGN10933 and REGN10987 (synthesized on the basis of casirivimab and imdevimab sequences from Regeneron), and LY-CoV555 (synthesized on the basis of bamlanivimab sequences from Lilly). All individual mAbs tested efficiently neutralized the WA1/2020 D614G, WA1/2020 N501Y/D614G and B.1.1.7 strains, and several mAbs (COV2-2130, COV2-2196, S309, S2E12 and 47D11) showed little change in potency against the Wash-B.1.351, Wash-B.1.1.28, B.1.429 and B.1.526 strains (Fig. 1c, d). By comparison, REGN10987 and LY-CoV555 showed an approximately 10-fold and complete loss, respectively, in inhibitory activity against the B.1.429 and B.1.617.1 strains, which is consistent with previous studies that have identified L452 and adjacent residues as interaction sites for these mAbs8 (Supplementary Table 2). Moreover, REGN10933, LY-CoV555 and 2B04 exhibited a substantial or complete loss of neutralizing activity against Wash-B.1.351, Wash-B.1.1.28, B.1.617.1 and B.1.526 (E484K) viruses that contain mutations at residue E484 (Fig. 1c, d, Extended Data Fig. 1), which corresponds with structural and mapping studies (Supplementary Table 2). Analysis of mAb cocktails showed that COV2-2130 in combination with COV2-2196 (COV2-2130/COV2-2196), S309/S2E12 and REGN10933/REGN10987 neutralized all of the virus strains we tested; the last of these combinations retained potency corresponding to the mAb with inhibitory activity in the cocktail for a given virus. The 2B04/47D11 mAb combination efficiently neutralized WA1/2020 D614G, WA1/2020 N501Y/D614G, B.1.1.7 and B.1.429 strains, whereas the activity of this combination against Wash-B.1.351, Wash-B.1.1.28, B.1.617.1 and B.1.526 (E484K) reflected the less-potent 47D11 mAb component (half-maximal effective concentration (EC50) of 384–2,187 ng ml−1) (Fig. 1c, d). Additional mutations in B.1.617.1 decreased the potency of the 2B04/47D11 combination further. By contrast, almost all of the mAbs retained neutralizing potency against B.1.526 (S477N).

Prophylactic efficacy against variants

To evaluate the efficacy of the mAb combinations in vivo, we initially used K18-hACE2 mice in which human ACE2 expression is driven by the cytokeratin 18 gene promoter9,10. In previous studies it was established that low (2 mg per kg body weight (mg kg−1)) doses of several anti-RBD neutralizing human mAbs provide a threshold of protection when administered as prophylaxis11. Accordingly, we gave K18-hACE2 mice a single 40-μg (about 2 mg kg−1 total) dose of the mAb combinations (2B04/47D11, S309/S2E12, COV2-2130/COV2-2196 or REGN10933/REGN10987) or LY-CoV555 as monotherapy by intraperitoneal injection one day before intranasal inoculation with SARS-CoV-2 (WA1/2020 N501Y/D614G, B.1.1.7, Wash-B.1.351 or Wash-B.1.1.28). For these in vivo studies, we used a recombinant version of WA1/2020 that encodes N501Y for direct comparison to B.1.1.7, Wash-B.1.351 or Wash-B.1.1.28 (all of which contain this residue). This substitution increases infection in mice12,13, but did not substantively affect neutralization of the mAbs we tested (Fig. 1c).

Compared to a control human mAb, a single 40-μg prophylaxis dose of the anti-SARS-CoV-2 mAbs conferred substantial protection against WA1/2020-N501Y/D614G-induced weight loss and viral burden in the lungs, nasal washes, brain, spleen and heart in the K18-hACE2 mice at 6 days after infection (Fig. 2a–c, Extended Data Figs. 2, 3a). Although all of the anti-SARS-CoV-2 mAb cocktails protected against weight loss caused by B.1.1.7 (Fig. 2d), Wash-B.1.351 (Fig. 2g) or Wash-B.1.1.28 (Fig. 2j), LY-CoV555 monotherapy protected only against the B.1.1.7 strain (Fig. 2d, g, j). Some of the antibodies provided less virological protection against the B.1.1.7 (Fig. 2e, f), Wash-B.1.351(Fig. 2h, i) or Wash-B.1.1.28 (Fig. 2k, l) strains in specific tissues. Whereas all mAb groups protected against B.1.1.7 infection in the lung (Fig. 2e), 2B04/47D11 or LY-CoV555 did not perform as well in nasal washes (Fig. 2f), and LY-CoV555 showed reduced protection in the brain (Extended Data Fig. 2). Sanger sequencing analysis of the RBD region of viral RNA of brain, nasal wash and lung samples from mice treated with these mAbs did not show evidence of neutralization escape (Supplementary Table 3). Mice treated with 2B04/47D11 or LY-CoV555 also showed greater virus breakthrough than those treated with the other tested antibodies when challenged with Wash-B.1.351 (Fig. 2h, i) or Wash-B.1.1.28 (Fig. 2k, l) viruses: 2B04/47D11 reduced viral burden in the lungs, nasal washes and brain much less efficiently than the other mAb cocktails, and LY-CoV555 mAb treatment conferred virtually no protection in any tissue analysed (Fig. 2h, i, k, l, Extended Data Figs. 2, 3b). Compared to COV2-2130/COV2-2196 and S309/S2E12, REGN10933/REGN10987 also showed less ability to reduce viral RNA levels in nasal washes of mice infected with Wash-B.1.351 (Fig. 2i) or Wash-B.1.1.28 (Fig. 2l) viruses. To confirm that our findings with Wash-B.1.351 are similar to a bona fide B.1.351 strain, we tested mAbs from each cocktail for neutralization and the COV2-2130/COV2-2196 cocktail for protection in K18-hACE2 mice. Equivalent levels of neutralization and viral burden reduction were seen with B.1.351 and Wash-B.1.351 viruses (Extended Data Fig. 4).

Fig. 2: Antibody prophylaxis against SARS-CoV-2 variants in K18-hACE2 mice.
figure 2

al, Eight-to-ten-week-old female and male K18-hACE2 transgenic mice received 40 μg (about 2 mg kg−1) of the indicated mAb treatment by intraperitoneal injection one day before intranasal inoculation with 103 FFU of the SARS-CoV-2 N501Y/D614G (ac), B.1.1.7 (df), Wash-B.1.351 (gi) or Wash-B.1.1.28 (jl) strains. Tissues were collected at six days after infection. a, d, g, j, Weight change after infection with SARS-CoV-2 (mean ± s.e.m.; n = 6–12 mice per group, two experiments; one-way analysis of variance (ANOVA) with Dunnett’s test of area under the curve. NS, not significant; ****P < 0.0001). Viral RNA levels in the lung (b, e, h, k) and nasal washes (c, f, i, l) were measured (line indicates median; in order from left to right, n = 9, 6, 7, 6, 6 and 6 (a); n = 11, 5, 5, 6, 6 and 6 (b); n = 9, 5, 5, 6, 6 and 6 (c); n = 12, 6, 6, 6, 5 and 6 (df); n = 12, 6, 6, 6, 6 and 6 (g, h, j, k); n = 12, 6, 6, 6, 6 and 5 (i); n = 10, 6, 6, 6, 6 and 5 (l) mice per group, two experiments; one-way ANOVA with Dunnett’s test with comparison to control mAb. NS, not significant, ****P < 0.0001; *P = 0.026 (c); **P = 0.0016, ***P = 0.0002 (l)). Dotted line indicates the limit of detection of the assay.

Source data

To evaluate further the extent of protection conferred by the different mAb groups against the SARS-CoV-2 variant viruses, we measured pro-inflammatory cytokine and chemokines in lung homogenates collected at six days after infection (Extended Data Figs. 5, 6). This analysis showed a strong correspondence with viral RNA levels in the lung: (1) compared to the control mAb, S309/S2E12, COV2-2130/COV2-2196 and REGN10933/REGN10987 combinations showed markedly reduced levels of pro-inflammatory cytokines and chemokines (G-CSF, IFNγ, IL-6, CXCL10, LIF, CCL2, CXCL9, CCL3 and CCL4) after infection with WA1/2020 N501Y/D614G, B.1.1.7, Wash-B.1.351 or Wash-B.1.1.28; and (2) prophylaxis with 2B04/47D11 or LY-CoV555 resulted in reduced inflammatory cytokine and chemokine levels in mice infected with WA1/2020 N501Y/D614G and B.1.1.7, with less improvement in mice infected with Wash-B.1.351 or Wash-B.1.1.28.

Given that a 40-μg dose of the S309/S2E12, COV2-2130/COV2-2196 and REGN10933/REGN10987 combinations prevented infection and inflammation caused by the different SARS-CoV-2 strains, we tested a tenfold-lower (4 μg) dose (about 0.2 mg kg−1) to assess for possible differences in protection. Prophylaxis with COV2-2130/COV2-2196, S309/S2E12, REGN10933/REGN10987 or 2B04/47D11 protected K18-hACE2 mice against weight loss caused by all four viruses (Extended Data Fig. 7a–d). Whereas the COV2-2130/COV2-2196, S309/S2E12 and REGN10933/REGN10987 mAb combinations reduced viral RNA levels in the lung at six days after infection in K18-hACE2 mice infected with WA1/2020 N501Y/D614G, B.1.1.7, Wash-B.1.351 or Wash-B.1.1.28, the 2B04/47D11 treatment conferred protection against B.1.1.7 and WA1/2020 N501Y/D614G but not against Wash-B.1.351 or Wash-B.1.1.28 viruses at this lower dose (Extended Data Fig. 7e–h). By comparison, in nasal washes, all four mAb cocktails resulted in relatively similar reductions in viral RNA levels at six days after infection of mice inoculated with WA1/2020 N501Y/D614G, B.1.1.7, Wash-B.1.351 or Wash-B.1.1.28 (Extended Data Fig. 7i–l). Even at this low treatment dose, with the exception of some breakthrough events (>6 log10(copies of N per mg): COV2-2130/COV2-2196 (2 of 24 mice); S309/S2E12 (6 of 24 mice); REGN10933/REGN10987 (1 of 24 mice); and 2B04/47D11 (6 of 24 mice)), the mAb combinations generally prevented infection of the brain (Extended Data Fig. 7m–p, Supplementary Table 3). Overall, the neutralization activity of mAbs in vitro against SARS-CoV-2 variants correlated with lung viral burden after prophylactic administration (Extended Data Fig. 8).

As an alternative model for evaluating mAb efficacy, we tested immunocompetent, inbred 129S2 mice, which are permissive to infection by SARS-CoV-2 strains that encode an N501Y substitution without human ACE2 expression12,13; presumably, the N501Y adaptive mutation enables engagement of mouse ACE2. We administered a single 40-μg (about 2 mg kg−1) dose of mAb cocktails (COV2-2130/COV2-2196, S309/S2E12 or REGN10933/REGN10987) or a control mAb via intraperitoneal injection one day before intranasal inoculation with 103 focus-forming units (FFU) of WA1/2020 N501Y/D614G, Wash-B.1.351 or Wash-B.1.1.28, and 105 FFU of B.1.1.7 (Extended Data Fig. 9). A higher inoculating dose of B.1.1.7 was required to obtain equivalent levels of viral RNA in the lung compared to the other three viruses. At three days after infection, we collected tissues for viral burden analyses; at this time point, reproducible weight loss was not observed. All three mAb cocktails tested (COV2-2130/COV2-2196, S309/S2E12 and REGN10933/REGN10987) protected 129S2 mice against infection in the lung by all of the SARS-CoV-2 strains (Extended Data Fig. 9a–d); despite some variability, we observed a trend towards less complete protection in mice infected with Wash-B.1.351 and Wash-B.1.1.28 strains (Extended Data Fig. 3c–f). When we evaluated the nasal washes, reductions in viral RNA levels were diminished with the Wash-B.1.351 virus, especially for the COV2-2130/COV2-2196 and REGN10933/REGN10987 combinations (Extended Data Fig. 9e–h). Sequencing analysis of lung samples from the infected 129S2 mice also did not reveal evidence of acquisition of mutations in the RBD (Supplementary Table 3).

The immunocompetent Syrian golden hamster has previously been used to evaluate mAb activity against SARS-CoV-2 infection14,15. We used this model to assess independently the inhibitory activity and possible emergence of resistance of one of the mAb combinations (COV2-2130/COV2-2196) against viruses containing the B.1.351 spike protein. One day before intranasal inoculation with 5 × 105 FFU of Wash-B.1.351 or WA1/2020 D614G, we treated hamsters with a single 800-μg (about 10 mg kg−1) or 320-μg (about 4 mg kg−1) dose of the COV2-2130/COV2-2196 cocktail or isotype control mAb by intraperitoneal injection (Extended Data Fig. 10). Weights were followed for four days, and tissues were collected for virological and cytokine analysis. At the 800-μg mAb cocktail dose, hamsters treated with COV2-2130/COV2-2196 and infected with WA1/2020 D614G or Wash-B.1.351 showed protection against weight loss and reduced viral burden levels in the lungs, but not nasal swabs, compared to the isotype control mAb (Extended Data Fig. 10a–d). Correspondingly, quantitative PCR with reverse transcription analysis of a previously described set of cytokines and inflammatory genes11 showed reduced mRNA expression in the lungs of hamsters treated with COV2-2130/COV2-2196 (Extended Data Fig. 10e–h). Consensus sequencing of the RBD region of viral RNA samples from the lungs of hamsters treated with COV2-2130/COV2-2196 did not show evidence of mutation or escape (Supplementary Table 3). When the lower 320-μg dose of COV2-2130/COV2-2196 was administered, we observed a trend towards protection against weight loss in hamsters infected with WA1/2020 D614G and Wash-B.1.351. Consistent with a partially protective phenotype, hamsters treated with the lower 320-μg dose of COV2-2130/COV2-2196 and inoculated with either WA1/2020 D614G or Wash-B.1.351 showed a trend towards reduced viral RNA in the lungs at 4 days after infection and markedly diminished (about 104- to 105-fold) levels of infectious virus as determined by plaque assay (Extended Data Fig. 10j, k). The reduction in lung viral load conferred by the lower dose of COV2-2130/COV2-2196 corresponded with diminished inflammatory gene expression after infection with either WA1/2020 D614G or Wash-B.1.351 (Extended Data Fig. 10m–p). In contrast to the protection seen in the lung, differences in viral RNA were not observed in nasal washes between COV2-2130/COV2-2196 and isotype control mAb-treated hamsters, regardless of the infecting strain (Extended Data Fig. 10l). Sequencing of the RBD of viral RNA from the lungs of COV2-2130/COV2-2196 or isotype mAb-treated hamsters also did not detect evidence of escape mutation selection after infection (Supplementary Table 3). Overall, these studies in hamsters with near-threshold dosing of the COV2-2130/COV2-2196 mAb cocktail establish protection and an absence of rapid escape against SARS-CoV-2 containing spike proteins from historical or variant strains.

Therapeutic efficacy against variants

As mAbs are being developed clinically as therapeutic agents, we assessed their post-exposure efficacy against the SARS-CoV-2 strain expressing the B.1.351 spike protein using K18-hACE2 mice. We administered a single, higher (200-μg; about 10 mg kg−1) dose of COV2-2130/COV2-2196, S309/S2E12, REGN10933/REGN10987 or 2B04/47D11 by intraperitoneal injection one day after inoculation with WA1/2020 N501Y/D614G (Fig. 3a–c) or Wash-B.1.351 (Fig. 3d–f). Compared to the control mAb-treated mice, which lost at least 15% of their starting weight over the 6 days of the experiment, each of these mAb cocktails prevented weight loss induced by WA1/2020 N501Y/D614G or Wash-B.1.351 infection (Fig. 3a, d). Histopathological analysis of lung sections from control mAb-treated mice showed interstitial oedema, immune cell infiltration and collapsed alveolar spaces, consistent with the inflammation induced by SARS-CoV-2 infection (Extended Data Fig. 11). By contrast, COV2-2130/COV2-2196-, S309/S2E12-, REGN10933/REGN10987- and 2B04/47D11-treated mice showed reduced or minimal lung pathology. COV2-2130/COV2-2196, S309/S2E12 and REGN10933/REGN10987 mAb cocktail treatments resulted in reduced infectious virus and viral RNA levels in lung homogenates, and viral RNA levels in nasal washes and brain homogenates from mice infected with either WA1/2020 N501Y/D614G or Wash-B.1.351 (Fig. 3b, c, e, f, Extended Data Figs. 3g, h, 12). By comparison, although the 2B04/47D11 mAb cocktail reduced viral RNA levels in the lungs, it showed less protection in the nasal washes of WA1/2020-N501Y/D614G- or Wash-B.1.351-infected mice. Although neutralizing capacity correlated with lung viral burden when mAbs were administered as prophylaxis, this association was less direct when mAbs were given in the post-exposure setting (Extended Data Fig. 8). This result was not entirely unexpected, as effector functions of some mAbs are required for optimal activity when given as post-exposure therapy11,16. Indeed, recent studies with influenza and SARS-CoV-2 neutralizing antibodies suggest that Fc engagement mitigates inflammation, improves respiratory mechanics and promotes viral clearance in the lung11,17.

Fig. 3: Post-exposure antibody therapy against SARS-CoV-2 variants in K18-hACE2 mice.
figure 3

af, Eight-to-ten-week-old female and male K18-hACE2 transgenic mice were administered 103 FFU of the SARS-CoV-2 N501Y/D614G (ac) or Wash-B.1.351 (df) strains by intranasal inoculation. One day later, mice received 200 μg (about 10 mg kg−1) of the indicated mAb treatment by intraperitoneal injection. Tissues were collected at six days after infection. a, d, Weight change after infection with SARS-CoV-2 (mean ± s.e.m.; in order from left to right n = 15, 6, 6, 6, 6, 7 and 7 (a, d) mice per group, two experiments; one-way ANOVA with Dunnett’s test of area under the curve. NS, not significant, ****P < 0.0001). Viral RNA levels in the lung (b, e) and nasal wash (c, f) (line indicates median; in order from left to right n = 7, 6, 6, 6, 6, 7 and 7 (b); n = 9, 6, 6, 6, 6, 7 and 7 (c, e, f) mice per group, 2 experiments; one-way ANOVA with Dunnett’s test with comparison to control mAb. NS, not significant, ****P < 0.0001; **P = 0.0014 (left) and 0.0088 (right), ***P = 0.0007 (left) and 0.0003 (right) (b); **P = 0.0026 (left), 0.0041 (middle) and 0.0049 (right) (c); **P = 0.0049, ***P = 0.0004 (e); **P = 0.0094, ***P = 0.0005, *P = 0.0442 (f)). Dotted line indicates the limit of detection of the assay.

Source data

With the emergence of several SARS-CoV-2 variants, it remains uncertain whether vaccines and antibody-based therapies will lose efficacy18. Cell-culture-based studies have shown that several of the mutations in variant strains (especially those at positions 452 and 484) result in reduced neutralization by antibodies derived from infected or vaccinated individuals1,2,3,19,20. Here we evaluated antibodies forming the basis of five mAb therapies in clinical development for in vivo efficacy against infection by SARS-CoV-2 variants. Monotherapy with LY-CoV555, an antibody that corresponds to bamlanivimab21, showed complete neutralization escape in cell culture and did not confer protection against viruses containing E484 substitutions. By contrast, all cocktails of two neutralizing mAbs conferred protection to varying degrees, even if one of the constituent mAbs showed reduced activity owing to resistance. Moreover, the higher doses of mAbs used in patients (for example, about 35 mg kg−1 for casirivimab and imdevimab (corresponding to the REGN mAbs)) could compensate for loss in neutralization potency. Generally, in mice and hamsters, mAb-mediated protection was better in the lung than in nasal washes, possibly because IgG levels in the respiratory mucosa are lower than in plasma22,23.

In our study, combination therapy with several mAbs was protective in mice and hamsters against variant strains, which highlights the importance of using multiple mAbs that recognize distinct epitopes (rather than monotherapy) to control SARS-CoV-2 infection. Indeed, the Emergency Use Authorization for bamlanivimab (LY-CoV555) as monotherapy was revoked, as the antibody does not reduce SARS-CoV-2 infection of several variants of concern2,24,25; instead, a combination of bamlanivimab (LY-CoV555) and etesevimab (LY-CoV016) is recommended, even though strains that contain E484 and K417 mutations (such as B.1.351, B.1.1.28, B.1.617.1 and B.1.526) will probably have resistance to both mAb components2,26 (Extended Data Fig. 13). Indeed, in K18-hACE2 mice, we found that LY-CoV555/LY-CoV016 had therapeutic activity against WA1/2020 N501Y/D614G (Fig. 3a–c) but not against Wash-B.1.351 (Fig. 3d–f). This failure of protection occurred because both of the mAbs in the cocktail lost neutralizing activity. Other cocktails may fare better against variants of concern as long as one of the mAbs in each pair retains substantial inhibitory activity. Beyond a loss of potency against already circulating resistant variants, antibody monotherapy can be compromised within an individual by rapid selection of escape mutations. Consistent with this idea, in other animal experiments with SARS-CoV-2, we observed the emergence of resistance against antibody monotherapy, resulting in the accumulation of mutations at RBD residues 476, 477, 484, or 487 (ref. 27, and M.S.D., unpublished data). Despite amplifying sequences from 99 brain, nasal wash and lung samples from mice and hamsters treated with the different mAb combinations in this study, we did not detect a single escape mutant. Combination mAb treatment may prevent escape through synergistic interactions in vivo or by driving selection of mutants with compromised fitness.

At the lower doses of the mAbs we tested, we observed some differences in mAb cocktail efficacy between rodent models, which could be due to host variation, viral variation or small differences in antibody levels. For example, mutations in the RBD can affect mAb binding as well as ACE2 binding28. Mutation at position 501 of the spike is of particular interest, as it enables mouse adaptation12,13 and is present in several variants of concern. The N501Y change associated with infection of conventional laboratory mice could facilitate virus engagement with mouse ACE2 or possibly other putative receptors. Beyond this, polymorphisms in or differences of expression of host receptors on key target cells also could affect SARS-CoV-2 infection in different hosts and the inhibitory effects of neutralizing antibodies. This complexity of antibody–spike protein–receptor interactions probably explains some of the variation in protection between K18-hACE2 mice, 129S2 mice and hamsters. Alternatively, the pharmacokinetics of human antibodies could vary between animals and affect efficacy. We observed small differences in serum antibody levels in the context of viral challenge that could affect relative protection (Supplementary Table 4).

Previous studies with pseudoviruses and authentic SARS-CoV-2 containing variant substitutions have suggested that adjustments to therapeutic antibody regimens might be necessary1,2,29,30,31. Although our studies with variant strains in vivo suggest that this conclusion probably holds for mAb monotherapy, four different mAb combinations performed well even when a virus was fully resistant to one mAb component. Thus, our results suggest that—as previously described for the historical WA1/2020 strain32—many, but not all, combination therapies with neutralizing mAbs should retain efficacy against emerging SARS-CoV-2 variants.

Methods

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

Cells

Vero-TMPRSS234 and Vero cells expressing human ACE2 and TMPRRS2 (Vero-hACE2-TMPRRS2)1 cells were cultured at 37 °C in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 10 mM HEPES pH 7.3, 1 mM sodium pyruvate, 1× non-essential amino acids 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 strains with substitutions (D614G or N501Y/D614G) were obtained from an infectious cDNA clone of the 2019n-CoV/USA_WA1/2020 strain, as previously described35. The B.1.351- and B.1.1.28-variant spike genes were introduced into the WA1/2020 backbone as previously described to create chimeric SARS-CoV-21. The B.1.1.7, B.1.429, B.1.351 and B.1.526 (S477N or E484K variant) isolates were obtained from nasopharyngeal isolates. The B.1.617.1 variant was plaque-purified from a midturbinate nasal swab and passaged twice on Vero-TMPRSS2 cells, as previously described36. All viruses were passaged once in Vero-TMPRSS2 cells and subjected to next-generation sequencing as previously described1 to confirm the introduction and stability of substitutions (Supplementary Table 1). Substitutions for each variant were as follows: B.1.1.7: deletion of 69 and 70 and 144 and 145, N501Y, A570D, D614G, P681H, T716I, S982A, and D1118H; B.1.351: D80A, T95I, D215G, deletion of 242-244, K417N, E484K, N501Y, D614G, and A701V; B.1.1.28: L18F, T20N, P26S, D138Y, R190S, K417T, E484K, N501Y, D614G, H655Y, and T1027I; B.1.429: S13I, W152C, L452R, and D614G; B.1.617.1: G142D, E154K, L452R, E484Q, D614G, P681R, Q1071H, and H1101D; and B.1.526 (S477N or E484K variant): L5F, T95I, D253G, S477N, E484K, D614G, and A701V or Q957R. All virus experiments were performed in an approved biosafety level 3 facility.

mAb purification

The mAbs studied in this paper (COV2-2196, COV2-2130, S309, S2E12, 2B04, 47D11, REGN10933, REGN10987, LY-CoV555 and LY-CoV016) have previously been described32,37,38,39,40,41,42,43. COV2-2196 and COV2-2130 mAbs were produced after transient transfection using the Gibco ExpiCHO Expression System (ThermoFisher Scientific) following the manufacturer’s protocol. Culture supernatants were purified using HiTrap MabSelect SuRe columns (Cytiva, (formerly GE Healthcare Life Sciences)) on an AKTA 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 EU per mg IgG). Endotoxin testing was performed using the PTS201F cartridge (Charles River), with a sensitivity range from 10 to 0.1 EU per ml, and an Endosafe Nexgen-MCS instrument (Charles River). S309, S2E12, REGN10933, REGN10987, LY-CoV016 and LY-CoV555 mAb proteins were produced in CHOEXPI 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. The final products were sterilized by filtration through 0.22-μm filters and stored at 4 °C.

Mouse experiments

Animal studies were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Institutional Animal Care and Use Committee at the Washington University School of Medicine (assurance number A3381–01). Virus inoculations were performed under anaesthesia that was induced and maintained with ketamine hydrochloride and xylazine, and all efforts were made to minimize animal suffering.

Heterozygous K18-hACE2 C57BL/6J mice (strain: 2B6.Cg-Tg(K18-ACE2)2Prlmn/J) and 129 mice (strain: 129S2/SvPasCrl) were obtained from The Jackson Laboratory and Charles River Laboratories, respectively. Mice were housed in groups and fed standard chow diets. Six-to-ten-week-old mice of both sexes were administered 103 or 105 FFU of the respective SARS-CoV-2 strain by intranasal administration. In vivo studies were not blinded, and mice were randomly assigned to treatment groups. No sample-size calculations were performed to power each study. Instead, sample sizes were determined on the basis of previous in vivo virus challenge experiments.

For antibody prophylaxis and therapeutic experiments, mice were administered the indicated mAb dose by intraperitoneal injection one day before or after intranasal inoculation with the indicated SARS-CoV-2 strain.

Hamster experiments

Six-month-old male Syrian hamsters were purchased from Charles River Laboratories and housed in microisolator units. All hamsters were allowed free access to food and water and cared for under United States Department of Agriculture (USDA) guidelines for laboratory animals. Hamsters were administered by intraperitoneal injection mAbs COV2-2130 and COV2-2196 or isotype control (4 or 10 mg kg−1, depending on the experiment). One day later, hamsters were given 5 × 105 FFU of indicated SARS-CoV-2 strain by the intranasal route in a final volume of 100 μl. All hamsters were monitored for body weight loss until being humanely euthanized at four days after infection. Nasal swabs were collected at three days after infection. All procedures were approved by the Washington University School of Medicine (assurance number A3381–01). Virus inoculations and antibody transfers were performed under anaesthesia that was induced and maintained with 5% isoflurane. All efforts were made to minimize animal suffering.

Focus reduction neutralization test

Serial dilutions of mAbs (starting at 10 μg ml−1 dilution) were incubated with 102 FFU of different strains or variants of SARS-CoV-2 for 1 h at 37 °C. Antibody–virus complexes were added to Vero-TMPRSS2 cell monolayers in 96-well plates and incubated at 37 °C for 1 h. Subsequently, cells were overlaid with 1% (w/v) methylcellulose in MEM. Plates were collected 30 h later by removing overlays and fixed with 4% PFA in PBS for 20 min at room temperature. Plates were washed and sequentially incubated with an oligoclonal pool of SARS2-2, SARS2-11, SARS2-16, SARS2-31, SARS2-38, SARS2-57 and SARS2-7144 anti-S antibodies and HRP-conjugated goat anti-mouse IgG (Sigma, 12-349) in PBS supplemented with 0.1% saponin and 0.1% bovine serum albumin. SARS-CoV-2-infected cell foci were visualized using TrueBlue peroxidase substrate (KPL) and quantitated on an ImmunoSpot microanalyzer (Cellular Technologies).

Measurement of viral burden

Tissues were weighed and homogenized with zirconia beads in a MagNA Lyser instrument (Roche Life Science) in 1000 μl of DMEM medium supplemented with 2% heat-inactivated FBS. Tissue homogenates were clarified by centrifugation at 10,000 rpm for 5 min and stored at −80 °C. RNA was extracted using the MagMax mirVana Total RNA isolation kit (Thermo Fisher Scientific) on the Kingfisher Flex extraction robot (Thermo Fisher Scientific). RNA was reverse-transcribed and amplified using the TaqMan RNA-to-CT 1-Step Kit (Thermo Fisher Scientific). Reverse transcription was carried out at 48 °C for 15 min followed by 2 min at 95 °C. Amplification was accomplished over 50 cycles as follows: 95 °C for 15 s and 60 °C for 1 min. Copies of SARS-CoV-2 N gene RNA in samples were determined using a previously published assay45. In brief, a TaqMan assay was designed to target a highly conserved region of the N gene (forward primer: ATGCTGCAATCGTGCTACAA; reverse primer: GACTGCCGCCTCTGCTC; probe: /56-FAM/TCAAGGAAC/ZEN/AACATTGCCAA/3IABkFQ/). This region was included in an RNA standard to allow for copy number determination down to 10 copies per reaction. The reaction mixture contained final concentrations of primers and probe of 500 and 100 nM, respectively.

Plaque assay

Vero-hACE2-TMPRSS2 cells were seeded at a density of 1 × 105 cells per well in 24-well tissue culture plates. The following day, medium was removed and replaced with 200 μl of material to be titrated diluted serially in DMEM supplemented with 2% FBS. One hour later, 1 ml of methylcellulose overlay was added. Plates were incubated for 72 h, then fixed with 4% paraformaldehyde (final concentration) in PBS for 20 min. Plates were stained with 0.05% (w/v) crystal violet in 20% methanol and washed twice with distilled, deionized water.

Cytokine and chemokine protein measurements

Lung homogenates were incubated with Triton-X-100 (1% final concentration) for 1 h at room temperature to inactivate SARS-CoV-2. Homogenates then were analysed for cytokines and chemokines by Eve Technologies using their Mouse Cytokine Array/Chemokine Array 31-Plex (MD31) platform.

Lung histology

Animals were euthanized before harvest and fixation of tissues. Lungs were inflated with about 2 ml of 10% neutral buffered formalin using a 3-ml syringe and catheter inserted into the trachea and kept in fixative for 7 days. Tissues were embedded in paraffin, and sections were stained with haematoxylin and eosin. Images were captured using the Nanozoomer (Hamamatsu) at the Alafi Neuroimaging Core at Washington University.

Statistical analysis

All statistical tests were performed as described in the indicated figure legends using Prism 8.0. Statistical significance was determined using an ordinary one-way ANOVA with Dunnett’s post-test when comparing three or more groups. The number of independent experiments used are indicated in the relevant figure legends.

Reporting summary

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