SARS-CoV-2 variants show resistance to neutralization by many monoclonal and serum-derived polyclonal antibodies

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused the global COVID-19 pandemic infecting more than 106 million people and causing 2.3 million deaths. The rapid deployment of antibody-based countermeasures has provided hope for curtailing disease and ending the pandemic1. However, the emergence of rapidly-spreading SARS-CoV-2 variants in the United Kingdom (B.1.1.7), South Africa (B.1.351), and elsewhere with mutations in the spike protein has raised concern for escape from neutralizing antibody responses and loss of vaccine efficacy based on preliminary data with pseudoviruses2–4. Here, using monoclonal antibodies (mAbs), animal immune sera, human convalescent sera, and human sera from recipients of the Pfizer-BioNTech (BNT162b2) mRNA vaccine, we report the impact on antibody neutralization of a panel of authentic SARS-CoV-2 variants including a B.1.1.7 isolate, a chimeric Washington strain with a South African spike gene (Wash SA-B.1.351), and isogenic recombinant variants with designed mutations or deletions at positions 69–70, 417, 484, 501, and/or 614 of the spike protein. Several highly neutralizing mAbs engaging the receptor binding domain (RBD) or N-terminal domain (NTD) lost inhibitory activity against Wash SA-B.1.351 or recombinant variants with an E484K spike mutation. Most convalescent sera and virtually all mRNA vaccine-induced immune sera tested showed markedly diminished neutralizing activity against the Wash SA-B.1.351 strain or recombinant viruses containing mutations at position 484 and 501. We also noted that cell line selection used for growth of virus stocks or neutralization assays can impact the potency of antibodies against different SARS-CoV-2 variants, which has implications for assay standardization and congruence of results across laboratories. As several antibodies binding specific regions of the RBD and NTD show loss-of-neutralization potency in vitro against emerging variants, updated mAb cocktails, targeting of highly conserved regions, enhancement of mAb potency, or adjustments to the spike sequences of vaccines may be needed to prevent loss of protection in vivo.


Full Text
To evaluate the effects of SARS-CoV-2 strain variation on antibody neutralization, we obtained or generated a panel of authentic infectious SARS-CoV-2 strains with sequence variations in the spike gene (Fig 1a-c). A B.1.1.7 isolate had signature changes in the spike gene 5  were propagated in Vero-TMPRSS2 and Vero-hACE2-TMPRSS2 cells expressing transmembrane protease serine 2 (TMPRSS2) and human angiotensin converting enzyme 2 (hACE2) to prevent the development of adventitious mutations in the spike, especially at or near the furin cleavage site, which accumulate rapidly in Vero E6 cells 6 and can impact entry pathways and virulence 7 . All viruses were used at low passage (p0 or p1) and deep sequenced to con rm mutations (Supplementary Table S1).
We performed neutralization tests with the different spike protein variants and the two cell types (Fig 1d-i and Extended Data Fig 2). With the parental WA1/2020 strain, which was derived in Vero CCL-81 cell cultures, neutralization by the majority of class 1 mAbs was similar in Vero-hACE2-TMPRSS2 or Vero-TMPRSS2 cells. In comparison, some class 2 mAbs (e.g., S309 and SARS2-44) showed a 4 to 6-fold loss in neutralization potency (EC 50 value) on Vero-hACE2-TMPRSS2 compared to Vero-TMPRSS2 cells. Moreover, NTD-reactive mAbs (COV2-2489 and COV2-2676) neutralized Vero CCL-81 cell-derived WA1/2020 virus on Vero-hACE2-TMPRSS2 cells but lost activity on Vero-TMPRSS2 cells or when viruses were derived from Vero-hACE2-TMPRSS2 cells (Fig  1f, h-i and Extended Data Fig 2). Given that the expression of hACE2 on recipient Vero cells and the cellular source of virus both can impact the neutralizing activity of mAbs that bind principally outside of the RBM on the spike protein, virus neutralization assays being used for correlation with in vivo e cacy of mAbs and vaccines may produce variable results depending on the cell substrate used for virus propagation and infection.
We next assessed the impact of spike protein mutations on mAb neutralization using Vero-hACE2-TMPRSS2 cells (Fig 1h) and Vero-TMPRSS2 cells (Fig 1i). We observed the following patterns with the variant viruses: (a) The D614G or P681H mutations (in the C-terminal region of S1) and the 69-70 deletion (in the NTD) had marginal effects on neutralization potency for the RBM and RBD mAbs we evaluated. It was di cult to assess their impact on the NTD mAbs we tested, since the recombinant viruses were generated in Vero-hACE2-TMPRSS2 cells, and the NTD mAbs neutralized them poorly at baseline; (b) The K417N mutation resulted in ~10-fold reductions in neutralization by mAbs COVOX-40 and SARS2-44 but did not negatively affect other mAbs in our panel. If anything, several class 1 mAbs showed slightly improved inhibitory activity (P = 0.002, Wilcoxon matched-pairs signed rank test) with this mutation; (c) Mutation at N501Y reduced the neutralizing activity of COVOX-40 and SARS2-44 slightly but did not alter the potency of other mAbs substantively; this result is consistent with data showing that human convalescent sera e ciently neutralize viruses with N501Y substitutions [16][17][18] ; (d) The E484K mutation negatively impacted the potency of several class 1 antibodies. Compared to the D614G virus, mAbs COV2-2196, COV2-3025, COV2-2381 and S2E12 showed 4-to 5-fold reduced activity against the E484K/D614G virus, and COV2-2050, COVOX-384, 1B07, and S2H58 lost virtually all neutralizing potential; (e) The combination of E484K/N501Y/D614G mutations, which is present in the circulating South African B.1.351 and Brazilian B.1.1.248 strains, showed even greater effects (6-to 13-fold reductions) on the activity of class 1 mAbs COV2-2196, COV2-3025, COV2-2381, and S2E12 mAbs; (f) When we tested class 1 mAbs for inhibition of the Wash SA-B.1.351 virus containing the full South African spike sequence, as expected, several mAbs (COV2-2050, COVOX-384, 1B07, and S2H58) lost activity in both Vero-hACE2-TMPRSS2 and Vero-TMPRSS2 cells. However, the reductions in neutralizing potential by other class 1 mAbs (COV2-2196, COV2-3025, COV2-2381, and S2E12) seen against the E484K/N501Y/D614G mutant virus were absent with Wash SA-B.1.351, which contains additional mutations. The K417N substitution, which is located at the edge of the RBM (Fig 1b) and enhances neutralization by some class 1 mAbs, may compensate for the negative effects on inhibition of the E484K/N501Y mutations. In comparison, we observed a distinct neutralization pattern with Wash SA-B.1.351 for class 2 and 3 mAbs. Because some of these mAbs neutralized Vero-hACE2-TMPRSS2 cell-derived virus poorly when tested in Vero-hACE2-TMPRSS2 cells, we performed parallel experiments in Vero-TMPRSS2 cells. Class 2 mAbs binding the base of the RBD showed small reductions in potency against the Wash SA-B.1.351. However, the two NTD mAbs in class 3 (COV2-2676 and COV2-2489) showed a loss of neutralizing activity against Wash SA-B.1.351 in Vero-TMPRSS2 cells, consistent with recent data with other NTD mAbs using pseudoviruses 4 ; (g) With one exception, none of the class 1 mAbs lost activity against the B.1.1.7 isolate on Vero-TMPRSS2 or Vero-hACE2-TMPRSS2 cells. However, we observed moderately diminished neutralizing activity (3-to 11-fold reduction) of some class 2 mAbs (SARS2-31, SARS2-44, and S309) against the B.1.1.7 strain depending on the cell substrate. The reduced potency of S309 mAb against B.1.1.7 strain contrasts with data showing it binds avidly to the B.1.1.7 spike protein on the surface of cells and neutralizes a vesicular stomatitis virus (VSV) pseudotyped with B.1.1.7 spike protein in Vero E6 cells (Extended Data Fig 3a-b). Moreover, one of the NTD class 3 mAbs (COV2-2489) also showed a marked loss of inhibitory activity against the B.1.1.7 strain in both cell types, possibly due to the deletions present in the NTD (69-70 and 144-145) 15 . Together, these data indicate that cell line selection (for both growth of virus stocks and neutralization assays) and hACE2 receptor expression are important variables in assessing the potency of antibodies against different SARS-CoV-2 variants.
We also evaluated the impact of cell substrate and hACE2 receptor expression on neutralizing activity of serum samples from convalescent adults enrolled at approximately one month after infection (Fig 5a-c) and from BNT162b2 mRNA-vaccinated individuals (Fig 5d-f). Given the limited remaining serum quantities, we performed neutralization experiments only with WA1/2020, B.1.1.7, and Wash SA-B.1.351 viruses using Vero-TMPRSS2 cells. Using nextgeneration sequence analysis of viral genomes, we con rmed additional mutations did not occur during passage in Vero-TMPRSS2 cells (Supplementary  Table S1). These experiments (Extended Data Fig 8) revealed the following: (a) Convalescent and vaccine sera showed reductions in neutralizing activity of B.1.1.7 compared to the WA1/2020 virus. Whereas sera from the vaccinated individuals inhibited B.1.1.7 virus infection less e ciently (3.6-fold, P < 0.01; Fig  5d), convalescent sera showed a trend towards reduced neutralization that did not attain statistical signi cance (2.4-fold, P = 0.08; Fig 5a). (b) In comparison, sera from both convalescent and vaccinated individuals showed a marked 10-to 13-fold reduction (P < 0.01) in neutralizing potency against the Wash SA-B.1.351 virus (Fig 5b, e). The results for the vaccine sera were similar in magnitude between Vero-hACE2-TMPRSS2 and Vero-TMPRSS2 cells (see also Fig 4a, d) and suggest that cellular expression of hACE2 does not markedly impact functional outcome. However, we generally observed greater decreases in neutralizing activity against the B.1.1.7 strain with convalescent sera in Vero-TMPRSS2 cells than in Vero-hACE2-TMPRSS2 cells (see Fig 2a), possibly because NTD antibodies are produced at higher levels during natural infection than vaccination and lose binding to B.1.1.7 viruses because of deletions in the NTD 25 .
Our in vitro experiments using a B.1.1.7 isolate and engineered variants in the backbone of the WA1/2020 strain establish that mutations in the spike can impact the potency of antibody neutralization. Some neutralizing mAbs targeting the base of the RBD or NTD showed reduced activity against the B.1.1.7 isolate, whereas others targeting the RBM or NTD failed to inhibit infection of Wash SA-B.1.351 or variants containing the E484K mutation. These nding are potentially important because the RBM has functional plasticity 26,27 , and additional mutations in this region that occur as the pandemic evolves could further impact the e cacy of mAb therapies or vaccines. Our results establishing the E484K substitution as a vulnerability for multiple neutralizing mAbs are consistent with deep mutational scanning or VSV-SARS-CoV-2-based neutralization escape screening campaigns 26,28,29 . However, several other highly neutralizing mAbs (e.g., COV2-2196, COV2-2381, COV2-3025, and S2E12) showed intact or only slightly diminished inhibitory activity against the suite of variant viruses we tested. Moreover, cocktails of mAbs binding different epitopes in spike protein overcame virus resistance to individual mAbs. Alternative approaches to addressing the diminished mAb neutralization activity by variant SARS-CoV-2 lineages include targeting of conserved regions of the spike and identifying clonal mAb variants with greater potency, such that a given dose of mAb can protect against a range of variants despite some decrease in neutralization activity.
Our studies with human sera from convalescent subjects and recipients of the BNT162b2 mRNA vaccine, and animal sera after immunization with a vaccine encoding a similar spike gene, demonstrate a lower potency of neutralization against E484K and N501Y-containing viruses (note: we did not perform studies with the single-mutation viruses, due to limited serum availability). This observation is unexpected given that antibody responses in animals and humans are polyclonal and in theory, should overcome resistance associated with individual mutations and loss of activity of particular B cell clones. Our studies focused exclusively on the impact of sequence changes in the spike protein on antibody neutralization in cell culture. Despite observing marked differences in serum neutralizing activity against authentic SARS-CoV-2 variant viruses, it remains unclear how this nding translates into effects on protection in the context of secondary infection or infection after vaccination with platforms using historical spike gene sequences. Although serum neutralizing titers are an anticipated correlate of protection 32 , this measurement does not account for Fc effector functions; Fcg receptor or complement protein engagement by non-, weakly-, or strongly-neutralizing antibodies that bind the SARS-CoV-2 spike protein on the surface of infected cells could confer substantial protection [33][34][35] . Also, the role of memory T or B cells in protection against variant viruses is unknown and could prevent severe infection even in the setting of compromised serum antibody responses [36][37][38] .
Moreover, the eld still does not know whether Vero or other cell-based neutralization assays predict antibody-mediated protection. Indeed, primary cells targeted by SARS-CoV-2 in vivo can express unique sets of attachment and entry factors 39 , which could impact receptor and entry blockade by speci c antibodies. We observed that cell line used for growth of viral stocks or neutralization assays affects the potency of monoclonal or serum-derived antibodies against different SARS-CoV-2 variants. Such results may impact the congruity of data across laboratories and interpretation of effects of viral variants on vaccine e cacy. As an example, a recent study with Vero E6 cell-derived SARS-CoV-2 with a spike protein containing some of the South African mutations (E484K, N501Y, and D614G) showed only a small 1.2-fold decrease in neutralization potency by BNT162b2 mRNA vaccine-elicited human sera 16 . When we compared neutralization of deep-sequenced con rmed p0 (Vero E6 cell-produced) and p1 (Vero-hACE2-TMPRSS2 cell-produced) K417N/E484K/N501Y/D614G viruses by immune serum from vaccinated mice, hamsters, or NHPs, or naturally infected humans in recipient Vero-hACE2-TMPRSS2 cells, the p0 viruses produced in Vero E6 cells were neutralized more e ciently (~3-fold, P < 0.05) than the p1 viruses produced in Vero-hACE2-TMPRSS2 cells (Extended Data Fig 9). We speculate that TMPRSS2 might modify the spike protein of authentic SARS-CoV-2 in the producer cell such that
Monoclonal antibodies. The human mAbs studied in this paper (COV2-2196, COV2-2072, COV2-2050, COV2-2381, COV2-2130, COVOX-384, COVOX-40,   S309, S2E12, S2H58, S2X333, VIR-7381, and S2X259) were isolated from blood samples from individuals in North America or Europe with previous laboratory-con rmed symptomatic SARS-CoV or SARS-CoV-2 infection. The original clinical studies to obtain specimens after written informed consent were previously described 10,13,14,25 and approved by the Institutional Review Board of Vanderbilt University Medical Center, the Institutional Review Board of the University of Washington, the Research Ethics Board of the University of Toronto, and the Canton Ticino Ethics Committee (Switzerland). Chimeric mAb 1B07 with a murine Fv and human Fc (human IgG1) were isolated from C57BL/6 mice immunized with recombinant spike and RBD proteins and described previously 12 . Murine mAbs were generated in BALB/c or C57BL/6 mice immunized with recombinant spike and RBD proteins and described previously 28,30 .
Human immune sera. Multiple sources of human serum samples were used in this study: Convalescent serum samples were obtained from a cohort to increase surface expression of the recombinant spike. Next, three contiguous overlapping fragments were fused by a rst overlap PCR (step 2) using the utmost external primers of each set, resulting in three larger fragments with overlapping sequences. A nal overlap PCR (step 3) was performed on the three large fragments using the utmost external primers to amplify the S gene and the anking sequences including the restriction sites KpnI and NotI. This fragment was digested and cloned into the expression plasmid phCMV1. For all PCR reactions, the Q5 Hot Start High delity DNA polymerase was used (New England Biolabs), according to the manufacturer's instructions and adapting the elongation time to the size of the amplicon. After each PCR step, the ampli ed regions were separated on agarose gel and puri ed using Illustra GFX™ PCR DNA and Gel Band Puri cation Kit (Merck KGaA).
Expi-CHO cells were transiently transfected with SARS-CoV-2-S expression vectors using Expifectamine CHO Enhancer. Two days later, cells were collected for immunostaining with mAbs. An Alexa647-labelled secondary antibody anti-human IgG Fc was used for detection. Binding of mAbs to transfected cells was analyzed by ow-cytometry using a ZE5 Cell Analyzer (Biorard) and FlowJo software (TreeStar). Positive binding was de ned by differential staining of CoV-S-transfectants versus mock-transfectants.
SARS-CoV-2 pseudotyped virus production. 293T/17 cells were seeded in 10-cm dishes for 80% next day con uency. The next day, cells were transfected with the plasmid pcDNA3.1(+)-spike-D19 (encoding the SARS-CoV-2 spike protein) or pcDNA3.1(+)-spike-D19 variants using the transfection reagent TransIT-Lenti according to the manufacturer's instructions. One day post-transfection, cells were infected with VSV-luc(VSV-G) at an MOI of 3. The cell supernatant containing SARS-CoV-2 pseudotyped virus was collected at day 2 post-transfection, centrifuged at 1,000 x g for 5 min to remove cellular debris, aliquoted and frozen at -80°C. The SARS-CoV-2 pseudotyped virus preparation was quanti ed using Vero E6 cells seeded at 20,000 cells/well in clear bottom black 96 well plates the previous day. Cells were inoculated with 1:10 dilution series of pseudotyped virus in 50 μL DMEM for 1 h at 37°C. An additional 50 μL of DMEM was added, cells were incubated overnight at 37°C. Luciferase activity was quanti ed with Bio-Glo reagent by adding 100 μL of Bio-Glo (diluted 1:1 in PBS), incubated at room temperature for 5 min, and relative light units (RLU) were read on an EnSight or EnVision plate reader.
Neutralization of SARS-CoV-2 pseudotyped virus. Vero E6 cells were seeded into clear bottom black-walled 96-well plates at 20,000 cells/well in 100 μL medium and cultured overnight at 37°C. Twenty-four hours later, 1:3 8-point serial dilutions of mAb were prepared in medium, with each dilution tested in duplicate on each plate (range: 10 μg/mL to 4 ng/mL nal concentration). Pseudovirus was diluted 1:25 in medium and added 1:1 to 110 μL of each antibody dilution. Pseudovirus:antibody mixtures were incubated for 1 h at 37°C. Media was removed from the Vero E6 cells and 50 μL of pseudovirus:antibody mixtures were added to the cells. One hour post-infection, 100 μL of medium was added to wells containing pseudovirus:antibody mixtures and incubated for 17 h at 37°C. Media then was removed and 100 μL of Bio-Glo reagent (diluted 1:1 in DPBS) was added to each well. The plate was shaken on a plate shaker at 300 RPM at room temperature for 20 min, and RLUs were read on an EnSight or EnVision plate reader.
Data availability. All data supporting the ndings of this study are available within the paper and are available from the corresponding author upon request.
Deep sequencing datasets of viral stocks are available at NCBI BioProject PRJNA698378 (https://dataview.ncbi.nlm.nih.gov/object/PRJNA698378? reviewer=g0mic4v5t4e1tpssk63p990suu).      EC50 values were calculated from one experiment each performed in duplicate with some exceptions due to limited sample (Wilcoxon matched-pairs signed rank test, *, P < 0.05; **, P < 0.01; ***, P < 0.001; all other comparisons, not signi cant). GMT values are shown above each graph. Dotted line represents the limit of detection of the assay. p-r, Heat map of the relative neutralizing activity of sera from individual mice (p), hamsters (q), and NHPs (r) against indicated SARS-CoV-2 viruses compared to WA1/2020 D614G. Blue, reduction; red, increase. e, Heat map of the relative neutralizing activity of sera from vaccinated individuals against indicated SARS-CoV-2 viruses compared to WA1/2020 D614G. Blue, reduction; red, increase.

Figure 5
Resistance of SARS-CoV-2 viral variants to neutralization by human serum from convalescent and vaccinated individuals in Vero-TMPRSS2 cells. Sera from individuals who had been infected with SARS-CoV-2 (a-c; n = 20, ~1-month post-infection) or vaccinated with the P zer-BioNTech mRNA vaccine (d-f; n = 10) were tested for neutralization of the indicated SARS-CoV-2 strains (WA1/2020 (a, b, d, e), B.1.1.7 (a, d), or Wash SA-B.1.351 (b, e) using a FRNT in Vero-TMPRSS2 cells. EC50 values were calculated from one experiment performed in duplicate (Wilcoxon matched-pairs signed rank test, **, P < 0.01). GMT values are shown above each graph. Dotted line represents the limit of detection of the assay. c, f, Heat maps of the relative neutralizing activity of sera from convalescent (c) or vaccinated (f) individuals against indicated SARS-CoV-2 viruses compared to WA1/2020. Blue, reduction; red, increase.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.