Decision letter: Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Neutralizing antibodies elicited by prior infection or vaccination are likely to be key for future protection of individuals and populations against SARS-CoV-2. Moreover, passively administered antibodies are among the most promising therapeutic and prophylactic anti-SARS-CoV-2 agents. However, the degree to which SARS-CoV-2 will adapt to evade neutralizing antibodies is unclear. Using a recombinant chimeric VSV/SARS-CoV-2 reporter virus, we show that functional SARS-CoV-2 S protein variants with mutations in the receptor-binding domain (RBD) and N-terminal domain that confer resistance to monoclonal antibodies or convalescent plasma can be readily selected. Notably, SARS-CoV-2 S variants that resist commonly elicited neutralizing antibodies are now present at low frequencies in circulating SARS-CoV-2 populations. Finally, the emergence of antibody-resistant SARS-CoV-2 variants that might limit the therapeutic usefulness of monoclonal antibodies can be mitigated by the use of antibody combinations that target distinct neutralizing epitopes. eLife digest The new coronavirus, SARS-CoV-2, which causes the disease COVID-19, has had a serious worldwide impact on human health. The virus was virtually unknown at the beginning of 2020. Since then, intense research efforts have resulted in sequencing the coronavirus genome, identifying the structures of its proteins, and creating a wide range of tools to search for effective vaccines and therapies. Antibodies, which are immune molecules produced by the body that target specific segments of viral proteins can neutralize virus particles and trigger the immune system to kill cells infected with the virus. Several technologies are currently under development to exploit antibodies, including vaccines, blood plasma from patients who were previously infected, manufactured antibodies and more. The spike proteins on the surface of SARS-CoV-2 are considered to be prime antibody targets as they are accessible and have an essential role in allowing the virus to attach to and infect host cells. Antibodies bind to spike proteins and some can block the virus’ ability to infect new cells. But some viruses, such as HIV and influenza, are able to mutate their equivalent of the spike protein to evade antibodies. It is unknown whether SARS-CoV-2 is able to efficiently evolve to evade antibodies in the same way. Weisblum, Schmidt et al. addressed this question using an artificial system that mimics natural infection in human populations. Human cells grown in the laboratory were infected with a hybrid virus created by modifying an innocuous animal virus to contain the SARS-CoV-2 spike protein, and treated with either manufactured antibodies or antibodies present in the blood of recovered COVID-19 patients. In this situation, only viruses that had mutated in a way that allowed them to escape the antibodies were able to survive. Several of the virus mutants that emerged had evolved spike proteins in which the segments targeted by the antibodies had changed, allowing these mutant viruses to remain undetected. An analysis of more than 50,000 real-life SARS-CoV-2 genomes isolated from patient samples further showed that most of these virus mutations were already circulating, albeit at very low levels in the infected human populations. These results show that SARS-CoV-2 can mutate its spike proteins to evade antibodies, and that these mutations are already present in some virus mutants circulating in the human population. This suggests that any vaccines that are deployed on a large scale should be designed to activate the strongest possible immune response against more than one target region on the spike protein. Additionally, antibody-based therapies that use two antibodies in combination should prevent the rise of viruses that are resistant to the antibodies and maintain the long-term effectiveness of vaccines and therapies. Introduction Neutralizing antibodies are a key component of adaptive immunity against many viruses that can be elicited by natural infection or vaccination (Plotkin, 2010). Antibodies can also be administered as recombinantly produced proteins or as convalescent plasma to confer a state of passive immunity in prophylactic or therapeutic settings. These paradigms are of particular importance given the emergence of SARS-CoV-2, and the devastating COVID19 pandemic that has ensued. Indeed, interventions to interrupt SARS-CoV-2 replication and spread are urgently sought, and passively administered antibodies are currently among the most promising therapeutic and prophylactic antiviral agents. Moreover, an understanding of the neutralizing antibody response to SARS-CoV-2 is critical for the elicitation of effective and durable immunity by vaccination (Kellam and Barclay, 2020). Recent studies have shown that related, potently neutralizing monoclonal antibodies that recognize the SARS-CoV-2 receptor-binding domain (RBD) are often elicited in SARS-CoV-2 infection (Robbiani et al., 2020; Brouwer et al., 2020; Cao et al., 2020; Chen et al., 2020; Chi et al., 2020; Rogers et al., 2020; Shi et al., 2020; Wu et al., 2020a; Wec et al., 2020; Kreer et al., 2020; Hansen et al., 2020; Ju et al., 2020; Seydoux et al., 2020; Liu et al., 2020; Zost et al., 2020). These antibodies have great potential to be clinically impactful in the treatment and prevention of SARS-CoV-2 infection. The low levels of somatic hypermutation and repetitive manner in which similar antibodies (e.g. those based on IGHV3-53 Robbiani et al., 2020; Barnes et al., 2020; Yuan et al., 2020) have been isolated from COVID19 convalescents suggests that potently neutralizing responses should be readily elicited. Paradoxically, a significant fraction of COVID19 convalescents, including some from whom potent neutralizing antibodies have been cloned, exhibit low levels of plasma neutralizing activity (Robbiani et al., 2020; Wu et al., 2020b; Luchsinger et al., 2020). Together, these findings suggest that natural SARS-CoV-2 infection may often fail to induce sufficient B-cell expansion and maturation to generate high-titer neutralizing antibodies. The degree to, and pace at which SARS-CoV-2 might evolve to escape neutralizing antibodies is unclear. The aforementioned considerations raise the possibility that SARS-CoV-2 evolution might be influenced by frequent encounters with sub-optimal concentrations of potently neutralizing antibodies during natural infection. Moreover, the widespread use of convalescent plasma containing unknown, and often suboptimal, levels of neutralizing antibodies might also increase the acquisition of neutralizing antibody resistance by circulating SARS-CoV-2 populations (Bloch et al., 2020; Al‐Riyami et al., 2020). Reinfection of previously infected individuals who have incomplete or waning serological immunity might similarly drive emergence of antibody escape variants. As human neutralizing antibodies are discovered and move into clinical development as therapeutics and prophylactics, and immunogens based on prototype SARS-CoV-2 spike protein sequences are deployed as vaccines, it is important to anticipate patterns of antibody resistance that might arise. Here, we describe a recombinant chimeric virus approach that can rapidly generate and evaluate SARS-CoV-2 S mutants that escape antibody neutralization. We show that mutations conferring resistance to convalescent plasma or RBD-specific monoclonal antibodies can be readily generated in vitro. Notably, these antibody resistance mutations are present at low frequency in natural populations. Importantly, the use of candidate monoclonal antibody combinations that target distinct epitopes on the RBD (and therefore have non-overlapping resistance mutations) can suppress the emergence of antibody resistance. Results Selection of SARS-CoV-2 S variants using a replication-competent VSV/SARS-CoV-2 chimeric virus To select SARS-CoV-2 S variants that escape neutralization by antibodies, we used a recently described replication-competent chimeric virus based on vesicular stomatitis virus that encodes the SARS-CoV-2 spike (S) protein and green fluorescent protein (rVSV/SARS-CoV-2/GFP) (Schmidt et al., 2020). Notably, rVSV/SARS-CoV-2/GFP replicates rapidly and to high-titers (107 to 108 PFU/ml within 48 hr), mimics the SARS-CoV-2 requirement for ACE-2 as a receptor, and is neutralized by COVID19 convalescent plasma and SARS-CoV-2 S-specific human monoclonal antibodies (Schmidt et al., 2020). The replication of rVSV/SARS-CoV-2/GFP can be readily monitored and measured by GFP fluorescence and the absence of proof-reading activity in the viral polymerase (VSV-L) results in the generation of virus stocks with greater diversity than authentic SARS-CoV-2, for an equivalent viral population size. These features facilitate experiments to investigate the ability S protein variants to escape antibody neutralization. We used two adapted, high-titer variants of rVSV/SARS-CoV-2/GFP, (namely rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1) (Schmidt et al., 2020) in attempts to derive antibody-resistant mutants. Virus populations containing 1 × 106 infectious particles were generated following three passages to generated sequence diversity. On the third passage, cells were infected at an MOI of ~ 0.5 and progeny harvested after as short a time as possible so as to minimize phenotypic mixing in the viral population and to maximize the concordance between the genome sequence and the S protein sequence represented in a given virion particle. Because the mutation rate of VSV is ~ 10−4 to 10−5/base per replication cycle (Steinhauer and Holland, 1986; Steinhauer et al., 1989; Combe and Sanjuán, 2014), we estimated that this procedure should generate a large fraction of the possible replication-competent mutants within a population size of 1 × 106. The viral populations were then incubated with antibodies to neutralize susceptible variants (Figure 1A). For monoclonal antibodies, viral populations were incubated with antibodies at 5 μg/ml or 10 μg/ml, (~1000 to 10,000 x IC50) so as to minimize the number of infection events by antibody sensitive variants, and enable rapid selection of the most resistant rVSV/SARS-CoV-2/GFP variants from the starting population. For plasma samples, the possibility existed that multiple different antibody specificities could be present, that might interfere with the outgrowth of rVSV/SARS-CoV-2/GFP variants that were resistant to the most prevalent or potent antibodies in the plasma. Therefore, in these selection experiments, viruses were incubated with a range of plasma dilutions (see materials and methods). Neutralized viral populations were then applied to 293T/ACE2(B) cells (Schmidt et al., 2020), which support robust rVSV/SARS-CoV-2/GFP replication, and incubated for 48 hr. We used three potent human monoclonal antibodies C121, C135, and C144 (Robbiani et al., 2020), that are candidates for clinical development (Table 1). In addition, we used four convalescent plasma samples, three of which were from the same donors from which C121, C135, and C144, were obtained (Robbiani et al., 2020; Table 1). Two of these plasmas (COV-47 and COV-72) were potently neutralizing while the third (COV-107) had low neutralizing activity. A fourth convalescent plasma sample (COV-NY) was potently neutralizing but did not have a corresponding monoclonal antibody (Table 1). Figure 1 Download asset Open asset Selection of SARS-CoV-2 S mutations that confer antibody resistance. (A) Outline of serial passage experiments with replication-competent VSV derivatives encoding the SARS-CoV-2 S envelope glycoprotein and a GFP reporter (rVSV/SARS-CoV-2/GFP) in 293T/ACE2(B) cells in the presence of neutralizing antibodies or plasma. Each passage experiment was performed twice (once each with rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1.) (B) Representative images of 293T/ACE2(B) cells infected with 1 × 106 PFU of rVSV/SARS-CoV-2/GFP in the presence or absence of 10 μg/ml of the monoclonal antibody C121. (C) Expanded view of the boxed areas showing individual plaques of putatively antibody-resistant viruses. Table 1 Plasma and monoclonal antibodies used in this study. DonorPlasma NT50 (rVSV-SARSCoV2/GFP)Plasma NT50 HIV/CCNGnLucMonoclonal antibodyCOV-4766228016C144COV-7262747982C135COV-107122334C121COV-NY126147505ND Infection with rVSV/SARS-CoV-2/GFP in the presence of the monoclonal antibodies C121 or C144 reduced the number of infectious units from 106 to a few hundred, as estimated by the frequency of GFP-positive cells (Figure 1B) a reduction of > 1000 fold. C135 reduced infection by ~ 100 fold. Imaging of wells infected with rVSV/SARS-CoV-2/GFP in the presence of C121 or C144 revealed a small number of foci (~10 to 20/well), that suggested viral spread following initial infection (Figure 1B). In the case of C135, a greater number of GFP-positive cells were detected, obscuring the visualization of focal viral spread following initial infection. Aliquots of supernatants from these passage-1 (p1) cultures were collected 48 hr after infection, diluted in the same concentrations of monoclonal antibodies that were initially employed, and used to infect fresh (p2) cultures (Figure 1A). For p2 cultures, almost all cells became infected within 48 hr, suggesting the possible outgrowth of monoclonal antibody escape variants that were present in the original viral populations. For selection in the presence of plasma, p1 supernatants were harvested at 48 hr after infection in the presence of the highest concentrations of plasma that permitted infection of reasonable numbers (approximately 10%) of cells. Then, p2 cultures were established using p1 supernatants, diluted in the same concentrations of plasma used in p1. This approach led to clear ‘escape’ for the COV-NY plasma with prolific viral growth in p2 as evidenced by a large increase in the number of GFP-positive cells. For COV-47, COV-72, and CO107, plasma clearly retained at least some inhibitory activity in p2. Thereafter, p3 cultures and p4 cultures were established for COV-47, COV-72, and COV-107 plasmas at 5-fold higher concentrations of plasma than were used in p1 and p2 cultures (Figure 1A). RNA was extracted from p2 supernatants (monoclonal antibodies and COV-NY plasma) as well as later passages for the COV-47, COV-72, and COV-107 plasma selections. Sequences encoding either the RBD or the complete S protein were amplified using PCR and analyzed by Sanger and/or Illumina sequencing. For all three monoclonal antibodies and two of the four plasmas, sequence analyses revealed clear evidence for selection, with similar or identical mutants emerging in the presence of monoclonal antibodies or plasma in both rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1 cultures (Figure 2A–D, Figure 2—figure supplement 1A,B, Figure 2—figure supplement 2A,B, Table 2). In the case of C121, mutations E484K and Q493K/R within the RBD were present at high frequencies in both p2 selected populations, with mutation at a proximal position (F490L) present in one p2 population (Figure 2A, Table 2). Viruses passaged in the presence of monoclonal antibody C144 also had mutations at positions E484 and Q493, but not at F490 (Figure 2C, Table 2). In contrast, virus populations passaged in the presence of monoclonal antibody C135 lacked mutations at E484 or Q493, and instead had mutations R346K/S/L and N440K at high frequency (Figure 2C, Table 2). Figure 2 with 2 supplements see all Download asset Open asset Analysis of S-encoding sequences following rVSV/SARS-CoV-2/GFP replication in the presence of neutralizing monoclonal antibodies. (A–D) Graphs depict the S codon position (X-axis) and the frequency of non-synonymous substitutions (Y-axis) following the second passage (p2) of rVSV/SARS-CoV-2/GFP on 293T/ACE2(B) cells in the absence of antibody or plasma (A), or in the presence of 10 μg/ml C121 (B), C135 (C) or C144 (D). Results are shown for both rVSV/SARS-CoV-2/GFP variants (One replicate each for rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1 - the frequency of 1D7 mutations is plotted as circles and 2E1 mutations as triangles). Table 2 Mutations occurring at high frequency during rVSV/SARS-CoV-2 passage in the presence of neutralizing antibodies or plasma. Mutant frequencyMutationp2p3p4Monoclonal antibodiesC121E484K*0.50, 0.39–†–F490L0.23Q493K0.12, 0.45C135N440K0.31, 0.30––R346S0.30, 0.17R346K0.22, 0.53R346M0.16C144E484K0.44, 0.18––Q493K0.31, 0.39Q493R0.17, 0.37PlasmasCOV47N148S0.16, 0.140.29, 0.300.72, 0.14K150R0.100.18K150E0.040.160.4K150T0.22K150Q0.160.22S151P0.10.180.2COV-NYK444R0.20,0.19––K444N0.14K444Q0.33V445E0.18 *Values represent the decimal frequency with which each mutation occurs ass assessed by NGS, two values indicate occurrences in both rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1 cultures, single values indicate occurrence in only one culture. † –, not done. Mutations at specific positions were enriched in viruses passaged in the presence of convalescent plasma, in two out of four cases (Figure 2—figure supplement 1A,B, Figure 2—figure supplement 2A,B, Table 2). Specifically, virus populations passaged in the presence of COV-NY plasma had mutations within RBD encoding sequence (K444R/N/Q and V445E) that were abundant at p2 (Figure 2—figure supplement 2A,B, Table 2). Conversely, mutations outside the RBD, specifically at N148S, K150R/E/T/Q and S151P in the N-terminal domain (NTD) were present at modest frequency in COV-47 p2 cultures and became more abundant at p3 and p4 (Figure 2—figure supplement 1A,B, Table 2). Replication in the presence of COV72 or COV107 plasma did not lead to the clear emergence of escape mutations, suggesting that the neutralization by these plasmas was not due to one dominant antibody specificity. In the case of COV107, the failure of escape mutants to emerge may simply be due to the lack of potency of that plasma (Table 1). However, in the case of COV-72, combinations of antibodies may be responsible for the potent neutralizing properties of the plasma in that case. Isolation and characterization of rVSV/SARS-CoV-2/GFP antibody escape mutants Based on the aforementioned analyses, supernatants from C121, C144, and C135 and COV-NY plasma p2 cultures, or COV47 p4 cultures, contained mixtures of putative rVSV/SARS-CoV-2/GFP neutralization escape mutants. To isolate individual mutants, the supernatants were serially diluted and individual viral foci isolated by limiting dilution in 96-well plates. Numerous individual rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1 derivatives were harvested from wells containing a single virus plaque, expanded on 293T/ACE2(B) cells, then RNA was extracted and subjected to Sanger-sequencing (Figure 3—figure supplement 1). This process verified the purity of the individual rVSV/SARS-CoV-2/GFP variants and yielded a number of viral mutants for further analysis (Figure 3—figure supplement 1). These plaque-purified viral mutants all encoded single amino-acid substitutions in S-coding sequences that corresponded to variants found at varying frequencies (determined by Illumina sequencing) in the antibody-selected viral populations. Notably, each of the isolated rVSV/SARS-CoV-2/GFP mutants replicated with similar kinetics to the parental rVSV/SARS-CoV-2/GFP1D7 and rVSV/SARS-CoV-2/GFP2E1 viruses (Figure 3A), suggesting that the mutations that emerged during replication in the presence of monoclonal antibodies or plasma did not confer a substantial loss of fitness, at least in the context of rVSV/SARS-CoV-2/GFP. Moreover, for mutants in RBD sequences that arose in the C121, C135, C144, and COV-NY cultures, each of the viral mutants retained approximately equivalent sensitivity to neutralization by an ACE2-Fc fusion protein, suggesting little or no change in interaction with ACE2 (Figure 3B). Figure 3 with 1 supplement see all Download asset Open asset Characterization of mutant rVSV/SARS-CoV-2/GFP derivatives. (A) Replication of plaque-purified rVSV/SARS-CoV-2/GFP bearing individual S amino-acid substitutions that arose during passage with the indicated antibody or plasma. 293T/ACE2cl.22 cells were inoculated with equivalent doses of parental or mutant rVSV/SARS-CoV-2/GFP isolates. Supernatant was collected at the indicated times after inoculation and number of infectious units present therein was determined on 293T/ACE2cl.22 cells. The mean of two independent experiments is plotted. One set of WT controls run concurrently with the mutants are replotted in the upper and lower left panels, A different set of WT controls run concurrently with the mutants is shown in the lower right panel (B) Infection 293T/ACE2cl.22 cells by rVSV/SARS-CoV-2/GFP encoding the indicated S protein mutations in the presence of increasing amounts of a chimeric ACE2-Fc molecule. Infection was quantified by FACS. Mean of two independent experiments is plotted. The WT controls are replotted in each panel. We next determined the sensitivity of the isolated RBD mutants to neutralization by the three monoclonal antibodies. The E484K and Q493R mutants that emerged during replication in the presence of C121 or C144, both caused apparently complete, or near complete, resistance to both antibodies (IC50 > 10 μg/ml, Figure 4A,B). However, both of these mutants retained full sensitivity (IC50 < 10 ng/ml) to C135. Conversely, the R346S and N440K mutants that emerged during replication in the presence of C135 were resistant to C135, but retained full sensitivity to both C121 and C144 (Figure 4A,B). The K444N, K444T, V445G, V445E, and V445L mutants that arose during replication in the presence of COV-NY plasma conferred partial resistance to C135, with IC50 values ranging from 25 to 700 ng/ml, but these mutants retained full sensitivity to both C121 and C144 (Figure 4A,B). The spatial distribution of these resistance-conferring mutations on the SARS-CoV-2 S RBD surface suggested the existence of both distinct and partly overlapping neutralizing epitopes on the RBD (Figure 4C). The C121 and C144 neutralizing epitopes appear to be similar, and include E484 and Q493, while a clearly distinct conformational epitope seems to be recognized by C135, that includes R346 and N440 residues. Antibodies that constitute at least part of the neutralizing activity evident in COV-NY plasma appear to recognize an epitope that includes and K444 and V445. The ability of mutations at these residues to confer partial resistance to C135 is consistent with their spatial proximity to the C135 conformational epitope (Figure 4C). Figure 4 Download asset Open asset Neutralization of rVSV/SARS-CoV-2/GFP RBD mutants by monoclonal antibodies. (A) Examples of neutralization resistance of rVSV/SARS-CoV-2/GFP mutants that were isolated following passage in the presence of antibodies. 293T/ACE2cl.22 cells were inoculated with WT or mutant rVSV/SARS-CoV-2/GFP in the presence of increasing amount of each monoclonal antibody, and infection quantified by FACS 16 hr later. Mean and SD from two technical replicates, representative of two independent experiments. (B) Neutralization sensitivity/resistance of rVSV/SARS-CoV-2/GFP mutants isolated following replication in the presence of antibodies. Mean IC50 values were calculated for each virus-monoclonal antibody combination in two independent experiments. (C) Position of neutralization resistance-conferring substitutions. Structure of the RBD (from PDB 6M17 Yan et al., 2020) with positions that are occupied by amino acids where mutations were acquired during replication in the presence of each monoclonal antibody or COV-NY plasma indicated. To test whether neutralization escape mutations conferred loss of binding to the monoclonal antibodies, we expressed conformationally prefusion-stabilized S-trimers (Hsieh et al., 2020), appended at their C-termini with NanoLuc luciferase (Figure 5A). The S-trimers were incubated in solution with the monoclonal antibodies, complexes were captured using protein G magnetic beads, and the amount of S-trimer captured was measured using NanoLuc luciferase assays (Figure 5A). As expected, C121, C135, and C144 monoclonal antibodies all bound the WT S-trimer (Figure 5B). The E484K and Q493R trimers exhibited complete, or near complete loss of binding to C121 and C144 antibodies but retained WT levels of binding to C135 (Figure 5B). Conversely, the R346S and N440K mutants exhibited complete loss of binding to C135, but retained WT levels of binding to C121 and C144. The K444N and V445E mutants retained near WT levels of binding to all three antibodies, despite exhibiting partial resistance to C135 (Figure 5A,B). Presumably the loss of affinity of these mutants for C135 was sufficient to impart partial neutralization resistance but insufficient to abolish binding in the solution binding assay. Figure 5 Download asset Open asset Loss of binding to monoclonal antibodies by neutralization escape mutants. (A) Schematic representation of the binding assay in which NanoLuc luciferase is appended to the C-termini of a conformationally stabilized S-trimer. The fusion protein is incubated with antibodies and complexes captured using protein G magnetic beads (B) Bound Nanoluc luciferase quantified following incubation of the indicated WT or mutant Nanoluc-S fusion proteins with the indicated antibodies and Protein G magnetic beads. Mean of three technical replicates at each S-Nanoluc concentration. Analysis of mutants that were isolated from the virus population that emerged during rVSV/SARS-CoV-2/GFP replication in the presence of COV-47 plasma (specifically N148S, K150R, K150E, S151P) revealed that these mutants exhibited specific resistance to COV-47 plasma. Indeed, the COV-47 plasma NT50 for these mutants was reduced by 8- to 10-fold (Figure 6A). This finding indicates that the antibody or antibodies responsible for majority of neutralizing activity in COV-47 plasma target an NTD epitope that includes amino acids 148–151, even though the highly potent monoclonal antibody (C144) isolated from COV-47 targets the RBD. Mutants in the 148–151 NTD epitope exhibited marginal reductions in sensitivity to other plasmas (Figure 6A), indicating that different epitopes are primarily targeted by plasmas from the other donors. Figure 6 Download asset Open asset Neutralization of rVSV/SARS-CoV-2/GFP RBD mutants by convalescent plasma. (A, B) Neutralization of rVSV/SARS-CoV-2/GFP mutants isolated following replication in the presence COV-47 plasma (A) or COV-NY plasma (B). 293T/ACE2cl.22 cells were inoculated with WT or mutant rVSV/SARS-CoV-2/GFP in the presence of increasing amounts of the indicated plasma, and infection quantified by flow cytometry, 16 hr later. Mean of two technical replicates, representative of two independent experiments (C) Plasma neutralization sensitivity/resistance of rVSV/SARS-CoV-2/GFP mutants isolated following replication in the presence of monoclonal antibodies or convalescent plasma. Mean NT50 values were calculated for each virus-plasma combination from two independent experiments. The viral population that emerged during replication in COV-NY plasma yielded mutants K444N or T and V445G, E or L. Each of these mutations conferred substantial resistance to neutralization by COV-NY plasma, with ~ 10 to 30-fold reduction in NT50 (Figure 6B). Thus, the dominant neutralizing activity in COV-NY plasma is represented by an antibody or antibodies recognizing an RBD epitope that includes K444 and V445. As was the case with COV-47 resistant mutants, viruses encoding the mutations conferring resistance to COV-NY plasma retained almost full sensitivity to neutralization by other plasmas (Figure 6B). Interestingly, the mutations that conferred complete or near complete resistance to the potent RBD-specific monoclonal antibodies C144, C135, and C121 conferred little or no resistance to neutralization by plasma from the same individual, or other individuals (Figure 6C). These RBD-specific antibodies represent the most potent monoclonal antibodies isolated from COV-47, COV-72, and COV-107, respectively, but the retention of plasma sensitivity by the monoclonal antibody-resistant mutants suggests that these antibodies contribute little to the overall neutralization activity of plasma from the same individual. This finding is consistent with the observation that memory B cells producing these antibodies are rare (Robbiani et al., 2020), and with the results of the selection experiments in which rVSV/SARS-CoV-2/GFP replication in the presence of COV-47, COV-72, and COV-107 plasma did not enrich for mutations that correspond to the neutralization epitopes targeted by the monoclonal antibodies obtained from these individuals (Figure 2—figure supplement 1A,B, Figure 2—figure supplement 2A,B. Table 2). Overall, analysis of even this limited set of monoclonal antibodies and plasmas shows that potent neutralization can be conferred by antibodies that target diverse SARS-CoV-2 epitopes. Moreover, the most potently neutralizing antibodies generated in a given COVID19 convalescent individual may contribute in only a minor way to the overall neutralizing antibody response in that same individual (see discussion). Natural occurrence of antibody-resistance RBD mutations The aforementioned neutralizing antibody escape mutations were artificially generated during in vitro replication of a recombinant virus. However, as monoclonal antibodies are developed for therapeutic and prophylactic app