SARS-CoV-2 escape in vitro from a highly neutralizing COVID-19 convalescent plasma.

To investigate the evolution of SARS-CoV-2 in the immune population, we co-incubated authentic virus with a highly neutralizing plasma from a COVID-19 convalescent patient. The plasma fully neutralized the virus for 7 passages, but after 45 days, the deletion of F140 in the spike N-terminal domain (NTD) N3 loop led to partial breakthrough. At day 73, an E484K substitution in the receptor-binding domain (RBD) occurred, followed at day 80 by an insertion in the NTD N5 loop containing a new glycan sequon, which generated a variant completely resistant to plasma neutralization. Computational modeling predicts that the deletion and insertion in loops N3 and N5 prevent binding of neutralizing antibodies. The recent emergence in the United Kingdom and South Africa of natural variants with similar changes suggests that SARS-CoV-2 has the potential to escape an effective immune response and that vaccines and antibodies able to control emerging variants should be developed. ONE SENTENCE
SUMMARY: Three mutations allowed SARS-CoV-2 to evade the polyclonal antibody response of a highly neutralizing COVID-19 convalescent plasma.

The SARS-CoV-2 virus, causative agent of COVID-19, accounts for over 78.5 million cases of infections and almost 2 million deaths worldwide. Thanks to an incredible scientific and financial effort, several prophylactic and therapeutic tools, such as vaccines and monoclonal antibodies (mAbs), have been developed in less than one year to combat this pandemic (1–4). The main target of vaccines and mAbs is the SARS-CoV-2 spike protein (S-protein), a large class I trimeric fusion protein which plays a key role in viral pathogenesis (3, 5, 6). The SARS-CoV-2 S-protein is composed of two subunits: S1, which contains the receptor-binding domain (RBD) responsible for the interaction with receptors on the host cells, and S2, which mediates membrane fusion and viral entry (7, 8). The S1 subunit presents two highly immunogenic domains, the N-terminal domain (NTD) and the RBD, which are the major targets of polyclonal and monoclonal neutralizing antibodies (4, 9, 10). The continued spread in immune-competent populations has led to adaptations of the virus to the host and generation of new SARS-CoV-2 variants. Indeed, S-protein variants have been recently described in the United Kingdom and South Africa (11, 12), and the Global Initiative on Sharing All Influenza Data (GISAID) database, reports more than 1,100 amino acid changes in the S-protein (13, 14).

An important question for vaccine development is whether the authentic virus, under the selective pressure of the polyclonal immune response in convalescent or vaccinated people, can evolve to escape herd immunity and antibody treatment. To address this question, we collected plasma from 20 convalescent patients and incubated the authentic SARS-CoV-2 wild-type (WT) virus for more than 90 days in the presence of a potent neutralizing plasma. Enzyme-linked immunosorbent assay (ELISA) showed that all plasmas collected bound the SARS-CoV-2 S-protein trimer and most of them also bound the S1 and S2 subunits. However, a broad range of reactivity profiles were noticed ranging from weak binders with titers of 1/80 to strong binders with titers of 1/10240 (Fig S1A; Table S1). PT008, PT009, PT015, PT122 and PT188 showed the strongest binding towards the S trimer and among them PT188 had also the highest binding to the S1 and S2 subunits. All but one plasma sample (PT103) were able to bind the S-protein S1 subunit. Neutralization activity tested against the SARS-CoV-2 WT and D614G variant also showed variable titers. Most of the plasma samples neutralized the viruses with titers ranging from 1/20 to 1/320. Four samples had extremely low titers (1/10), whereas sample PT188 showed extremely high titers (1/10240). Four plasma samples did not show neutralization activity against the SARS-CoV-2 WT and SARS-CoV-2 D614G variant. Plasma from subject PT188, which had the highest neutralizing titer and ELISA binding reactivity (Fig. S1B, C, D; Table 1), was selected to test whether SARS-CoV-2 can evolve to escape a potent humoral immunity.

Two-fold dilutions of plasma PT188 ranging from 1/10 to 1/20480 were co-incubated with 105 TCID50 of the wild type virus in a 24-well plate. This viral concentration is approximately three logs more than what is conventionally used in microneutralization assays (15–19). The plasma/virus mixture was co-incubated for 5–8 days. Then, the first well showing cytopathic effect (CPE) was diluted 1:100 and incubated again with serial dilutions of plasma PT188 (Fig. 1A; Table S2). For 6 passages and 38 days PT188 plasma neutralized the virus with a titer of 1/640 and did not show any sign of escape. However, after 7 passages and 45 days, the neutralizing titer decreased to 1/320. Sequence analyses revealed a deletion of the phenylalanine in position 140 (F140) on the S-protein NTD N3 loop in 36% of the virions (Fig. 1, B and C; Table S2). In the subsequent passage (P8), this mutation was observed in 100% of the sequenced virions and an additional 2-fold decrease in neutralization activity was observed reaching an overall neutralization titer of 1/160. Following this initial breakthrough, a second mutation occurred after 12 passages and 80 days of plasma/virus co-incubation (P12). This time, the glutamic acid in position 484 of the RBD was substituted with a lysine (E484K). This mutation occurred in 100% of sequenced virions and led to a 4-fold decrease in neutralization activity which reached a titer of 1/40 (Fig. 1, B and C; Table S2). The E484K substitution was rapidly followed by a third and final change comprising an 11-amino-acid insertion between Y248 and L249 in the NTD N5 loop (248aKTRNKSTSRRE248k). The insertion contained an N-linked glycan sequon (248dNKS248f), and this viral variant resulted in complete abrogation of neutralization activity by the PT188 plasma sample. Initially this insertion was observed in only 49% of the virions but when the virus was kept in culture for another passage (P14) the insertion was fully acquired by the virus (Fig. 1, B and C; Table S2).

Fig. 1.

Evolution of an authentic SARS-CoV-2 escape mutant.

(A) Schematic representation of the 24-well plate format used to select the authentic SARS-CoV-2 escape mutant. Blue, red, green and yellow wells show feeder cells protect from PT188 neutralization, CPE, authentic virus on Vero E6 cells and Vero E6 alone, respectively. (B) The graph shows the PT188 neutralization titer after each mutation acquired by the authentic virus. Specific mutations, fold decrease and days to which the mutations occur are reported in the figure. (C) SARS-CoV-2 S-protein gene showing type, position of mutations and frequency of mutations.

To evaluate the ability of the SARS-CoV-2 PT188 escape mutant (PT188-EM) to evade the polyclonal antibody response, all twenty plasma samples from COVID-19 convalescent patients were tested in a traditional CPE-based neutralization assay against this viral variant using the virus at 100 TCID50. All samples showed at least a 2-fold decrease in neutralization activity against SARS-CoV-2 PT188-EM (Fig. 2A; Fig. S1B, C, D; Table S1). As expected, the plasma used to select the escape mutant showed the biggest neutralization decrease against this escape mutant with a 256-fold decrease compared to wild-type SARS-CoV-2. Plasma PT042, PT006, PT005, PT012 and PT041 also showed a substantial drop in neutralization efficacy (Table S1). In addition, we observed that a higher response towards the S-protein S1-subunit correlates with loss of neutralization activity against SARS-CoV-2 PT188-EM (Fig. S2A) whereas a high response towards the S-protein S2-subunit did not show correlation (Fig. S2B).

Fig. 2.

Neutralization efficacy of plasma and thirteen mAbs to SARS-CoV-2 PT188-EM.

(A) Heat-map showing the neutralization activity of tested plasma samples to the SARS-CoV-2 WT, D614G and PT188-EM variants. (B) Heat-maps showing neutralization profiles of tested mAbs. (C) Negative stain EM 2D class averages showing J13, I21 and H20 Fabs bound to the SARS-CoV-2 S-protein. (D) 3D reconstruction of J13 bound to the NTD domain of the S-protein viewed looking along (left panel) or toward (right panel) the viral membrane.

We also tested a previously identified panel of thirteen neutralizing monoclonal antibodies (nAbs) (19) by CPE-based neutralization assay to assess their neutralization efficacy against SARS-CoV-2 PT188-EM. These antibodies were classified in three groups based on their binding profiles to the S-protein. Group I nAbs were able to bind the S1-RBD, Group II targeted the S1-subunit but not the RBD, and Group III nAbs were specific for the S-protein trimer (Table S3). These antibodies also showed a variable neutralization potency against the SARS-CoV-2 WT and D614G viruses ranging from 3.9 ng/mL to 500.0 ng/mL (Fig. 2B; Fig. S2E, F, G; Table S3). The three mutations selected by SARS-CoV-2 PT188-EM to escape the highly neutralizing plasma completely abrogated the neutralization activity of two of the six tested RBD-directed antibodies (F05 and G12) (Fig. 2B; Fig. S2E, F, G; Table S3), suggesting that their epitopes include E484. In contrast, the extremely potent neutralizing antibody J08 was the most potently neutralizing antibody against this escape mutant with an IC100 of 22.1 ng/mL. Interestingly, the S1-RBD-directed antibody C14 showed a 2-fold increase in neutralization activity compared to the SARS-CoV-2 WT virus whereas I14 and B07 showed a 16- and 2-fold decrease, respectively. All tested antibodies derived from Group II (S1-specific not RBD) and Group III (S-protein trimer specific) completely lost their neutralization ability against SARS-CoV-2 PT188-EM (Fig. 2B; Fig. S2E, F, G; Table S3). To better understand the abrogation of activity of some of the tested antibodies, J13, I21 and H20 were co-complexed with SARS-CoV-2 WT S-protein and structurally evaluated by negative-stain EM. 2D class averages of the three tested antibodies showed that they all bind to the NTD of the S-protein (Fig. 2C). A 3D reconstruction for the J13 Fab complex provided further evidence that this antibody binds to the NTD (Fig. 2D).

Computational modeling and simulation of the WT and PT188-EM spikes provides a putative structural basis for understanding antibody escape. The highly antigenic NTD is more extensively mutated, containing both the F140 deletion as well as the 11-amino-acid insertion in loop N5 that introduces a novel N-glycan sequon at position N248d (Fig. 3, A – C). In contrast, the single mutation in the RBD (E484K) swaps the charge of the sidechain, which would significantly alter the electrostatic complementarity of antibody binding to this region (Fig. 3D). Upon inspection of molecular dynamics simulations of the NTD escape mutant model, we hypothesize that the F140 deletion alters the packing of the N1, N3 and N5 loops (Fig. S3), where the loss of the bulky aromatic sidechain would overall reduce the stability of this region (Table S1). Subsequently, the extensive insertion within the N5 loop appears to remodel this critical antigenic region, predicting substantial steric occlusion with antibodies targeting this epitope, such as antibody 4A8 (Fig. 3B) (20). Furthermore, introduction of a new N-glycan at position N248d (mutant numbering scheme) would effectively eliminate neutralization by such antibodies (Fig. 3B and S4).

Fig. 3.

In-silico modeling of the PT188-EM spike NTD and RBD.

(A) In-silico model of the NTD of the SARS-CoV-2 PT188-EM spike protein based on PDB id 7JJI. This model accounts for the 11-amino-acid insertion (yellow ribbon) and F140 deletion (highlighted with a yellow bead). N5 loop as in the wild type cryo-EM structure (PDB id: 7JJI) is shown as a transparent red ribbon. (B) Close-up of the PT188-EM spike NTD model in complex with antibody 4A8. Both heavy chain (HC, light gray) and light chain (LC, dark gray) of 4A8 are shown. The 11-amino-acid insertion (yellow ribbon) within N5 loop introduces a new N-linked glycan (N248d) that sterically clashes with 4A8, therefore disrupting the binding interface. The N-glycan at position N149 is however compatible with 4A8 binding. (C) Conformational dynamics of the PT188-EM spike NTD model resulting from 100 ns of molecular dynamics simulation is shown by overlaying multiple frames along the generated trajectory. (D) In-silico model of the PT188-EM spike RBD based on PDB id 6M17, where the E484K mutation is shown with licorice representation.

To determine the extent to which the escape mutations were detrimental to the infectivity of SARS-CoV-2 PT188-EM, the viral fitness was evaluated. Four different measures were assessed: visible CPE, viral titer, RNA-dependent RNA polymerase (RdRp) and nucleocapsid (N) RNA detection by reverse transcription polymerase chain reaction (RT-PCR) (Fig. S5). Initially, the SARS-CoV-2 WT virus and the PT188-EM variant were inoculated at a multiplicity of infection (MOI) of 0.001 on Vero E6 cells. Every day, for four consecutive days, a titration plate was prepared and optically assessed after 72 hours of incubation to evaluate the CPE effect on Vero E6 cells and viral titer. Furthermore, the RNA was extracted to asses RdRp and N-gene levels in the supernatant. We collected pictures at 72 hours post-infection to evaluate the morphological status of non-infected Vero E6 cells and the CPE on infected feeder cells. Vero E6 cells were confluent at 72 hours and no sign of CPE was optically detectable (Fig S5A). On the contrary, SARS-CoV-2 WT and PT188-EM showed significant and comparable amount of CPE (Fig. S5A). Viral titers were evaluated for both SARS-CoV-2 WT and PT188-EM, and no significant differences were observed as the viruses showed almost identical growth curves (Fig. S5B). A similar trend was observed when RdRp and N-gene levels in the supernatant were detected, even if slightly higher levels of RdRp and N-gene were detectable for SARS-CoV-2 PT188-EM at Day 0 and Day 1 (Fig. S5C). Finally, strong correlations between viral titers and RdRp/N-gene levels were observed for both SARS-CoV-2 WT and PT188-EM (Fig. S5D – E).

In conclusion, we have shown that the authentic SARS-CoV-2 virus, if constantly pressured, has the ability to escape even a potent polyclonal serum targeting multiple neutralizing epitopes. These results are remarkable because while escape mutants can be easily isolated when viruses are incubated with single monoclonal antibodies, usually a combination of two mAbs is sufficient to eliminate the evolution of escape variants and because SARS-CoV-2 shows a very low estimated evolutionary rate of mutation as this virus encodes a proofreading exoribonuclease machinery (21, 22)(23, 24). The recent isolation of SARS-CoV-2 variants in the United Kingdom and South Africa with deletions in or near the NTD loops shows that what we describe here can occur in the human population. The ability of the virus to adapt to the host immune system was also observed in clinical settings where an immunocompromised COVID-19 patient, after 154 days of infection, presented different variants of the virus including the E484K substitution (25). Therefore, we should be prepared to deal with virus variants that may be selected by the immunity acquired from infection or vaccination. This can be achieved by developing second-generation vaccines and monoclonal antibodies, possibly targeting universal epitopes and able to neutralize emerging variants of the virus.

Our data also confirm that the SARS-CoV-2 neutralizing antibodies acquired during infection target almost entirely the NTD and the RBD. In the RBD, the possibility to escape is limited and the mutation E484K that we found is one of the most frequent mutations to escape monoclonal antibodies and among the most common RBD mutations described in experimental settings as well as in natural isolates posted in the GISAD database (13, 26, 27). This is likely due to residue E484 being targeted by antibodies derived from IGHV3–53 and closely related IGHV3–66 genes, which are the most common germlines for antibodies directed against the RBD (28). On the other hand, the NTD loops can accommodate many different changes, such as insertions, deletions and amino-acid alterations. Interestingly, in our case, the final mutation contained an insertion carrying an N-glycosylation site which has the potential to hide or obstruct the binding to neutralizing epitopes. The introduction of a glycan is a well-known immunogenic escape strategy described in influenza (29), HIV-1 and other viruses (30–32), although to our knowledge this finding presents the first patient-derived escape mutant utilizing this mechanism for SARS-CoV-2. Surprisingly, only three mutations, which led to complete rearrangement of NTD N3 and N5 loops and substitution to a key residue on the RBD, were sufficient to eliminate the neutralization ability of a potent polyclonal serum. Fortunately, not all plasma and mAbs tested were equally affected by the three mutations suggesting that natural immunity to infection can target additional epitopes that can still neutralize the PT188-EM variant. Therefore, it will be important to closely monitor which epitopes on the S-protein are targeted by the vaccines against SARS-CoV-2 that are going to be deployed in hundreds of millions of people around the world.