The IgG titres provided by the CLIA test are shown in Figure 1, Figure 2, and Supplementary Table S4. According to the manufacturer, output results were considered positive for the presence of anti-SARS-CoV-2-IgG antibodies for values >15 AU/mL, negative for values <12 AU/mL, and as borderline for values between 12-15 AU/mL. Among convalescent COVID-19 patients tested by CLIA, 34/40 (85%) sera were IgG positive, 2/40 (5%) (SA-1, SA-6) were considered borderline, and 4/40 (10%) were negative (Supplementary Table S1). By arbitrarily classifying the patients with high titres of antibodies (IgG titres >100), the prevalence of the high titres decreased among patients infected by the original/B genotype (8/9, 89%), the UK/B.1.1.7 genotype (3/10, 30%), the Marseille-4/B.1.160 genotype (1/9, 11%) and the South African/B.1.351 genotype (0/12, 0%). Moreover, a significant proportion (3/12, 25%) of the patients infected with the South African/B.1.351 variant were negative for IgG by CLIA, and these samples represented 75% of all samples that tested negative.
For MNT, the observed titres were low, ranging from no seroneutralization (<1/5) to 1/160 (Figure 1). The same was found for CLIA in terms of the reactivity of the different groups of convalescent sera. By arbitrarily classifying the patients with high IgG antibody titres > 1/10 against the strain that the patients were infected with, the decreasing order of the prevalence of titres were: the original/B genotype (9/11, 82%), UK/B.1.1.7 genotype (8/10, 80%), Marseille-4/B.1.160 genotype (2/9, 22%) and South African/B.1.351 genotype (3/12, 25%). The presence of neutralizing antibodies in each group of patients against the variants, excluding those variants responsible for the patients’ infection, varied according to the variant tested. Without taking into account the sera of patients convalescing for the South African/B.1.351 variant that react nearly only to South African/B.1.351 and Brazilian/B.220.127.116.11 variants, at a 1/5 MSN titre, 14/30 (47%) of the samples reacted against the Original/B genotype, 10/19 (53%) reacted against the Marseille-4/B.1.160 genotype, 13/20 (65%) reacted against the UK/B.1.1.7 genotype, 20/30 (66%) reacted against the Marseille-501/A.27 variant, 18/30 (60%) reacted against the Brazilian/B.1.28.1 variant, 15/60 (50%) reacted against the Marseille-484K.V1/R.1 variant, 24/30 (80%) reacted against the Belgian/B.1.214 variant and 12/60 (50%) reacted against the Indian B.6K.V1 variant. The South African/B.1.351 variant was the least recognized variant after excluding the variants that caused the infections, as only 8/30 (27%) had detectable seroneutralising antibodies against this variant.
Vaccinated patients’ sera
Regarding the vaccinated participants and their CLIA serology test results, both participants who received the 2 AZD122 injections were IgG positive, and the majority (11/12) of the participants were vaccinated with 2 doses of the Pfizer/BioNTech vaccine (Figure 1, Figure 2 and Supplementary table S5). However, one patient (V-Pfizer-10) who showed no detectable reaction by the seroneutralization tests, even against the original Original/B strain despite the patient previously receiving 2 doses of the Pfizer/BioNTech vaccine, was also negative for IgG in the CLIA. This patient was not immunocompromised but was an elderly patient (88 years old). The other patient with a low antibody titre (V-Pfizer 4) had a previous splenectomy.
The neutralizing profiles of most patients who had the Pfizer/BioNTech vaccine also showed neutralization gaps in the South African/B.1.351 variant (Figure 1 and Figure 2). Otherwise, these sera appeared to inhibit the in vitro CPE for 9 out of the 10 SARS-CoV-2 strains until the sera dilutions were 1:40 and 1:80. Both persons who received two shots of the AZD122 vaccine displayed limited to completely absent neutralization on all the tested SARS-CoV-2 isolates. One Astra-2 serum showed a stronger reaction with the Belgian/B.1.214 variant isolate. The results of individuals who received the Pfizer/BioNTech vaccine also showed heterogeneity in the neutralization profiles, as some had much weaker antibody titres than those vaccinated and had very high IgG titres (>400 AU/mL). Of interest for the current period, 8/11 patients vaccinated by the Pfizer/BioNTech vaccine had MSN titres > 1/10 against both of the Indian/B.1.617.2 variant strains tested.
Human monoclonal antibody LY-CoV555
We assayed the neutralizing activity of the commercial monoclonal antibody bamlanivimab (LY-CoV555) at an initial concentration of 35 mg/mL, and the results of the neutralization activity of this tested mAb are summarized in the first row in Figure 3. LY-CoV555 significantly neutralized the Original/B strain and the Marseille-4/B.1.160 variant (neutralizing titre of 0.224 μg/mL) and less significantly neutralized the UK/B.1.1.7 and Belgian/B.1.214 variants (neutralizing titre of 1.12 μg/mL). Additionally, it had almost no neutralizing activity on the Marseille-501/A.27 and Indian/B.1.617.2 variants (very low neutralizing titre of 3500 μg/mL). Moreover, the South African/B.1.351, Brazilian/B.18.104.22.168 and Marseille-484K.V1/R.1 variants were profoundly resistant to neutralization by bamlanivimab (LY-CoV555).
Molecular mechanisms of the neutralization escape of SARS-CoV-2 variants
Most neutralizing antibodies (nAbs) against SARS-CoV-2 are directed against the RBD and the NTD of the spike protein. As references, we used the LY-CoV 555 nAb (bamlanivimab) and the 4A8 nAb, which recognize the principal neutralization determinants of the RBD and the NTD, respectively.
The E484K substitution (GluàLys substitution) in the Marseille-484K.V1/R.1 variant induces a dramatic rearrangement of the RBD surface that results in a complete lack of interaction with the bamlanivimab nAb (Figure 4A and 4B). This molecular mechanism explains the dramatic decrease in the affinity (85%) of the bamlanivimab nAb for the RBD of the Marseille-484K.V1/R.1 variant (Table 1). A similar mechanism also accounts for all variants that display the E484K mutation, including the South African/B.1.351 (70% decrease) and Brazilian/B.22.214.171.124 (67% decrease) variants (Table 1).
The L452R substitution is present in both the Marseille-501/A.27 (Figure 4C) and the Indian/B.1.617.2 (Figure 4D) variants, yet in a distinct mutational context. In the case of the Marseille-501/A.27 variant, the L452R mutation is associated with N501Y. The loss affinity of Bamlanivimab’s nAb for this variant was estimated to be 76% (Table 1). The molecular mechanism of this effect could be attributed to a reorientation of the cationic side chain of R452 (compared to L452), which takes Y449 away from the antibody heavy chain residue N31 (Figure 5A and 5B).
The case of the Indian/B.1.617.2 variant is more puzzling since, in this case, the substitution L452R is associated with T478K instead of N501Y. As shown in Figure 6A, in the Original/B strain, T478 is close to F486, a key amino acid controlling bamlanivimab recognition. Indeed, the methyl group of T478 points in the direction of the aromatic ring of F486, which allows the formation of a cluster of π-π interactions with Y32 and Y92 of the light chain of the antibody. The clamp of Y32 and Y92 is particularly visible when the amino acid atoms are represented in spheres (Figure 6A, upper panel). When T478 is substituted by K478 (T478K substitution), F486 is attracted by the cationic group of K478, preventing any contact with the aromatic amino acids Y32 and Y92 of the antibody (Figure 6B). This mechanism largely contributes to the 72% loss of affinity of bamlanivimab for the RBD of the Indian/B.1.617.2 variants (Table 1).
Finally, we evaluated the impact of mutations in the NTD on antibody recognition (Table 1). Interestingly, some mutational patterns did not seem to decrease the affinity of the 4A8 nAb for the NTD, and in some cases, the affinity was even slightly increased, as shown for the Marseille-484K, V1/R1 and UK/B.1.1.7 variants (Table 1). In other cases, a significant decrease in the antibody affinity, compatible with the neutralization escape, was calculated and ranged from 64% for the Brazilian/B.126.96.36.199 variant to 47% for the South African/B.1.351 variant (Table 1). At the opposite end of the scale, the affinity of 4A8 for the UK/B.1.1.7 variant was slightly increased (-241 kJ.mol-1 vs. -225 kJ.mol-1 for the Original/B strain). A detailed analysis of the 4A8 epitope provided a molecular explanation for such a range of effects (Figure 7). This epitope is divided into two prominent and flexible regions of the NTD, the N3 and N5 loops, which adopt a crescent-like shape recognized by the antibody (Figure 7A). Key residues involved in 4A8 binding belong either to the N3 loop (K147, K150 and W152) or to the N5 loop (R246, Y248 and L249). Inasmuch as both loops are accessible at the NTD surface, the 4A8 antibody can bind to the NTD, as shown for the Origjnal/B strain (Figure 7A) and the UK/B.1.1.7 variant (Figure 7B). In the case of the South Africa/B.1.351 variant, the only part of the epitope preserved from this dramatic reorganization of the NTD is the tip of the N3 loop harbouring K147 and K150, which may explain the residual affinity of some anti-NTD nAbs (such as those elicited by vaccination) for this variant (Figure 7C). However, this truncated epitope may lose most of its immunogenicity. Thus, patients infected by the Original/B or the UK/B.1.1.7 strains may elicit nAbs against several variants, including the South African/B.1.351, but the reverse is not true, as sera from patients infected by the South African/B.1.351 variant have poor neutralizing activities. Subtle conformational changes in the NTD affecting the relative orientations of the K147 and R246 side chains were consistent with the slightly decreased affinity of anti-NTD nAbs for the Indian 2 variant vs. the Indian 1 variant (Figure 8). Since both variants have the same RBD but display distinct mutational patterns in the NTDs, these data underscored the importance of the NTD as a key neutralizing determinant of SARS-CoV-2.