Isolation and characterization of spike-specific SARS-CoV-2 HuMabs.
In previous work, we screened plasma samples obtained early in the pandemic (March 2020) from 165 convalescent Danish healthcare workers for SARS-CoV-2 spike-specific IgG and spike-RBD/ACE2 binding inhibition (28). Through this analysis we selected one subject (K501) who had spike-specific IgG, RBD-specific IgG and spike-RBD/ACE2 inhibition (Figure S2). Based on these data, PBMC’s were collected at the September 2020 time-point with the aim of isolating spike-specific HuMabs.
A total of seven SARS-CoV-2 spike-specific HuMabs were isolated using EBV immortalization and their binding to the target antigen was confirmed by ELISA (Fig. 1A) (30). To characterize the ability of the anti-spike HuMabs to block ACE2 binding to the spike protein, the HuMabs were screened for in vitro inhibition and IC50 was determined (Fig. 1B and Fig. 1C). Four of the seven HuMabs (K501SP2, K501SP3, K501SP4, and K501SP5) competed with ACE2 for binding. No spike/ACE2 inhibition was observed with HuMabs K501SP1, K501SP6 and K501SP7.
Sequence analysis of the seven HuMabs revealed a polyclonal response with all HuMabs being derived from different germlines (Fig. 1D and Table S2). Four HuMabs had kappa light chains (K501SP1, K501SP2, K501SP3, and K501SP5) whereas three HuMabs had lambda light chains (K501SP4, K501SP6 and K501SP7) (Table S1). Similar to previous studies (51), the HuMabs isolated here had relatively low levels of somatic mutation which ranged between 92.4% and 96.9% identity to germline with CDRH3 length that ranged from 11 to 23 a.a. (Table S1). For the light chains, somatic mutation ranged from 94.1–98.6%.
RBD-specific HuMabs have cross variant binding and inhibition.
The majority of the protective antibodies to SARS-CoV-2 have been shown to target the RBD and inhibit ACE2 binding. Therefore, RBD-specific binding ELISAs and ACE2 competition assays were performed on the WT variant of the virus to assess whether our spike-specific HuMabs bound to and inhibited the RBD from binding to ACE2. The same four HuMabs that inhibited spike/ACE2 binding (K501SP2, K501SP3, K501SP4 and K501SP5) bound to WT RBD (Fig. 2A) and inhibited its binding to ACE2 (Fig. 2B). No WT RBD binding or ACE2 binding-inhibition was observed with HuMabs K501SP1, K501SP6 and K501SP7, indicating that these HuMabs are not RBD specific and bind elsewhere on the spike protein.
Next, to assess the variant transcending capacity of the RBD-specific HuMabs, RBD-specific binding ELISAs and ACE2 competition assays were performed on Alpha, Beta, Gamma and Omicron (BA.1) variants. All four HuMabs bound to and competed with the Alpha variant, however, inhibition was reduced when compared to WT RBD (Fig. 2A and Fig. 2B). Only HuMabs K501SP3 and K501SP4 bound to the Beta, Gamma and Omicron variants, and competed with the ACE2 binding (Fig. 2A, Fig. 2B and Figure S3), but with a reduced binding capacity compared to the WT (Fig. 2C). Overall, HuMab K501SP3 had the broadest and most potent inhibition capacity when analyzed in our RBD/ACE2 inhibition assay and therefore we measured binding affinity for this HuMab.
To assess the binding affinity of K501SP3, HuMab/RBD-binding was tested in a concentration titration series using a QCMbiosensor (Fig. 2D and Figure S4) for the WT and the four VoCs. HuMab K501SP2 was also tested as a binding control. The results confirmed the binding and inhibition data above and showed that K501SP3 bound to all the VoCs tested in nM range, however with decreased affinity when compared to the WT where affinity was in the pM range (Fig. 2D and Figure S4). As expected, HuMab K501SP2 bound to the WT and Alpha variant only.
Taken together, our data report the discovery of a highly potent RBD-specific HuMab with the ability to bind and inhibit four VoCs. Furthermore, we find three spike-specific HuMabs (K501SP1, K501SP6 and K501SP7) that are unable to bind to RBD or inhibit RBD/ACE2 binding.
Spike-specific HuMabs have cross variant and sub-lineage neutralization.
Encouraged by the RBD/ACE2 inhibition breadth and potency results presented above, and to determine whether both our non-RBD and RBD-specific HuMabs had broad and potent virus neutralization capacity, we next examined breadth of neutralization for each of the seven HuMabs. Neutralization was performed against eight SARS-CoV-2 isolates representing the WT D614G, Alpha, Delta and five of the Omicron sub-lineages (BA.1, BA.2, BA.5, BQ.1.1 and XBB.1), two of which (BQ.1.1 and XBB.1) became dominant while undertaking this study. The analysis showed that HuMab K501SP6 had the broadest neutralizing capacity, since it neutralized all isolates tested including the more recent Omicron BQ.1.1 and XBB.1 isolates. These isolates could not to be neutralized by any of the other seven HuMabs or the emergency use authorized (EUA) control Bebtelovimab at < 10 µg/mL (Fig. 3, Table S3). Nevertheless, the IC50 value of K501SP6 was much higher than the other neutralizing HuMabs, with an average IC50 of 17.4 µg/mL across all variants tested (Table S3). HuMabs K501SP1 and K501SP7 did not neutralize any of the isolates at the dilutions tested. HuMabs K501SP2, K501SP3, K501SP4 and K501SP5 could effectively neutralize the WT D614G, Alpha and Delta isolates tested (Fig. 3, Table S3). However, HuMabs K501SP2 and K501SP5 were not found to neutralize any of the Omicron sub-lineage isolates at 100 µg/ml. K501SP4 was found to neutralize BA.2, BA.5, BQ.1.1 and XBB.1 Omicron isolates, albeit with an IC50 between 25–50 µg/ml and thus were not very potent. K501SP3 could effectively neutralize all the isolates tested with the exception of the BA.5, BQ.1.1 and XBB.1 isolates, which were not observed to be neutralized at 100 µg/ml.
Overall, HuMab K501SP3 was found to be potent, supporting results observed above using the RBD/ACE2 inhibition assay, however this antibody failed to neutralize newer circulating variants. In conclusion, non-RBD specific HuMab K501SP6 was the broadest HuMab, inhibiting all isolates tested including the new and currently circulating BQ.1.1 and XBB.1 variants.
Investigation into BA.5 resistance to neutralization by HuMab K501SP3
Given the potency of neutralization observed towards earlier circulating variants by HuMab K501SP3, further investigation was conducted into BA.5 neutralization resistance. Firstly, to confirm that the neutralization observed was directed to the RBD, we performed neutralization assays using the panel of mutant viruses exhibiting the RBD of either the Alpha, Delta and Omicron BA.1, BA.2 or BA.5 variants. Neutralization by K501SP3 of these viruses showed IC50 comparable with the respective variant culture isolates, and no neutralization of the BA.5 recombinant was observed, validating that substitutions in the BA.5 RBD were responsible for preventing neutralization by this HuMab (Fig. 4A and Table S4). Compared to the Alpha, Delta, BA.1 and BA.2 spike proteins, the BA.5 spike protein contains substitution F486V. To investigate if this specific residue was responsible for abolishing the neutralization from K501SP3, we performed neutralizations with four additional mutant viruses containing E484A (which is present in all Omicron sub-lineages and is associated with escape from mAbs (52)), F486V, E484A and F486V or the complete BA.5 spike without F486V (reversion to V486F). Neutralization of K501SP3 to these viruses showed that, compared to the IC50 observed against the WT D614G isolate, the E484A recombinant had a 2.5-fold higher IC50, the F486V recombinant had a 7-fold higher IC50 and the E484A + F486V recombinant had a 44-fold higher IC50 (Fig. 4B and Table S4). For the BA.5 recombinant with a V486F reversion, neutralization was now possible with an IC50 comparable to the WT D614G isolate (Fig. 4B and Table S4).
Next, to understand the structural basis of both the potent neutralization, and resistance to BA.5 observed in HuMab K501SP3, we resolved the Cryo-EM structure of SARS-CoV-2 Omicron BA.1 spike trimer (Fig. 4C) in complex with HuMab K501SP3 (Fig. 4D). In the trimeric structure, the RBDs of subunit 1 and 2 were observed to be in ‘up’ conformation with each binding to a Fab region of HuMab K501SP3. The RBD of subunit 3 was observed to be in ‘down’ conformation (Fig. 4D). Although the resolution was insufficient to map the K501SP3 contact point, map density in that region shows that the Fab regions of K501SP3 are bound on top of the RBDs particularly in the receptor binding motif (RBM) (438–506 a.a.) which mediates binding of the spike protein to human ACE2.
Together, these observations are consistent in showing that K501SP3 is bound to the RBM and that the mutations F486V and E484A are essential for neutralization resistance, with F486V having a major impact. Furthermore, these properties are suggestive of a Barnes classification class 1 binding antibody (53).
Genetic analysis of HuMabs K501SP3 and K501SP6
To further characterize our broadest (K501SP6) and most potent (K501SP3) HuMabs, we investigated whether the genetic characteristics of each of these HuMabs were unique. We used the coronavirus antibody database (CoV-AbDab) to compare the sequences of our HuMabs to 12036 entries of antibodies known to bind SARS-CoV-2 (38). Sequence analysis of HuMab K501SP3 revealed that it was derived from an IGHV1-58 heavy chain gene and an IGKV3-20 light chain with a CDRH3 region of 16 a.a. in length (Table S2). Whereas HuMab K501SP6 was derived from an IGHV4-39 heavy chain, an IGLV1-40 light chain and a CRHR3 region of 21 a.a. Of the 12036 SARS-CoV-2 antibody sequences banked in the CoV-AbDab database, 105 use the K501SP3 combination of IGHV1-58 and IGKV3-20. Further analysis revealed that 81 antibodies had identical IGHV, IGHJ, IGLV, IGLJ gene usage and CDRH3 length to K501SP3 (Fig. 5A and Table S5). The closest antibody (P008-081 (54)) shared 86% a.a. sequence identity in their heavy and light chains.
For HuMab K501SP6, a total of 45 antibodies that used the combination of IGHV4-39 and IGLV1-40 were identified (38). We were not able to identify any antibody with identical IGHV, IGHJ, IGLV, IGLJ gene usage and CDRH3 length to HuMab K501SP6. On further examination two antibodies with the same IGHV, IGLV and CDRH3 length as HuMab K501SP6, but alternate IGHJ and IGLJ usage, were identified. Another six antibodies with identical IGHV, IGHJ, IGLV, IGLJ gene usage, but different CDRH3 length, were also found (Fig. 5B and Table S5).
Comparison of HuMab K501SP3 with the emergency use approved therapeutic mAbs revealed that HuMab K501SP3 had the same combination of IGHV1-58 and IGKV3-20 as well as CDRH3 region length as Tixagevimab (COV2-2196) (Fig. 5C) (55, 56). The sequence identity to Tixagevimab for the heavy chain was 78.2% and for the light chain was 89.2%. No FDA approved therapeutic mAbs were found to have the same heavy and light chain genes as HuMab K501SP6.
Due to the breadth of neutralization and genetic uniqueness of HuMab K501SP6 we used the CoV-AbDab to search for HuMabs that target the non-RBD area of the spike and neutralize the Omicron BQ.1.1 and XBB.1.4 variants. We found only one other HuMab, C1717 with these characteristics. This mAb targets the N-terminal domain (NTD)-SD2 and utilizes IGHV1-69 and IGLV1-44 (19, 57).
Overall, limited antibodies with similar genetic characteristics to HuMab K501SP6 were observed, indicating genetic uniqueness. Furthermore, we found that HuMab K501SP3 gene usage and CDRH3 length are common for RBD specific HuMabs.