Generation of SARS-CoV-2 escape mutations by monoclonal antibody therapy

Abstract

COVID-19 patients at risk of severe disease may be treated with neutralising monoclonal antibodies (mAbs). To minimise virus escape from neutralisation these are administered as combinations e.g. casirivimab+imdevimab or, for antibodies targeting relatively conserved regions, individually e.g. sotrovimab. Unprecedented genomic surveillance of SARS-CoV-2 in the UK has enabled a genome-first approach to detect emerging drug resistance in Delta and Omicron cases treated with casirivimab+imdevimab and sotrovimab respectively. Mutations occur within the antibody epitopes and for casirivimab+imdevimab multiple mutations are present on contiguous raw reads, simultaneously affecting both components. Using surface plasmon resonance and pseudoviral neutralisation assays we demonstrate these mutations reduce or completely abrogate antibody affinity and neutralising activity, suggesting they are driven by immune evasion. In addition, we show that some mutations also reduce the neutralising activity of vaccine-induced serum.

Fig. 1. P values for differences in spike amino-acid frequencies between pre- and post-treatment sequences.

a Patients infected with Delta and treated with casirivimab/imdevimab (n = 1557 pre-treatment and n = 67 post-treatment), b patients infected with BA.1 and treated with sotrovimab (n = 3221 pre-treatment and n = 148 post-treatment), and c patients infected with BA.2 and treated with sotrovimab (n = 1338 pre-treatment and n = 25 post-treatment). Amino acid frequencies were compared between pre-and post-treatment samples (at least 10 days after treatment) at each site in the spike sequence alignment. P values for each site were calculated using a one-sided Fisher’s test, and p values were log-transformed and inversed for visualisation so that sites with diverging values appear higher up on the figure. Only sites with some variability (>1 amino acid) are shown. The horizontal lines indicate p value thresholds of p < 0.001, p < 0.0001 etc. Residues with diverging frequencies (p < 0.001) are highlighted in red, with the observed amino acid change indicated in text. Residues known to interact with each drug are indicated in blue and purple at the top of the figure. Nine sites are highlighted in red in the figure: E406D/Q (p = 9 × 10⁻⁴), G446S/V (p < 10⁻¹⁶), Y453F (p = 0.000809) and L455F/S (p = 9 × 10⁻⁶) in patients infected with Delta and treated with casirivimab+imdevimab; P337R/S (p < 10⁻¹⁶) and E340A/D/K/V (p < 10⁻¹⁶), K356T (p < 10⁻¹⁶) and R493Q (p = 1.8 × 10⁻⁵) in patients infected with BA.1 and treated with sotrovimab; and E340K (p = 0.000369) in patients infected with BA.2 and treated with sotrovimab. See also Figure S1. No adjustments were made for multiple comparisons.

Fig. 2. Structural modelling of mutations mapped to the spike RBD.

a Model of the Omicron RBD (7TLY) docked with S309 (sotrovimab). Omicron RBD is shown as a grey surface from an approximate front view, S309 as cartoon ribbons with heavy and light chains coloured separately. Mutation sites mapped to the RBD surface are coloured magenta and labelled. b Close-up view of the interface between the P337, E340, K356 patch of residues with the S309 heavy chain. Potential hydrogen bonds and hydrophobic interactions are shown as green dashed lines. c Model of the Delta RBD docked with REGEN-COV mAbs casirivimab and imdevimab shown from approximate front (left) and back (right) views. Delta RBD is shown as a grey surface and mutation sites E406, G446, Y453 and L455 are coloured magenta and labelled. d Close-up view of the interface between E406, Y453 and L455 with casirivimab. e Close-up view of the interface between G446 with imdevimab. Potential hydrogen bonds and hydrophobic interactions are shown as green dashed lines.

Fig. 3. Neutralisation escape caused by RBD mutations.

A Pseudoviral neutralisation curves of the indicated Delta variants with REGEN-COV mAbs. Comparison is made with Omi-12 A VH1-58 mAb, which is not sensitive to the mutations found following REGEN-COV treatment. Data are presented as mean values ± SEM. n = 2 biologically independent experiments. B Pseudovirus neutralisation curves for BA.1 sotrovimab mutants. Data are presented as mean values ± SEM. n = 2 biologically independent experiments. C, D Neutralisation mean IC50 ± SEM titres for neutralisations shown in A, B.

Fig. 4. Pseudoviral neutralisation IC50 titres of third dose Pfizer-BioNTech vaccine serum.

IC50 reciprocal plasma dilution titres for Delta REGEN-COV induced mutations and BA.1 sotrovimab induced mutations are compared with titres for ancestral strain Victoria, Delta and BA.1. Geometric mean of titres are shown above each column. The Wilcoxon matched-pairs signed rank test was used for the analysis and two-tailed P values were calculated. Different colours indicate different pseudovirus, which is indicated underneath the graph.