Shifting mutational constraints in the SARS-CoV-2 receptor-binding domain during viral evolution

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has evolved variants with substitutions in the spike receptor-binding domain (RBD) that affect its affinity for angiotensin-converting enzyme 2 (ACE2) receptor and recognition by antibodies. These substitutions could also shape future evolution by modulating the effects of mutations at other sites-a phenomenon called epistasis. To investigate this possibility, we performed deep mutational scans to measure the effects on ACE2 binding of all single-amino acid mutations in the Wuhan-Hu-1, Alpha, Beta, Delta, and Eta variant RBDs. Some substitutions, most prominently Asn⁵⁰¹→Tyr (N501Y), cause epistatic shifts in the effects of mutations at other sites. These epistatic shifts shape subsequent evolutionary change-for example, enabling many of the antibody-escape substitutions in the Omicron RBD. These epistatic shifts occur despite high conservation of the overall RBD structure. Our data shed light on RBD sequence-function relationships and facilitate interpretation of ongoing SARS-CoV-2 evolution.

Fig. 1.. Deep mutational scanning maps of ACE2-binding affinity for all single amino acid mutations in five SARS-CoV-2 RBD variants.

The impact on ACE2 receptor-binding affinity (Δlog₁₀(K _D)) of every single amino-acid mutation in SARS-CoV-2 RBDs, as determined by high-throughput titration assays (fig. S1). The wildtype amino acid in each variant is indicated with an “x”, and gray squares indicate missing mutations in each library. An interactive version of this map is at https://jbloomlab.github.io/SARS-CoV-2-RBD_DMS_variants/RBD-heatmaps, and raw data are in data S1. The effects of mutations on RBD surface expression are in fig. S2.

Fig. 2.. Epistatic shifts in mutational effects across RBD variants.

(A) The shift in mutational effects on ACE2 binding at each RBD site between the indicated variant and Wuhan-Hu-1. An interactive version of this plot is at https://jbloomlab.github.io/SARS-CoV-2-RBD_DMS_variants/epistatic-shifts. The epistatic shift is calculated as the Jensen-Shannon divergence in the set of Boltzmann-weighted affinities for all amino acids at each site. Gray shading indicates sites of strong antibody escape based on prior deep mutational scanning of the Wuhan-Hu-1 RBD ( 11 ). (B) Ribbon diagram of the Wuhan-Hu-1 RBD structure (PDB 6M0J) colored according to epistatic shifts. Labeled spheres indicate residues that are mutated in each RBD variant. (C) Mutation-level plots of epistatic shifts at sites of interest. Each scatter plot shows the measured affinity of all 20 amino acids in the Beta versus Wuhan-Hu-1 RBD. Red dashed lines mark the parental RBD affinities, and the gray dashed line indicates the additive (non-epistatic) expectation. Epistatic shifts can reflect idiosyncratic mutation-specific shifts (e.g., site 498) or global changes in mutational sensitivity at a site (e.g., site 449). Site 484 does not have a substantial epistatic shift and is shown for comparison. See fig. S3 for scatterplots of additional sites of interest. See fig. S4 for epistatic shifts in mutational effects on RBD expression.

Fig. 3.. Functional and evolutionary relevance of epistatic interactions.

(A) Double mutant cycle diagram illustrating the positive epistasis interaction between N501Y and Q498R. Asterisk indicates expected double-mutant binding affinity assuming additivity. (B) Affinity-buffering of Omicron BA.1 mutations. Each diagram shows the cumulative addition of individually measured effects on ACE2-binding affinity (Δlog₁₀(K _D)) for each single RBD substitution in Omicron BA.1 as measured in the Wuhan-Hu-1 (left) or Beta (right) RBDs. Mutation effect is calculated in the labeled direction even when the reference state in a background differs, e.g., N501Y in the Beta background is the opposite-sign effect of the measured Y501N mutation. Red line marks the Wuhan-Hu-1 affinity, and asterisk the actual affinity of the Omicron BA.1 RBD relative to Wuhan-Hu-1 as measured in ( 12 ). See also fig. S5, A to C. (C) Efficiency of entry of Omicron BA.1 (or reversion mutant) spike-pseudotyped lentivirus on a HEK-293T cell line expressing low levels of ACE2 (fig. S6, A and B). Labels indicate fold-decrease in geometric mean (red bar) of biological triplicate measurements. (D) Double mutant cycle illustrating positive epistasis between N501Y and Y449H. (E) Impact of epistasis on SARS-CoV-2 sequence evolution. Plot illustrates the change in a mutation’s effect between Alpha (N501Y) versus Wuhan-Hu-1 deep mutational scanning data, versus the ratio in number of observed occurrences of the substitution in genomes containing N501 versus Y501 in a global SARS-CoV-2 phylogeny as of 25 May, 2022 ( 22 ). Note that we are counting substitution occurrence as an event on the phylogeny independent of the number of offspring of a node, and not the raw number of sequenced genomes with which a mutation is observed. A pseudocount was added to all substitution counts to enable ratio comparisons, and substitutions that were observed <2 times in total are excluded. Color scale reinforces the ΔΔlog₁₀(K _D) metric on the y-axis. Labeled mutations are those with |ΔΔlog₁₀(K _D)| > 0.9. Vertical line at x ~ 0.6 marks equal relative occurrence on Y501 versus N501 genomes given the larger number of substitutions that had been observed on N501 genomes.

Fig. 4.. Epistatic shifts are not accompanied by large structural perturbations.

(A) Global alignment of the Wuhan-Hu-1 (PDB 6M0J) and Beta (PDB 7EKG) RBD backbones. Key sites are labeled. (B) Correlation between the extent of epistatic shift in mutational effects at a site and its structural perturbation in Beta versus Wuhan-Hu-1 RBDs (backbone Cα or all-atom average displacement from aligned X-ray crystal structures). See figs. S7 and S8 for additional details. (C) Molecular dynamics simulation of RBD variants bound to ACE2. Volumetric maps (top) show the 3D space occupied by key residues over the course of simulation. Cartoon diagrams (bottom) illustrate the fraction of simulation frames in which a salt bridge (black arrow) or polar or nonpolar (gray arrow) contact is formed between residue pairs (fig. S9C). See fig. S9A for equivalent diagrams for Omicron+Y501N (R498/N501 for comparison with Wuhan-Hu-1+Q498R) and fig. S9B for apo ACE2. See fig. S9C for histograms of contact distances over the course of the simulations.