Evolution of the SARS-CoV-2 Mutational Spectrum

Abstract

SARS-CoV-2 evolves rapidly in part because of its high mutation rate. Here, we examine whether this mutational process itself has changed during viral evolution. To do this, we quantify the relative rates of different types of single-nucleotide mutations at 4-fold degenerate sites in the viral genome across millions of human SARS-CoV-2 sequences. We find clear shifts in the relative rates of several types of mutations during SARS-CoV-2 evolution. The most striking trend is a roughly 2-fold decrease in the relative rate of G→T mutations in Omicron versus early clades, as was recently noted by Ruis et al. (2022. Mutational spectra distinguish SARS-CoV-2 replication niches. bioRxiv, doi:10.1101/2022.09.27.509649). There is also a decrease in the relative rate of C→T mutations in Delta, and other subtle changes in the mutation spectrum along the phylogeny. We speculate that these changes in the mutation spectrum could arise from viral mutations that affect genome replication, packaging, and antagonization of host innate-immune factors, although environmental factors could also play a role. Interestingly, the mutation spectrum of Omicron is more similar than that of earlier SARS-CoV-2 clades to the spectrum that shaped the long-term evolution of sarbecoviruses. Overall, our work shows that the mutation process is itself a dynamic variable during SARS-CoV-2 evolution and suggests that human SARS-CoV-2 may be trending toward a mutation spectrum more similar to that of other animal sarbecoviruses.

Keywords: SARS-CoV-2; equilibrium frequencies; mutation rate; mutational spectrum.

Fig. 1.

Mutation spectrum of SARS-CoV-2 clades at 4-fold degenerate sites. (A) Principal component analysis (PCA) of mutation spectra of Nextstrain clades with sufficient sequences (each point is a clade). (B) Fraction of mutations at 4-fold degenerate sites that are of each type for each clade. (C) Relative rates of each mutation type, calculated as the fraction of mutations of that type divided by the fraction of sites with the parental nucleotide. (D) Phylogenetic tree of clade founder sequences, with plots showing mutation rates for that clade (ordered as in C) minus rates for clade 20A. Interactive versions of A–C at https://jbloomlab.github.io/SARS2-mut-spectrum/ enable easier identification of individual clades. Supplementary figure S1, Supplementary Material online, shows that the PCA is robust to subsetting on sequences from different geographic locations, excluding top mutations, and partitioning the genome. The numerical values in (B) and (C) are at https://github.com/jbloomlab/SARS2-mut-spectrum/blob/main/results/synonymous_mut_rates/rates_by_clade.csv.

Fig. 2.

The changes in relative mutation rates correlate with the phylogenetic relationships among clades. The plots show the correlation between the square root of the phylogenetic distance separating each pair of clades and the Euclidean distance between the relative mutation rates for those clades. Assuming that mutation rates evolve neutrally according to a Brownian motion model, mutation rate distances should scale linearly with the square root of phylogenetic distance. The P-values are computed using the Mantel test with 100,000 permutations. The plots show that the mutation rates are significantly correlated (P < 0.05) with phylogenetic distance even if we exclude G→T or C→T mutations individually (although not together) or do the analysis only among Omicron or non-Omicron clades.

Fig. 3.

Predicted equilibrium frequencies from mutation rates versus actual nucleotide frequencies at 4-fold degenerate sites in sarbecoviruses. (A) Predicted equilibrium frequencies of nucleotides at 4-fold degenerate sites as calculated from the mutation rates for all of the SARS-CoV-2 clades analyzed here. (B) Actual frequencies of nucleotides at 4-fold degenerate sites in various sarbecoviruses. (C) Manhattan between predicted equilibrium frequencies (from mutation rates) and actual empirically observed nucleotide frequencies at 4-fold degenerate sites for a variety of viruses. SARS-CoV-2 clades are shown in purple squares. See supplementary figure S2, Supplementary Material online for per-nucleotide frequencies for the viruses shown in C.