SARS-CoV-2 evolution in animals suggests mechanisms for rapid variant selection

Abstract

SARS-CoV-2 spillback from humans into domestic and wild animals has been well documented, and an accumulating number of studies illustrate that human-to-animal transmission is widespread in cats, mink, deer, and other species. Experimental inoculations of cats, mink, and ferrets have perpetuated transmission cycles. We sequenced full genomes of Vero cell-expanded SARS-CoV-2 inoculum and viruses recovered from cats (n = 6), dogs (n = 3), hamsters (n = 3), and a ferret (n = 1) following experimental exposure. Five nonsynonymous changes relative to the USA-WA1/2020 prototype strain were near fixation in the stock used for inoculation but had reverted to wild-type sequences at these sites in dogs, cats, and hamsters within 1- to 3-d postexposure. A total of 14 emergent variants (six in nonstructural genes, six in spike, and one each in orf8 and nucleocapsid) were detected in viruses recovered from animals. This included substitutions in spike residues H69, N501, and D614, which also vary in human lineages of concern. Even though a live virus was not cultured from dogs, substitutions in replicase genes were detected in amplified sequences. The rapid selection of SARS-CoV-2 variants in vitro and in vivo reveals residues with functional significance during host switching. These observations also illustrate the potential for spillback from animal hosts to accelerate the evolution of new viral lineages, findings of particular concern for dogs and cats living in households with COVID-19 patients. More generally, this glimpse into viral host switching reveals the unrealized rapidity and plasticity of viral evolution in experimental animal model systems.

Keywords: SARS-CoV-2; companion animals; host adaptation; spillover; viral variants.

Fig. 1.

A total of 88 unique variants were detected in ≥3% of sequences of in vitro– and in vivo–derived SARS-CoV-2. The position, predicted effect (A), and allele frequency (B) of SNV and SV variants detected across the SARS-CoV-2 genome in sequences obtained from 13 experimentally inoculated animals and three passages of the viral inoculum. Each point represents an SNV (circle) or SV (triangle). (A) All variants detected in ≥3% of sequences, demonstrating a majority of SNVs and a slightly increased occurrence of modifications in the spike protein. (B) All variants detected in ≥ 25% of sequences, revealing the presence of higher-frequency variants in the spike protein of all datasets, excluding the P1 stock virus, and the prevalence of higher-frequency variants across the entire genome of SARS-CoV-2 recovered from dogs. The frequency is indicated by scale from blue (low) to red (high). The schematic of the SARS-CoV-2 genome is illustrated for orientation.

Fig. 2.

SARS-CoV-2 cell culture variants revert rapidly during in vivo experimental infection. SARS-CoV-2 isolate USA-WA1/2020 was passaged three times in Vero E6 cell line. Five SNV substitutions across the genome reached >93% frequency; the variant proportion recovered from each supernatant stock is indicated by P1, P2, and P3. Cats, dogs, hamsters, and the ferret (n = 6, 3, 3, and 1, respectively) were inoculated with 10⁵ to 10⁶ pfu intranasally. Virus was recovered 1 to 3 d postinoculation, sequenced using a tiled amplicon technique and analyzed with a pipeline for calling SNVs and SVs in viral populations. Cell culture variants decreased in frequency in all animals with the exception of the ferret (n = 1). Variants are indicated in the reference to the consensus residue in USA-WA1/2020 and their position within the coding sequence of a SARS-CoV-2 protein. Each point represents the mean of two technical replicates, aside from one cat, for which a replicate was not sequenced.

Fig. 3.

SARS-CoV-2 viral evolution differs across species, gene regions, and individuals. (A) Each point indicates the number of unique variants detected at ≥3% frequency in SARS-CoV-2 genomes recovered from individual animals. There is no significant difference in the number of unique variants detected in different species (ANOVA, P = 0.22). (B) Analysis of variant distribution within species reveals that the majority of variants were detected in just one individual within each species. Subplot a shows the distribution for all variants, while b illustrates only variants not occurring in the P3 inoculum at ≥3%. (C) Variants are distributed across viral genes in relation to each gene’s length as a proportion of the entire genome length (linear model, R² = 0.69, and P < 0.0001). The spike protein contained a notably higher proportion of all variants in comparison to its share of genome length. Gray shading represents the 95% CI for the slope of the regression line. (D) SARS-CoV-2 spike protein variant “residues of concern” are in NTD, RBD, and furin cleavage site. Residues described in the main text and Table 1 in the SARS-CoV-2 spike trimer are highlighted on structure 6VXX. Blue indicates NTD and yellow indicates RBD. The furin cleavage site and adjacent residue 686 are in the indicated loop, which was not resolved in this structure.

Fig. 4.

Signatures of positive selection are detected in SARS-CoV-2 genome sequences recovered from experimentally inoculated cats, dogs, hamsters, and the ferret. (A) Comparison of nonsynonymous (piN) and synonymous (piS) nucleotide diversity reveals that piN is significantly greater than piS, indicating positive selection. Each point represents a measurement of piN (red) or piS (blue) calculated for the entire SARS-CoV-2 genome from sequences recovered from an individual animal, relative to the reference sequence of USA-WA1/2020. The analysis of the same measures within each species reveals that piN > piS for viral genomes is greater in dogs and hamsters. (B) Orf1ab, S, and M are undergoing positive selection in mammalian hosts. Each point represents piN or piS calculated for a specific SARS-CoV-2 gene or open reading frame from sequences recovered from an individual animal.