Genetic consequences of effective and suboptimal dosing with mutagenic drugs in a hamster model of SARS-CoV-2 infection

Abstract

Mutagenic antiviral drugs have shown promise against multiple viruses, but concerns have been raised about whether their use might promote the emergence of new and harmful viral variants. Recently, genetic signatures associated with molnupiravir use have been identified in the global SARS-COV-2 population. Here, we examine the consequences of using favipiravir and molnupiravir to treat SARS-CoV-2 infection in a hamster model, comparing viral genome sequence data collected from (1) untreated hamsters, and (2) from hamsters receiving effective and suboptimal doses of treatment. We identify a broadly linear relationship between drug dose and the extent of variation in treated viral populations, with a high proportion of this variation being composed of variants at frequencies of less than 1 per cent, below typical thresholds for variant calling. Treatment with an effective dose of antiviral drug was associated with a gain of between 7 and 10 variants per viral genome relative to drug-free controls: even after a short period of treatment a population founded by a transmitted virus could contain multiple sequence differences to that of the original host. Treatment with a suboptimal dose of drug showed intermediate gains of variants. No dose-dependent signal was identified in the numbers of single-nucleotide variants reaching frequencies in excess of 5 per cent. We did not find evidence to support the emergence of drug resistance or of novel immune phenotypes. Our study suggests that where onward transmission occurs, a short period of treatment with mutagenic drugs may be sufficient to generate a significant increase in the number of viral variants transmitted.

Keywords: SARS-CoV-2; favipiravir; molnupiravir; mutagenesis.

Figure 1.

Statistics of virus diversity in treated and untreated populations. (A) Increases in the mean number of variants per viral genome were observed in populations treated with increasing doses of favipiravir. Black dots show genome-wide statistics calculated from populations: Each dot represents the viral population in a single hamster. The red square shows the mean value for populations treated with 300 mg favipiravir and 150 mg molnupiravir. The blue square at 500 mg molnupiravir shows a datapoint that was omitted from the regression calculation; this point is discussed in the main text. Shading indicates a 95 per cent confidence interval for the regression line. A horizontal dashed line indicates the model threshold at the point of effective treatment. (B) Increases in the mean number of variants per viral genome per virus were observed in populations treated with increasing doses of molnupiravir. (C) Composition of variants denoted by variant frequency, ordered by favipiraivir dose. Bars show the total number of variants per genome (plotted as points in Fig. 1A), decomposed into the contributions to this total from variants of different frequencies. Each bar describes statistics for the viral population in a single hamster. (D) Composition of variants denoted by variant frequency, ordered by molnupiravir dose. Bars show the total number of variants per genome (plotted as points in Fig. 1A), decomposed into the contributions to this total from variants of different frequencies.

Figure 2.

Mutational spectra of low-frequency variants. (A) The proportion of low-frequency variation (variant frequency <1 per cent) that was comprised of C to U mutations was higher for treated populations than for untreated populations. Squares show treatments for which a single data point was collected. (B) The proportion of low-frequency variation that was comprised of G to A mutations was higher for treated populations than for untreated populations.

Figure 3.

Genomic locations of variants in treated viral populations. The locations of variants which reached a frequency of at least 5 per cent in the viral population are shown in red (non-synonymous variant), blue (synonymous variant) or black (nonsense variant). With the exception of variants transmitted through standing variation, very little replication of variants between treated populations was observed.

Figure 4.

πN/πS was lower in untreated than in treated populations. Error bars were calculated using a likelihood model.

Figure 5.

Variants observed in RdRp at frequencies of 5 per cent or greater. Sites reported to convey resistance to remdesivir are shown in red vdW representation (Szemiel et al. 2021; Gandhi et al. 2022; Stevens et al. 2022). Other colours show variants observed in hamster populations at frequencies of 5 per cent or more. Image created in VMD (Humphrey, Dalke, and Schulten 1996) based on the PDB structure 6m71 (Gao et al. 2020).