Synonymous mutations reduce genome compactness in icosahedral ssRNA viruses

Abstract

Recent studies have shown that single-stranded (ss) viral RNAs fold into more compact structures than random RNA sequences with similar chemical composition and identical length. Based on this comparison, it has been suggested that wild-type viral RNA may have evolved to be atypically compact so as to aid its encapsidation and assist the viral assembly process. To further explore the compactness selection hypothesis, we systematically compare the predicted sizes of >100 wild-type viral sequences with those of their mutants, which are evolved in silico and subject to a number of known evolutionary constraints. In particular, we enforce mutation synonynimity, preserve the codon-bias, and leave untranslated regions intact. It is found that progressive accumulation of these restricted mutations still suffices to completely erase the characteristic compactness imprint of the viral RNA genomes, making them in this respect physically indistinguishable from randomly shuffled RNAs. This shows that maintaining the physical compactness of the genome is indeed a primary factor among ssRNA viruses' evolutionary constraints, contributing also to the evidence that synonymous mutations in viral ssRNA genomes are not strictly neutral.

Figure 1

(a) Example of a typical fold of the entire brome mosaic virus (BMV) RNA2 sequence. The maximum ladder distance (MLD) of the folded sequence is highlighted. (b) Thermally averaged MLD, 〈MLD〉, of the WT BMV RNA2 sequence (blue line) and the distribution of 〈MLD〉 values obtained for random RNA sequences of same length and composition as the WT sequence. (c) 〈MLD〉 value of viral ssRNA sequences versus the sequence length N (in nucleotides). Different virus families are represented by different colors and symbols. (Red solid line) Power law of Eq. 3 for the expected values of 〈MLD〉 for random RNA sequences, constrained only by their overall viral-like nucleotide composition. Due to their atypical nucleotide composition, Tymoviridae are not represented by Eq. 3, and the corresponding scaling law for Tymoviridae-like random RNA sequences, ${\overline{\langle \text{MLD}\rangle }}_{Ty}\left(N\right)=\left(0.92\pm 0.44\right)\times N^{0.669\pm 0.054}$, is shown (orange dashed line). See the Supporting Material for further information. (To see this figure in color, go online.)

Figure 2

(a) Influence of synonymous point mutations on MLD. (Gray circles) The 〈MLD〉 values of WT viral sequences from Fig. 1b; (blue triangles) $\overline{\langle \text{MLD}\rangle }$ values of synonymously mutated sequences. Scaling laws for $\overline{\langle \text{MLD}\rangle }$ values of random RNA sequences with viral-like and Tymoviridae-like composition are shown as in Fig. 1. (b) The average degree of sequence identity between the mutated and WT sequences. (Gray-shaded area) Values one would expect if only one in three nucleotides were allowed to mutate in the coding regions of the genomes. Note that Tymoviridae genomes (green) are more conserved than the others. This is due to the presence of overlapping reading frames covering, on average, 30% of their genome. (To see this figure in color, go online.)

Figure 3

Mutation dynamics trajectories for four viral ssRNA sequences. (Top to bottom) BMV RNA2 and RNA1 segments from the tripartite genome of BMV (Bromoviridae), OnYMV (Tymoviridae), and ERBV1 (Picornaviridae). Each panel shows nine 〈MLD〉 trajectories and their average value (blue) for each sequence in units of MC steps, N/100. (Red dot-dashed lines and green dashed lines) 〈MLD〉 values of WT RNAs and the $\overline{\langle \text{MLD}\rangle }$ values of random RNAs (for viral-like composition, Eq. 3), respectively. Note that in the case of OnYMV, a Tymovirus, we must consider the appropriate asymptotic value of $\overline{\langle \text{MLD}\rangle }$ for random RNAs with Tymoviridae-like composition (see Fig. 1). This value is shown in the figure (orange short-dashed line). (To see this figure in color, go online.)

Figure 4

Color-coded heat maps for the probability density of finding mutated sequences with given 〈MLD〉 and sequence identity with the WT sequence. The probability density for each virus is computed and normalized over the whole length of the nine mutation trajectories (1500 MC steps) shown in Fig. 3. (Red dot-dashed lines and green dashed lines) 〈MLD〉 values of WT RNA and the $\overline{\langle \text{MLD}\rangle }$ values of random RNAs (with viral-like composition, Eq. 3), respectively. (Orange short-dashed line) In the OnYMV case, the random $\overline{\langle \text{MLD}\rangle }$ value for Tymoviridae-like composition is shown. (To see this figure in color, go online.)

Figure 5

The $\overline{\langle \text{MLD}\rangle }$ values for the synonymous constraint only (upward triangles), and for the additional constraints of preserving UTR sequences (downward triangles) and UTR sequences with codon biases (squares). The $\overline{\langle \text{MLD}\rangle }$ values for these last two cases are evaluated over a set of 150 mutated sequences for each virus. Data are presented in the same manner as in Fig. 2 (see also Fig. S3 for UTRs preserving synonymous point mutations of Tymoviridae). (To see this figure in color, go online.)