Positive Selection or Free to Vary? Assessing the Functional Significance of Sequence Change Using Molecular Dynamics

Abstract

Evolutionary arms races between pathogens and their hosts may be manifested as selection for rapid evolutionary change of key genes, and are sometimes detectable through sequence-level analyses. In the case of protein-coding genes, such analyses frequently predict that specific codons are under positive selection. However, detecting positive selection can be non-trivial, and false positive predictions are a common concern in such analyses. It is therefore helpful to place such predictions within a structural and functional context. Here, we focus on the p19 protein from tombusviruses. P19 is a homodimer that sequesters siRNAs, thereby preventing the host RNAi machinery from shutting down viral infection. Sequence analysis of the p19 gene is complicated by the fact that it is constrained at the sequence level by overprinting of a viral movement protein gene. Using homology modeling, in silico mutation and molecular dynamics simulations, we assess how non-synonymous changes to two residues involved in forming the dimer interface-one invariant, and one predicted to be under positive selection-impact molecular function. Interestingly, we find that both observed variation and potential variation (where a non-synonymous change to p19 would be synonymous for the overprinted movement protein) does not significantly impact protein structure or RNA binding. Consequently, while several methods identify residues at the dimer interface as being under positive selection, MD results suggest they are functionally indistinguishable from a site that is free to vary. Our analyses serve as a caveat to using sequence-level analyses in isolation to detect and assess positive selection, and emphasize the importance of also accounting for how non-synonymous changes impact structure and function.

Fig 1. Tombusvirus genome structure and p19 crystal structure.

(A) Schematic diagram showing the arrangement of the tombusvirus genome, including the overprinting of p19 on the movement protein (MP). (B) Crystal structure of the tomato bushy stunt (TBS) virus p19 dimer (PDB ID 1R9F) [24] bound to a 19 bp RNA fragment. The protein subunits are drawn in cartoon style and coloured dark and light grey; the two RNA strands are drawn as van der Waals spheres and coloured red and orange. Residues found to be under positive selection in at least two analyses are coloured cyan (cartoon representation), and residues identified as being under positive selection by all four analyses are drawn explicitly and coloured according to atom type (cyan: carbon; red: oxygen; blue: nitrogen; white: hydrogen). The dashed square indicates the region shown in panel C. (C) Close up view of the dimer interface with residues Arg139 and Glu143 of each subunit drawn explicitly.

Fig 2. Alignment of amino acid sequences of p19 from 11 different tombusvirus species.

The single-letter residue codes are coloured according to the nature of the amino acid side chain: (red) negatively-charged (D,E); (dark blue) aromatic (F,Y); (blue) positively-charged (K, R); (cyan) large polar amide-containing (Q,N); (orange) small polar hydroxyl-containing (S,T); (yellow) sulfur-containing (C,M); (grey) small aliphatic (A,G); (green) medium aliphatic (I,V,L); (purple-blue) imidazole (H); (violet) indole (W); (pink-brown) cyclised secondary amine (P). Sites identified as being under positive selection by all four analyses are highlighted in pale green. The secondary structure elements are indicated above the sequences: (barrels) α-helices; (arrows) β-strands. Viral species and NCBI accession numbers are as follows: CIR (Carnation italian ringspot virus, NC003500), TBS (Tomato bushy stunt virus, NC001554), AMC (Artichoke mottle crinkle virus, NC001339), LNV (Lisianthus necrosis virus, NC007983), PLV (Pear latent virus, NC004723), PNV (Pelagornium necrotic streak virus, NC005285), CNV (Cucumber necrosis virus, NC001469), GAL (Grapevine algerian latent virus, AY830918), CRV (Cymbidium ringspot virus, NC003532), CBL (Cucumber bulgarian latent virus, NC004725), MNV (Maize necrotic streak virus, NC007729).

Fig 3. Residues identified as being under positive selection.

See text for details of each of the four analyses. Light green indicates residues that only some of the analyses identified as being under positive selection. Dark green indicates residues found to be under positive selection in all four analyses. Note that some residues predicted to be under positive selection by HyPhy FEL2 (36, 43, 67, 107, 145) actually show no variation (see Fig 2).

Fig 4. Potential energies and number of hydrogen bonds between different components of simulated p19 systems.

Systems comprise wild-type and all permissible mutations of p19 alone or in complex with the siRNA. (A) Average potential energies (E_pot) of the interaction between the two protein subunits making up the p19 dimer (prot:prot). (B) Average potential energies (E_pot) of the interaction between the protein dimer and the RNA (prot;RNA). (C) Average potential energies (E_pot) of the complete protein/RNA complex (prot+RNA). The electrostatic and van der Waals contributions are shown separately alongside the total potential energy, as indicated above the graphs. (D) Average number of hydrogen bonds during the entire simulation between the two protein subunits making up the p19 dimer. (E) Average number of hydrogen bonds during the entire simulation between the protein dimer and the RNA. Solid bars correspond to simulations with RNA included, and empty bars to simulations without RNA present. Averages were calculated from the first 50 ns of simulation, after 10 ns equilibration, so that the averages calculated from the simulations of p19 with RNA bound (200 ns) are comparable to those of the simulations without RNA bound (50 ns). The error bars correspond to the standard deviation in all cases.