Parallel evolution in the emergence of highly pathogenic avian influenza A viruses

Abstract

Parallel molecular evolution and adaptation are important phenomena commonly observed in viruses. Here, we exploit parallel molecular evolution to understand virulence evolution in avian influenza viruses (AIV). Highly-pathogenic AIVs evolve independently from low-pathogenic ancestors via acquisition of polybasic cleavage sites. Why some AIV lineages but not others evolve in this way is unknown. We hypothesise that the parallel emergence of highly-pathogenic AIV may be facilitated by permissive or compensatory mutations occurring across the viral genome. We combine phylogenetic, statistical and structural approaches to discover parallel mutations in AIV genomes associated with the highly-pathogenic phenotype. Parallel mutations were screened using a statistical test of mutation-phenotype association and further evaluated in the contexts of positive selection and protein structure. Our resulting mutational panel may help to reveal new links between virulence evolution and other traits, and raises the possibility of predicting aspects of AIV evolution.

Fig. 1. Summary of analytical approach.

A summary of our approach to identifying HAPMs (HP-cluster associated parallel mutations) in AIV genomes. 1. Large-scale data curation and phylogenetic analysis to detect genotypes and phenotypes (HP lineages highlighted in red). 2. Phylogenetically informed subsampling to reduce alignment sizes, whilst retaining HP clusters and key mutational pathways along internal branches. 3. Detecting candidate HAPMs across multiple HP clusters (highlighted in orange) through the reconstruction of ancestral states. 4. Selecting HAPMs by statistical testing for genotype-to-phenotype associations using an evolutionary-informed model (represented in the matrix by aa Y occurring in Site X). 5. Identifying HAPMs evolving under positive selection using different dN/dS (ω) estimation methods. 6. Inferring the possible effects of HAPMs on protein structure stability and function, using existing structural data and experimental literature.

Fig. 2. Historical occurrence of documented HP AIV outbreaks, and independent emergence of HP lineages.

a For each H7NX HP case, the recorded date of emergence and duration (in years), the country of emergence (highlighted in red), the virus subtype causing the outbreak, and the reported avian host species are indicated in the map. b Phylogenetic distribution of HP lineages in the H7 subsampled phylogeny. Branches leading to LP sequences are shown in black. HP sequences (highlighted in red) correspond to unique genotypes defined by a pCS (shown next to the phylogeny), as they form independent well-supported lineages (as defined in ‘Methods’, indicated by C1-C9, showing posterior probability and bootstrap support values >80 or 0.8 for nodes of interest), matching the historical occurrence of the documented HP outbreaks. c Historical occurrence of documented H5NX HP outbreaks, following the description used for panel (a). Data for C8 is not included, as since 1996, this lineage circulates worldwide. d Phylogenetic distribution of HP lineages in the H5 subsampled phylogeny, following the description used for panel (b). Distinct, well-supported HP clusters are highlighted in red (C1-C8). The H5NX HP viruses within cluster C8 (clade 2.3.4 and descending lineages) are highlighted in blue. Maps for the global distribution for the H5NX and H7NX HPAIV were generated with the tool http://gamapserver.who.int/mapLibrary/default.aspx under the CC BY-NC-SA 3.0 IGO license with preprint permission authorisation 362444 for WHO copyrighted material. Maps were further edited using Adobe Illustrator for the purposes needed.

Fig. 3. Reconstruction of amino acid evolution at selected HAPMs.

Maximum clade credibility (MCC) trees showing the reconstruction of ancestral states for four illustrative HAPMs (a−d). Branches are coloured according to the inferred ancestral amino acid states (tip states not shown). Nodes representing the immediate LP ancestors and nodes from which the HP clusters directly descend are shown with circles. The posterior probabilities for a given amino acid state occurring at the specified node are indicated. The HP sequences (HP clusters) are highlighted in orange, whilst the immediate ancestral LP sequences are highlighted in grey. For details on sampling date, location, and host species, see Supplementary Figs. 5–8.

Fig. 4. Functional relevance of HAPMs of H7NX viruses.

Protein structure models for the a HA (Hemagglutinin) and the b PB2/PB1 (Polymerase complex) of the H7NX viruses. Functionally relevant HAPMs are highlighted in red within the structural models. Only those HAMPs that are significantly associated with the HP phenotype are indicated. Estimated values for free energy change (ddG) are indicated for these sites. Negative values or those close to zero can be interpreted as fairly stabilising/neutral mutations with respect to protein structure.

Fig. 5. Functional relevance of HAPMs of H5NX viruses.

Protein structure models for the a HA (Hemagglutinin) and the b PB2 (Polymerase complex) of the H5NX viruses. See Fig. 4 legend for further details.