Background: The segmented RNA genome of avian Influenza viruses (AIV) allows genetic reassortment between co-infecting viruses, providing an evolutionary pathway to generate genetic innovation. The genetic diversity (16 haemagglutinin and 9 neuraminidase subtypes) of AIV indicates an extensive reservoir of influenza viruses exists in bird populations, but how frequently subtypes reassort with each other is still unknown. Here we quantify the reassortment patterns among subtypes in the Eurasian avian viral pool by reconstructing the ancestral states of the subtypes as discrete states on time-scaled phylogenies with respect to the internal protein coding segments. We further analyzed how host species, the inferred evolutionary rates and the dN/dS ratio varied among segments and between discrete subtypes, and whether these factors may be associated with inter-subtype reassortment rate. Results: The general patterns of reassortment are similar among five internal segments with the exception of segment 8, encoding the Non-Structural genes, which has a more divergent phylogeny. However, significant variation in rates between subtypes was observed. In particular, hemagglutinin-encoding segments of subtypes H5 to H9 reassort at a lower rate compared to those of H1 to H4, and Neuraminidase-encoding segments of subtypes N1 and N2 reassort less frequently than N3 to N9. Both host species and dN/dS ratio were significantly associated with reassortment rate, while evolutionary rate was not associated. The dN/dS ratio was negatively correlated with reassortment rate, as was the number of negatively selected sites for all segments. Conclusions: These results indicate that overall selective constraint and host species are both associated with reassortment rate. These results together identify the wild bird population as the major source of new reassortants, rather than domestic poultry. The lower reassortment rates observed for H5N1 and H9N2 may be explained by the large proportion of strains derived from domestic poultry populations. In contrast, the higher rates observed in the H1N1, H3N8 and H4N6 subtypes could be due to their primary origin as infections of wild birds with multiple low pathogenicity strains in the large avian reservoir.

XML of Eurasian AIVXML were created in BEAUti v1.6.2 for analysis in BEAST v1.7.3 (Drummond & Rambaut, 2007, Drummond et al. 2012). For each file, one internal segment of the same 344 subsampled Eurasian AIV (specified with subtype, host, location and isolate dates) was used as input data, labelled with the segment name. Files labelled with “original” represent the original time-scaled phylogenies of different segments. The substitution, clock and coalescent models are provided along with the priors placed on the model parameters. The other files labelled “empirical” and “H, N or HN” were generated to estimate the transition rate among discrete subtypes (HA, NA or HA-NA combined), with the implementation of a discrete trait model and BSSVS (Lemey et al. 2009) using empirical trees for each segment. These other files contain additional XML code not generated by BEAUTi in order to allow the re-use of the previously generated empirical tree sets.