Assessing the role of live poultry trade in community-structured transmission of avian influenza in China

Yang, Qiqi; Zhao, Xiang; Lemey, Philippe; Suchard, Marc A.; Bi, Yuhai; Shi, Weifeng; Liu, Di; Qi, Wenbao; Zhang, Guogang; Stenseth, Nils Chr.; Pybus, Oliver G.; Tian, Huaiyu

It is difficult to exclude the potential impact on AIV dispersal of interspecies transmission between wild bird and domestic poultry populations (Fig. 2) . We attempted here to focus on transmission dynamics among poultry by reconstructing interspecies transmission events and removing virus lineages dominated by sequences from wild birds (SI Appendix, Materials and Methods). However, the success of this approach (SI Appendix, Fig. S7 ) may be affected by the underrepresentation of AIV genetic diversity coming from wild birds. In China, domestic poultry are sampled more intensively than wild birds and samples from wild birds are concentrated in only a few locations, such as Qinghai and Hong Kong. Thus, active surveillance of both domestic poultry and wild birds is needed for a better understanding of AIV dispersal dynamics in the country. We also directly compared the ability of empirical wild bird migration and poultry trade networks in China to explain AIV lineage movement, using Bayesian phylogeography inference with a GLM extension. We found that wild bird migration may be weakly associated with H5N1 virus dissemination in poultry but not associated with the spatial spread of H5N6 and H7N9 subtypes. This is consistent with the observation that most H5N6 and H7N9 strains to date are mainly circulating in domestic poultry, human, and environmental samples in China rather than in wild birds. We acknowledge two additional caveats. First, although we have proved the feasibility of using a gravity model to reconstruct the live poultry trade network and have found a live poultry trade network that could explain the observed gene flow, uncertainties between the gravity model and disease spread (31) still exist and should be addressed in future work. Second, even though we have tried to avoid the limits of multiresolution modularity and have demonstrated community structure robustness (Dataset S1, Fig. 5 , and SI Appendix, Figs. S8-S10), we note that the detected LPTCs are not unique; that is, there is no unique way to define the best partition of a network.