The Role of Human Transportation Networks in Mediating the Genetic Structure of Seasonal Influenza in the United States

Abstract

Recent studies have demonstrated the importance of accounting for human mobility networks when modeling epidemics in order to accurately predict spatial dynamics. However, little is known about the impact these movement networks have on the genetic structure of pathogen populations and whether these effects are scale-dependent. We investigated how human movement along the aviation and commuter networks contributed to intra-seasonal genetic structure of influenza A epidemics in the continental United States using spatially-referenced hemagglutinin nucleotide sequences collected from 2003-2013 for both the H3N2 and H1N1 subtypes. Comparative analysis of these transportation networks revealed that the commuter network is highly spatially-organized and more heavily traveled than the aviation network, which instead is characterized by high connectivity between all state pairs. We found that genetic distance between sequences often correlated with distance based on interstate commuter network connectivity for the H1N1 subtype, and that this correlation was not as prevalent when geographic distance or aviation network connectivity distance was assessed against genetic distance. However, these patterns were not as apparent for the H3N2 subtype at the scale of the continental United States. Finally, although sequences were spatially referenced at the level of the US state of collection, a community analysis based on county to county commuter connections revealed that commuting communities did not consistently align with state geographic boundaries, emphasizing the need for the greater availability of more specific sequence location data. Our results highlight the importance of utilizing host movement data in characterizing the underlying genetic structure of pathogen populations and demonstrate a need for a greater understanding of the differential effects of host movement networks on pathogen transmission at various spatial scales.

Fig 1. Aviation (A) and commuter (B) network models for the continental US.

Edge colors represent the number of individuals traveling between each state pair per day. Bar plots directly below each network depict the weight (total number of individuals moving in (colored red) and out (colored blue) of a state; top) and degree (total number of connections in (colored red) and out (colored blue) of a state; bottom) for each of that network’s nodes, ordered from left to right by the longitude of each state’s population center.

Fig 2. US commuting communities.

Two realizations using the simulated annealing algorithm to partition the US into communities based on an unweighted network (A) and weighted network (B) of county-to-county commuter flows. Modularity is similar across all realizations for a given network type, although exact community compositions differ. In all realizations, community boundaries do not neatly coincide with state borders.