Agricultural and geographic factors shaped the North American 2015 highly pathogenic avian influenza H5N2 outbreak

Abstract

The 2014-2015 highly pathogenic avian influenza (HPAI) H5NX outbreak represents the largest and most expensive HPAI outbreak in the United States to date. Despite extensive traditional and molecular epidemiological studies, factors associated with the spread of HPAI among midwestern poultry premises remain unclear. To better understand the dynamics of this outbreak, 182 full genome HPAI H5N2 sequences isolated from commercial layer chicken and turkey production premises were analyzed using evolutionary models able to accommodate epidemiological and geographic information. Epidemiological compartmental models embedded in a phylogenetic framework provided evidence that poultry type acted as a barrier to the transmission of virus among midwestern poultry farms. Furthermore, after initial introduction, the propagation of HPAI cases was self-sustainable within the commercial poultry industries. Discrete trait diffusion models indicated that within state viral transitions occurred more frequently than inter-state transitions. Distance and sample size were very strongly supported as associated with viral transition between county groups (Bayes Factor > 30.0). Together these findings indicate that the different types of midwestern poultry industries were not a single homogenous population, but rather, the outbreak was shaped by poultry industries and geographic factors.

Fig 1. Evolutionary history of HPAI H5N2 isolated from commercial poultry premises, 2015.

(A) Bayes factor (BF) tests between molecular clock and coalescent evolutionary models. For each coalescent model (exponential growth [Expo] and extended Bayesian skyline plot [EBSP]), BF was calculated using the constant coalescent model as reference (Const, indicated with asterisk) under the same molecular clock model. Two horizontal gray reference lines denote log(BF) = 1 and log(BF) = 5, which represent support and very strong support, respectively, for improved fit over the reference. (B) Molecular clock rate (substitutions per site per year) comparison between molecular clock and coalescent evolutionary models. (C) The estimated time of the most recent common ancestor (TMRCA; decimal year) compared between molecular clock and coalescent evolutionary models. (D) Maximum clade credibility tree representing the ancestral reconstruction of poultry industry (layer chicken vs. turkey) across the evolutionary history of the outbreak. The ancestral reconstruction assumed an EBSP coalescent and strict molecular clock evolutionary model. Tree branches are colored based on the most probable poultry industry of the descendant node. Thin gray node bars represent the 95% highest posterior density (HPD) of the node height (i.e., the time at which that ancestor is estimated to have existed).

Fig 2. Comparison of hypothesized HPAI H5N2 epidemiological compartmental models.

(A) Each compartmental model represents a Susceptible-Infectious-Removed (SIR) model with varied population heterogeneity: 1) a single, closed, homogenous population, 2) a single, homogenous population with a continual external source of virus (U), 3) a closed population, stratified by poultry system (turkeys (T) and layer chickens (C)), and 4) the stratified population with a continual external source of virus. (B) Compartmental model fit for the midwestern highly pathogenic avian influenza (HPAI) H5N2 outbreak, 2015. Akaike’s information criteria for Markov chain Monte Carlo (AICM) calculated based on the posterior distribution of the structured tree likelihood was used to evaluate the relative model fit for the four assessed compartmental models under differing molecular clock assumptions. Under both molecular clocks, Model 3 provided the best model fit. (C) Estimated infectious period of layer chicken and turkey farms during the 2015 midwestern highly pathogenic avian influenza (HPAI) H5N2 outbreak. During model specification, an informative prior was provided for the Bayesian process. This prior probability distribution was based on the reported average time from HPAI confirmation to depopulation plus 5 days to allow for delay between infection and HPAI confirmation. Model 3 estimated the infectious period for layer chickens to be longer than expected given the prior information.

Fig 3. Discrete trait diffusion model of HPAI H5N2 among midwestern county groups.

(A) Maximum clade credibility tree representing the ancestral reconstruction of county groups across the evolutionary history of the outbreak. The ancestral reconstruction was based on an EBSP coalescent and strict molecular clock evolutionary model. Tree branches are colored based on the most probable county group of the descendant node. Thin gray node bars represent the 95% highest posterior density (HPD) of the node height (i.e., the time at which that ancestor is estimated to have existed). (B) Diffusion rate summary among county groups. County groups were defined based on state and composition of host type within the county. Counties with only turkey cases (turkey exclusive; T) were grouped separately from counties with at least one layer chicken case (mixed poultry; C). Arrows represent transition rates with strong support (Bayes factor > 10) with arrow thickness proportional to the magnitude of transition rate. (C) Conditional effect size of environmental and geographic covariates within the generalized linear model (GLM). Conditional effect size represents the effect size of the variable coefficient given inclusion in the GLM. Supported covariates (Bayes factor > 3) are bolded. Covariates are ordered by Bayes factor. The dashed gray line represents a conditional effect size of 0, signifying little impact of the covariate on viral dispersal.