Characterising routes of H5N1 and H7N9 spread in China using Bayesian phylogeographical analysis

Abstract

Avian influenza H5N1 subtype has caused a global public health concern due to its high pathogenicity in poultry and high case fatality rates in humans. The recently emerged H7N9 is a growing pandemic risk due to its sustained high rates of human infections, and recently acquired high pathogenicity in poultry. Here, we used Bayesian phylogeography on 265 H5N1 and 371 H7N9 haemagglutinin sequences isolated from humans, animals and the environment, to identify and compare migration patterns and factors predictive of H5N1 and H7N9 diffusion rates in China. H7N9 diffusion dynamics and predictor contributions differ from H5N1. Key determinants of spatial diffusion included: proximity between locations (for H5N1 and H7N9), and lower rural population densities (H5N1 only). For H7N9, additional predictors included low avian influenza vaccination rates, low percentage of nature reserves and high humidity levels. For both H5N1 and H7N9, we found viral migration rates from Guangdong to Guangxi and Guangdong to Hunan were highly supported transmission routes (Bayes Factor > 30). We show fundamental differences in wide-scale transmission dynamics between H5N1 and H7N9. Importantly, this indicates that avian influenza initiatives designed to control H5N1 may not be sufficient for controlling the H7N9 epidemic. We suggest control and prevention activities to specifically target poultry transportation networks between Central, Pan-Pearl River Delta and South-West regions.

Fig. 1. Phylogeography models of H5N1 and H7N9.

Bayesian MCC phylogeneies and between-region diffusion networks on HA gene segments of H5N1 (left panel) and H7N9 (right panel) in China. The sequences are classified according to their location grouped by economic division. Bohai Economic Rim (BER): Beijing, Hebei, Shandong; Central (CT): Anhui, Henan, Hubei, Hunan, Jiangxi, Shanxi; North-east (NE): Heilongjiang, Jilin, Liaoning; North-west (NW): Gansu, Ningxia, Qinghai, Shaanxi, Xinjiang; Pan-Pearl River Delta (PRD): Fujian, Guangdong; South-west (SW): Guangxi, Guizhou, Sichuan, Tibet, Yunnan, Chongqing; Yangtze-River Delta (YRD): Jiangsu, Shanghai, Zhejiang. Trees have been scaled according to taxa dates (representing sample collection date)

Fig. 2. Level of Bayes Factor support for each transmission route.

The left and right panels display the level of Bayes Factor (BF) support for each of the transmission routes considered for H5N1 and H7N9 analyses respectively. The x-axis represents the origin location and the y-axis represents the destination. Level of BF support is coloured according to classifications described in Table S1.

Fig. 3. Generalised linear model.

From left to right, the two panels show (i) Bayes Factor (BF), and (ii) Mean Coefficients (meanCeffects) and their 95% highest posterior credible intervals. Plots for H5N1 and H7N9 are displayed side by side. For all predictors excluding Distance, green and orange colours represent origin and destination locations respectively. In the BF plots, the dashed line indicates BF = 3. In the meanCeffects plot, the dashed line indicates 0. Note BF results are displayed on a log10 scale