Epidemiological dynamics and phylogeography of influenza virus in southern China

Abstract

Background: Understanding the epidemiological dynamics of influenza virus is central to surveillance and vaccine strain selection. It has been suggested that tropical and subtropical regions represent the global source of influenza epidemics. However, our understanding of the epidemiological dynamics of influenza virus in these regions is limited by a relative lack of long-term data.

Methods: We analyzed epidemiological and virological data on influenza recorded over a period of 15 years from the metropolitan city of Shenzhen in subtropical southern China. We used wavelet analysis to determine the periodicity of influenza epidemics and molecular phylogeographic analysis to investigate the role of Shenzhen and southern China in the global evolution of influenza virus.

Results: We show that southern China is unlikely to represent an epicenter of global influenza activity, because activity in Shenzhen is characterized by significant annual cycles, multiple viral introductions every year, limited persistence across epidemic seasons, and viruses that generally are not positioned on the trunk of the global influenza virus phylogeny.

Conclusions: We propose that novel influenza viruses emerge and evolve in multiple geographic localities and that the global evolution of influenza virus is complex and does not simply originate from a southern Chinese epicenter.

Figure 1.

Seasonality of influenza epidemics in Shenzhen, China, 1995–2009. A, Time series of the standardized number of influenza virus–positive specimens (calculated as the monthly number of influenza virus–positive specimens divided by the annual number of influenza virus–positive specimens; purple line, left y-axis), the “percentage positivity” (calculated as the monthly number of influenza virus–positive specimens divided by the monthly number of all specimens tested; blue line, left y-axis), and influenza-like illness (ILI) consultation rates (calculated as the weekly number of consultations for ILI divided by the weekly number of all consultations; red line, right y-axis). All time series have been square-root transformed. B, Monthly standardized number of influenza virus–positive specimens (purple) and percentage positivity (blue) during 1995–2008 (the 2009 pandemic influenza A virus subtype H1N1 season was omitted). Color bars represent the average proportion of influenza virus–positive patients identified in each month of the year. Error bars show standard errors based on variation between years. C, Same as B, but for monthly ILI consultation rate (in red) during 2003–2008. D, Wavelet power spectrum of standardized influenza positive counts during 1995–2009, identifying changes in periodicities over time (left) and average periodicity (right). Power increases from blue to red so that red indicates stronger periodicities. Black lines highlight periodicities reaching statistical significance (here, 1-year periodicities). Shaded areas indicate the presence of edge effects. E, Same as panel D, but for ILI consultation rate, 2003–2009.

Figure 2.

Weekly distribution of influenza virus subtypes in Shenzhen, China, 2003–2009. Blue: seasonal influenza A virus subtype H1N1 (A/H1N1); red: seasonal influenza A virus subtype H3N2 (A/H3N2); green: influenza B virus (B); black: 2009 pandemic influenza A virus subtype H1N1 (A/H1N1pdm09). A/H1N1 predominated in 2005, 2006, and 2008; A/H3N2 predominated in 2003, 2007, and in 2009 before the emergence of A/H1N1pdm09. B was predominant in 2004 but showed higher activity in late 2007 and the beginning of 2008.

Figure 3.

Phylogeographic analysis of influenza A virus subtype H3N2. A, Maximum likelihood phylogeny of global isolates of the virus, with the trunk lineages shown by bold lines. Dashed lines indicate a magnification of that part of the tree where isolates from Shenzhen and other areas in southern China fall and indicated by light and dark orange colors, respectively (branches extended to the same level). B, Estimated proportions of geographic regions in the trunk lineages were plotted against their genetic distance from the tree root (units of nucleotide substitutions per site).