The genomic evolutionary dynamics and global circulation patterns of respiratory syncytial virus

Abstract

Respiratory syncytial virus (RSV) is a leading cause of acute lower respiratory tract infection in young children and the second leading cause of infant death worldwide. While global circulation has been extensively studied for respiratory viruses such as seasonal influenza, and more recently also in great detail for SARS-CoV-2, a lack of global multi-annual sampling of complete RSV genomes limits our understanding of RSV molecular epidemiology. Here, we capitalise on the genomic surveillance by the INFORM-RSV study and apply phylodynamic approaches to uncover how selection and neutral epidemiological processes shape RSV diversity. Using complete viral genome sequences, we show similar patterns of site-specific diversifying selection among RSVA and RSVB and recover the imprint of non-neutral epidemic processes on their genealogies. Using a phylogeographic approach, we provide evidence for air travel governing the global patterns of RSVA and RSVB spread, which results in a considerable degree of phylogenetic mixing across countries. Our findings highlight the potential of systematic global RSV genomic surveillance for transforming our understanding of global RSV spread.

Fig. 1. Maximum likelihood reconstructions of RSVA (1482 genomes/taxa; 2006–2020) and RSVB (1543 genomes/taxa; 1997–2020) complete genome phylogenies and genotypes identification.

Lineages that are not assigned to a genotype are shown in light grey. The SH-aLRT and UFB support values for the genotypes are provided in Supplementary Table S1.

Fig. 2. Posterior estimates of time-homogeneous predictor contributions to RSV diffusion between countries.

The predictors include the number of passengers travelling by air between each pair of countries represented in the data set (air travel, in dark red), population size at the origin and destination location (pop size ori & pop size dest, in blue), geographic distance (geo distance, in light green), absolute differences in latitude (lat diff, in dark orange) and sample sizes at the origin and destination locations (# taxa ori & # taxa dest, in dark green). The Y-axis represents the product of the coefficient (on a log scale) and the inclusion probability for the predictors (coefficient * Inclusion). (A, B: RSVA. C, D: RSVB. The plots on the left and right distinguish between analyses without and with sample size predictors respectively. E and F summarise the estimates for a single GLM-diffusion model applied to the combined RSVA and RSVB data sets at the country level. The grey boxes in the violin plots represent the median and quantile estimates. Violin plots are based on n = 507 (A), n = 535 (B), n = 45002 (C), n = 45002 (D), n = 452 (E) and n = 452 (F) post-burnin samples from the respective MCMC chains. Source data are provided as a Source Data file.

Fig. 3. Posterior estimates of the normalised entropy for RSVA and RSVB phylogeographic clustering by country during the most recent season (2019–2020) of INFORM-RSV sampling.

The normalised entropy ranges between 0 (~no mixing of lineages by country) and 1 (~random mixing with respect to country). Circles and error bars refer to the mean and 95% Highest Posterior Density (HPD) interval of the normalised entropy estimates respectively. The size of the circles is proportional to what fraction of the highest mean estimate each average estimate represents. The same is indicated by the colours of the circles, which range from blue for an average estimate that represents 0% of the highest value to bright red for the highest mean estimate. Entropy estimates are based on n = 901 post-burnin samples from the stationary MCMC chain. Source data are provided as a Source Data file.