Dynamics and ecology of a multi-stage expansion of Oropouche virus in Brazil

Abstract

In March 2024, the Pan American Health Organization (PAHO) issued an alert in response to a rapid increase in Oropouche fever cases across South America. Brazil has been particularly affected, reporting a novel reassortant lineage of the Oropouche virus (OROV) and expansion to previously non-endemic areas beyond the Amazon Basin. Utilising phylogeographic approaches, we reveal a multi-scale expansion process with both short and long-distance dispersal events, and diffusion velocities in line with human-mediated jumps. We identify forest cover, banana and cocoa cultivation, temperature, and human population density as key environmental factors associated with OROV range expansion. Using ecological niche modelling, we show that OROV circulated in areas of enhanced ecological suitability immediately preceding its explosive epidemic expansion in the Amazon. This likely resulted from the virus being introduced into simultaneously densely populated and environmentally favourable regions in the Amazon, such as Manaus, leading to an amplified epidemic and spread beyond the Amazon. Our study provides valuable insights into the dispersal and ecological dynamics of OROV, highlighting the role of human mobility in colonisation of new areas, and raising concern over high viral suitability along the Brazilian coast.

Keywords: Oropouche virus; ecological niche modelling; phylodynamics; phylogeography.

Figure 1.. Dispersal history and dynamics of OROV lineages in Brazil.

(A) Dispersal history of OROV lineages inferred through continuous phylogeographic reconstructions. Lineage dispersal events between Brazilian states with a posterior probability >=0.95 are displayed by solid arrows, and dispersal events with a posterior probability <0.95 are displayed by dashed arrows. Additionally, the location of the different areas is represented by transparent grey dots whose surface is proportional to the number of local lineage dispersal events, i.e. phylogenetic branches inferred as remaining in that state. Brazilian states are coloured according to the estimated date of the first invasion event (median date computed from the 100 trees sampled from the posterior distribution). (B) Evolution through time of the spatial wavefront distance, representing the maximal distance from the epidemic origin over time. (C) Evolution through time of the weighted diffusion coefficient, a dispersal metric that measures the dispersal capacity of viral lineages. (D) Kernel density plots with the branch-weighted diffusion coefficient against the geographic distance travelled by each branch (both axes being log-transformed).

Figure 2.. Environmental conditions associated with OROV lineage dispersal locations over time (for segment M).

Figure panels show the spatial distribution of six main environmental factors (units specified): evergreen broadleaf forest cover (%) (A), human population density (normalised between 0 and 255 per km² for visual clarity) (B), Ae. aegypti ecological suitability (probability of occurrence) (C), mean annual temperature (°C) (D), banana harvested area (in hectares, log-transformed) (E), and cocoa harvested area (in hectares, log-transformed) (F) in the top rows. Circles on the map depict the end node of dispersal locations inferred by continuous phylogeography, sized by the number of dispersal events in an area, and coloured by the timing of the event. Bottom rows of each figure panel are line graphs depicting the environmental covariates associated with the locations of OROV lineage dispersal events in Brazil. Each plot illustrates how specific ecological conditions have changed over time (2022–2024) at the sites of viral lineage dispersal. The embedded tables show the association between environmental conditions and the dispersal location of inferred OROV lineages. Based on the analysis of 100 posterior trees obtained from continuous phylogeographic inference, the table reports Bayes factor (BF) supports for association between environmental raster values and tree node locations. Following the scale of interpretation of Kass and Raftery (24), we highlight BF values >20 considered as strong supports.

Figure 3.. Ecological niche prediction for OROV local circulation in Brazil.

A) Predicted ecological suitability for OROV transmission across Brazil utilising all disease occurrence points. Suitability predictions range from unsuitable (0) to highly suitable (1). B) Ecological suitability prediction using input disease occurrence points sampled before 2024. C) Ecological suitability prediction using input disease occurrence points sampled before mid-2023. D) Response curves and relative importance (RI) for individual environmental factors obtained from the random forest (RF) suitability prediction model. These response curves (five iterations) depict the relationship between the environmental factors and the response (the ecological suitability of OROV transmission). E) Ecological suitability values of OROV dispersal locations (for segment M) overlayed on weekly recorded OROV cases in Brazil. Suitability values are estimated from three ecological niche models (ENM), as described in the text.