Genomic epidemiology unveils the dynamics and spatial corridor behind the Yellow Fever virus outbreak in Southern Brazil

Abstract

Despite the considerable morbidity and mortality of yellow fever virus (YFV) infections in Brazil, our understanding of disease outbreaks is hampered by limited viral genomic data. Here, through a combination of phylogenetic and epidemiological models, we reconstructed the recent transmission history of YFV within different epidemic seasons in Brazil. A suitability index based on the highly domesticated Aedes aegypti was able to capture the seasonality of reported human infections. Spatial modeling revealed spatial hotspots with both past reporting and low vaccination coverage, which coincided with many of the largest urban centers in the Southeast. Phylodynamic analysis unraveled the circulation of three distinct lineages and provided proof of the directionality of a known spatial corridor that connects the endemic North with the extra-Amazonian basin. This study illustrates that genomics linked with eco-epidemiology can provide new insights into the landscape of YFV transmission, augmenting traditional approaches to infectious disease surveillance and control.

Fig. 1.. Human incidence of YFV in Brazilian macroregions, 2015–2022.

(A) Time series of monthly reported human cases in four Brazilian macroregions (those reporting the majority of cases) and monthly mosquito-viral suitability measure (index P) aggregated (mean) across the macroregions. Index P is calibrated to YFV and Aedes aegypti, interpreted as spillover risk, and presented here index P is presented by the lines as baseline (black) and shifted by +1 month (blue). The inset plot shows the distribution (y axis) of human cases according to the value of index P (x axis) extracted at each time point of reported human YFV cases (shifted +1 month, standardized by yearly maximum). (B) Age distribution of human cases across the years (light blue) versus Census 2010 data (black), both aggregated across the Southeast and South macroregions. (C) Spatial distribution of human cases (red filled points) and one example of a spatial covariate (forest cover) used for modeling the Southeast macroregion. For all covariates, see fig. S1. Covariates were normalized by their maximum for visualization, with covariates population density, Atlantic and NHP cases first transformed with log₁₀ (color scale in the bottom right). Spatial boundaries (black) are states: São Paulo (SP), Espírito Santo (ES), Minas Gerais (MG), Rio de Janeiro (RJ). (D) Marginal posterior distributions (standardized to appear on the same scale) for individual regression weights for the 10 covariates explored.

Fig. 2.. Spatial hotspots as a mixture of past reporting and proportion unvaccinated.

(A) Probability of reporting of human YFV cases per municipality as predicted by the spatial regression model. (B) Proportion unvaccinated (1, vaccination coverage) per municipality. (A and B) Variables are discretized into three categories to provide a bivariate color scale in (C) with a maximum of nine colors for best interpretability. (C) Spatial distribution of both probability of occurrence and proportion unvaccinated presented in bivariate color scale. Spatial distribution of both probability of occurrence and proportion unvaccinated presented in bivariate color scale [i.e., (A) and (B) superimposed into a bivariate color scale]. The three largest cities in each state are highlighted (São Paulo: São Paulo, Guarulhos, Campinas; Minas Gerais: Belo Horizonte, Uberlândia, Contagem; Rio de Janeiro: Duque de Caxias, Rio de Janeiro, São Gonçalo; Espírito Santo: Serra, Vila Velha, Cariacica).

Fig. 3.. YFV South America I genotype in Brazil.

(A) MCC phylogeny inferred using the 147 novel sequences obtained in this study plus 296 publicly available sequences from GenBank. Colors represent different sampling locations (Brazilian states). Colored bars below the tree represent the (top) clade (middle) host and Brazilian macroregion of sampling. Clade-defining mutations have been highlighted on each branch. (B) Map of Brazil presenting the states under investigation highlighted within macroregions: RR, Roraima; PA, Pará; DF, Distrito Federal; GO, Goiás; BA, Bahia; MG, Minas Gerais; ES, Espírito Santo; RJ, Rio de Janeiro; SP, São Paulo state; PR, Paraná; SC, Santa Catarina; RS, Rio Grande do Sul. (C) Root-to-tip regression of sequence sampling date against genetic divergence from the root of the outbreak clade. Sequences from the endemic Amazon basin (states of Roraima, Pará, and Tocantins) are highlighted.

Fig. 4.. Spatiotemporal spread of South America I genotype Clade Ia in Brazil.

(A) Root-to-tip regression of sequence sampling date against genetic divergence from the root of the outbreak clade. (B) Phylogeographic reconstruction of the spread of the South America I genotype clade Ia in Brazil (n = 163). Circles represent nodes of the MCC phylogeny and are colored according to their inferred time of occurrence. Shaded areas represent the 80% highest posterior density interval and depict the uncertainty of the phylogeographic estimates for each node. Solid curved lines denote the links between nodes and the directionality of movement. Differences in forest coverage population density are shown on a green-white scale. (C) Map of Brazil highlighting the spatial area under investigation. MG, Minas Gerais state; ES, Espírito Santo state; RJ, Rio de Janeiro state; BA, Bahia state.

Fig. 5.. Spatiotemporal spread of South America I genotype Clade IIb in Brazil.

(A) Root-to-tip regression of sequence sampling date against genetic divergence from the root of the outbreak clade. (B) Phylogeographic reconstruction of the spread of South America I genotype clade IIb in Brazil (n = 270). Circles represent nodes of the MCC phylogeny and are colored according to their inferred time of occurrence. Shaded areas represent the 80% highest posterior density interval and depict the uncertainty of the phylogeographic estimates for each node. Solid curved lines denote the links between nodes and the directionality of movement. Differences in forest coverage population density are shown on a green-white scale. (C) Map of Brazil highlighting the spatial area under investigation. GO, Goiás state; MG, Minas Gerais state; ES, Espírito Santo state; RJ, Rio de Janeiro state; SP, São Paulo state; PR, Paraná state; SC, Santa Catarina state; RS, Rio Grande do Sul state.