Convergent trends and spatiotemporal patterns of Aedes-borne arboviruses in Mexico and Central America

Abstract

Background: Aedes-borne arboviruses cause both seasonal epidemics and emerging outbreaks with a significant impact on global health. These viruses share mosquito vector species, often infecting the same host population within overlapping geographic regions. Thus, comparative analyses of the virus evolutionary and epidemiological dynamics across spatial and temporal scales could reveal convergent trends.

Methodology/principal findings: Focusing on Mexico as a case study, we generated novel chikungunya and dengue (CHIKV, DENV-1 and DENV-2) virus genomes from an epidemiological surveillance-derived historical sample collection, and analysed them together with longitudinally-collected genome and epidemiological data from the Americas. Aedes-borne arboviruses endemically circulating within the country were found to be introduced multiple times from lineages predominantly sampled from the Caribbean and Central America. For CHIKV, at least thirteen introductions were inferred over a year, with six of these leading to persistent transmission chains. For both DENV-1 and DENV-2, at least seven introductions were inferred over a decade.

Conclusions/significance: Our results suggest that CHIKV, DENV-1 and DENV-2 in Mexico share evolutionary and epidemiological trajectories. The southwest region of the country was determined to be the most likely location for viral introductions from abroad, with a subsequent spread into the Pacific coast towards the north of Mexico. Virus diffusion patterns observed across the country are likely driven by multiple factors, including mobility linked to human migration from Central towards North America. Considering Mexico's geographic positioning displaying a high human mobility across borders, our results prompt the need to better understand the role of anthropogenic factors in the transmission dynamics of Aedes-borne arboviruses, particularly linked to land-based human migration.

Fig 1. Epidemiological trends of CHIKV in the Americas from 2013 to 2018.

(A) Monthly number of CHIKV cases reported to the Pan-American Health Organisation (PAHO) between 2013 and 2017, grouped by PAHO region. The largest epidemic peak in Mexico (PAHO North America region) occurred in 2015, highlighted in blue. (B) Monthly number of confirmed CHIKV cases in Mexico grouped by year, as reported by the SINAVE. (C) Monthly number of confirmed CHIKV cases per geographic region in Mexico during 2015. A map of the states included in each geographic region is shown below. Details of the states included in each region are available in Table 1. Plots were generated using the ggplot package (https://ggplot2.tidyverse.org/index.html) in R. Original base layer maps use as a source for geospatial base layer data public domain maps imported from the Natural Earth data project (https://cran.r-project.org/web/packages/rnaturalearth/index.html).

Fig 2. Time-scaled analysis for CHIKV in Mexico.

(A) ML phylogenetic tree of all CHIKV complete genome sequences from the Americas included in our analysis, with tips coloured by region of collection. (B) Time-calibrated phylogeographic analysis of the CHIKV CCNA. Tip and nodes for locations within Mexico and are coloured by region, with the main clusters identified numbered (1–6). The map in the inset shows pairs of locations where transition rates significantly explain the phylogenetic diffusion process, as inferred under a BSSVS analysis. Only transition rates with a posterior probability (PP) > 0.5 are shown, coloured by Bayes Factor (BF). (C) Posterior probability densities for the TMRCAs of six CHIKV transmission clusters in Mexico. Median values for each distribution are indicated, with the distributions coloured by the most probable location of each MRCA. Collection dates for InDRE04 (the first sequence corresponding to a CHIKV imported case into Mexico, collected on 2014-05-30), and InDRE51 (the first confirmed autochthonous case within Mexico, collected on 2014-10-15) are shown in red and grey, respectively. Plots were generated using the ggplot package (https://ggplot2.tidyverse.org/index.html) in R. Original base layer maps use as a source for geospatial base layer data public domain maps imported from the Natural Earth data project (https://cran.r-project.org/web/packages/rnaturalearth/index.html).

Fig 3. Epidemiological trends of DENV in Mexico from 2013 to 2019.

(A) The proportion of serotyped DENV cases between 2000 and 2018 in Mexico shown in reference to the total number of serotyped samples per year for all the country. Data provided by InDRE for this study is shown between the solid black lines (delimiting 2013 to 2018). Data from previous years was obtained from published sources [71]. The DENV-3 and DENV-4 serotypes together with potential co-infections (reported between 2013 and 2018) are grouped into the ‘Other’ category. (B) Monthly number of cases assigned to each DENV serotype in Mexico from 2013 to 2018. (C) Breakdown of serotype proportions in different geographic regions in Mexico between 2013 and 2019, relative to the total number of serotyped samples for each location. Plots were generated using the ggplot package (https://ggplot2.tidyverse.org/index.html) in R. Original base layer maps use as a source for geospatial base layer data public domain maps imported from the Natural Earth data project (https://cran.r-project.org/web/packages/rnaturalearth/index.html).

Fig 4. Time-scaled analysis for DENV-2 in Mexico.

(A) ML phylogenetic tree for DENV-2 in the Americas, with tips coloured according to the geographic region of collection. (B) Time-calibrated phylogeographic analysis for the DENV-2 CAM lineage, with the MCC tree displaying different sampling locations for Mexico. Tip and nodes shapes are included for locations within Mexico, coloured by region. Distinct clades identified within the CAM lineage are designated as Clade A (including one cluster from Mexico) and Clade B (including two clusters from Mexico). The map in the inset shows pairs of locations for which transition rates were inferred to be significant under a BSSVS analysis. Only transition rates with a posterior probability (PP) > 0.5 are shown, coloured by Bayes Factor (BF). Original base layer maps use as a source for geospatial base layer data public domain maps imported from the Natural Earth data project (https://cran.r-project.org/web/packages/rnaturalearth/index.html).