Phylodynamics of Yellow Fever Virus in the Americas: new insights into the origin of the 2017 Brazilian outbreak

Abstract

Yellow fever virus (YFV) strains circulating in the Americas belong to two distinct genotypes (I and II) that have diversified into several concurrent enzootic lineages. Since 1999, YFV genotype I has spread outside endemic regions and its recent (2017) reemergence in non-endemic Southeastern Brazilian states fuels one of the largest epizootic of jungle Yellow Fever registered in the country. To better understand this phenomenon, we reconstructed the phylodynamics of YFV American genotypes using sequences from nine countries sampled along 60 years, including strains from Brazilian 2017 outbreak. Our analyses reveals that YFV genotypes I and II follow roughly similar evolutionary and demographic dynamics until the early 1990s, when a dramatic change in the diversification process of the genotype I occurred associated with the emergence and dissemination of a new lineage (here called modern). Trinidad and Tobago was the most likely source of the YFV modern-lineage that spread to Brazil and Venezuela around the late 1980s, where it replaced all lineages previously circulating. The modern-lineage caused all major YFV outbreaks detected in non-endemic South American regions since 2000, including the 2017 Brazilian outbreak, and its dissemination was coupled to the accumulation of several amino acid substitutions particularly within non-structural viral proteins.

Figure 1

Time-scaled Bayesian phylogeographic MCC tree of the YFV genotype I prM/E gene sequences. Branches are colored according to the most probable location state of their descendent nodes as indicated at the legend (top left). Key ancestral nodes of modern-lineage subclades with high posterior probability support (clade credibility >0.99) are indicated with dark full triangles. All horizontal branch lengths are drawn to a scale of years. The tree is automatically rooted under the assumption of a relaxed molecular clock. (AR: Argentina, BR-CO: Brazil Central-West, BR-N: Brazil North, BR-NE: Brazil Northeast, BR-S: Brazil South, BR-SE: Brazil Southeast, CO: Colombia, EC: Ecuador, PA: Panama, TT: Trinidad and Tobago, VE: Venezuela). Viral migration events occurred within the old lineages (1946–1992) and the modern one (1989–2016) are summarized in the maps. Lines between locations represent branches in the Bayesian MCC tree along which location’s transitions occur. Maps were created with CorelDraw from templates obtained from d-maps.com. (South America: http://www.d-maps.com/carte.php?num_car=28522&lang=en; and Brazil: http://www.d-maps.com/carte.php?num_car=16016&lang=en).

Figure 2

Time-scaled Bayesian MCC phylogeographic tree of the YFV genotype II prM/E gene sequences. Branches are colored according to the most probable location state of their descendent nodes as indicated at the legend (top left). All horizontal branch lengths are drawn to a scale of years. The tree is automatically rooted under the assumption of a relaxed molecular clock. (BR-N: Brazil North, BO: Bolivia, EC: Ecuador, PE: Peru, TT: Trinidad and Tobago). Viral migration events occurred within genotype II (1956–1981) are summarized in the maps. Lines between locations represent branches in the Bayesian MCC tree along which location’s transitions occur. Maps were created with CorelDraw from templates obtained from d-maps.com. (South America: http://www.d-maps.com/carte.php?num_car=28522&lang=en; and Brazil: http://www.d-maps.com/carte.php?num_car=16016&lang=en).

Figure 3

Demographic history of YFV genotypes I and II in the Americas. Mean estimates of the effective number of infections (N _e) (solid line) are shown together with the 95% HPD intervals (shaded area) of the Bayesian skyline for whole Genotype I dataset (a), Genotype I subsampled dataset (b), Genotype I old-lineages (orange line) and modern-lineage (purple line) (c), and Genotype II (d). Vertical segmented and continuous lines in demographic plots of panels a, b and c represents the estimated mean T_MRCA for the genotype I modern-linage and its 95% HPD interval, respectively. Horizontal segmented line in panel b represents the median Ne value at present estimated with the complete Genotype I dataset. The vertical axes represent the estimated N _e on a logarithmic scale. Time scales are in calendar years.

Figure 4

Evolutionary analysis of YFV genotype I and II complete coding sequences (CDS). (a) Correlation between the sampling date of each sequence and the genetic distance of that sequence from the root of a maximum likelihood phylogeny of the YFV CDS (R² = 0.78) is displayed (top left). (b) Time-scaled Bayesian MCC tree of YFV CDS. The shaded box highlights the modern-lineage clade. Genotype I branches corresponding to old, modern-basal and modern-subclade 1E lineages were colored in blue, light purple and gray, respectively. Reconstructed ancestral key nodes of all genotype I sequences (N1), all modern-lineage sequences (N2) and the modern subclade 1E (N3) are indicated. Inferred amino acid substitutions between ancestral key nodes are shown in relation to the polyprotein positions.