Origin of the São Paulo Yellow Fever epidemic of 2017-2018 revealed through molecular epidemiological analysis of fatal cases

Abstract

The largest outbreak of yellow fever of the 21^st century in the Americas began in 2016, with intense circulation in the southeastern states of Brazil, particularly in sylvatic environments near densely populated areas including the metropolitan region of São Paulo city (MRSP) during 2017-2018. Herein, we describe the origin and molecular epidemiology of yellow fever virus (YFV) during this outbreak inferred from 36 full genome sequences taken from individuals who died following infection with zoonotic YFV. Our analysis revealed that these deaths were due to three genetic variants of sylvatic YFV that belong the South American I genotype and that were related to viruses previously isolated in 2017 from other locations in Brazil (Minas Gerais, Espírito Santo, Bahia and Rio de Janeiro states). Each variant represented an independent virus introduction into the MRSP. Phylogeographic and geopositioning analyses suggested that the virus moved around the peri-urban area without detectable human-to-human transmission, and towards the Atlantic rain forest causing human spill-over in nearby cities, yet in the absence of sustained viral transmission in the urban environment.

Figure 1

The current outbreak of yellow fever virus in Brazil (2016–2018). (A) Brazilian states with YFV cases recorded and sequenced in humans, non-human primates (NHP) and mosquitoes between 2017–2018. A grey circle marks the metropolitan region of São Paulo (MRSP). (B) Cycle threshold according to each of the 7 tissues analyzed for positive patients. Boxplots represent the 75th percentile, median, 25th percentile and the whiskers extend to the highest and lowest value in the 1.5x interquartile range. The different colors represent the different tissues analyzed. (C) Total cases recorded represented sylvatic cases of YFV (qRT-PCR positive cases) during the epidemiological weeks covered by the study (week 52 of 2017 to week 17 of 2018). (D) Relationship between the average coverage and the Ct values obtained for each sequenced sample. The data indicate that we obtained the expected direct inverse relationship between Ct and coverage parameters, as indicated by the trend line. (E) Combined coverage (normalized by the sample average) along all 36 sequenced YFV genomes generated in this study.

Figure 2

Time-stamped, MCC tree of YFV South American genotype I in Brazil recovered under the logistic-lognormal demographic model. The different colours indicate samples from different locations. The black circles represent posterior support upper than 0.7. The single synapomorphic change observed in Clade II [N1646T (NS3)] is shown in the box over the branch leading to Clade II-D. The three distinct introductions in the metropolitan region of São Paulo (MRSP) are shown (See also Fig. 3).

Figure 3

Highest posterior probability migration paths for the YFV Clades I and II from 2016 to 2018 towards the metropolitan region of São Paulo (MRSP), based on the analysis of 74 complete genomes. Although the sample size is small such that inferences should be made with caution, three distinct introductions in the MRSP are shown and strongly supported. The spatiotemporal spread was visualized with SPREAD3.

Figure 4

Spatial distribution of YFV deaths through time in non-human primates (NHP) and humans. Arrows indicate the general trend of movement around the metropolitan region of São Paulo (MRSP) estimated from distance matrices (see Fig. 5). The earliest cases in NHP are shown in the north, and later in the south and northeast of the MRSP. Most human cases are near sites with reported deaths of NPH, confirmed to be caused by YFV. The outbreak appears to have been confined mostly near the forested belt around the MRSP, contrasting with the almost empty, heavily urbanized center. Cardinal points are aligned according to the main axis of the page, (e.g., top being north, etc.). The figure was created by plotting the coordinates of reported cases to a satellite image available from Google Maps (google.com/maps) as background.

Figure 5

Neighbor joining tree calculated from pairwise geoposition distances among all the non-human primates and human cases available from the metropolitan region of São Paulo (MRSP).