The sporadic emergence of YFV in the Americas has been strongly associated with infection of red howler monkeys (Alouatta seniculus), which are particularly susceptible to disease. As exemplified in this study, nonhuman primate surveillance targeting this species remains an efficient strategy for monitoring enzootic YFV activity. Isolates made during a surveillance study aimed at investigating the ecology of infectious diseases in Venezuelan nonhuman primates were detected by cell culture and passaged once in Vero cells before sequencing, as previously described (4). Sequences were manually aligned in Se-Al (http://tree.bio.ed.ac.uk/software/seal/) and confirmed as nonrecombinant by using Recombination Detection Program (RDP4) (10). We obtained phylodynamic and phylogeographic estimates using Bayesian inference as implemented in BEAST v1.8.0 (11,12). We assessed the extent of geographic structuring using Bayesian tip-significance testing (13) based on the Markov chain Monte Carlo phylogenies estimated in BEAST. A Bayesian phylogeny was also inferred in Mr. Bayes (14) by using the General Time Reversible (GTR+I+Γ₄) model for the complete open reading frame sequences.

Figure 1

Figure 1. A) Bayesian maximum clade credibility (MCC) tree for YFV in the Americas based on 654 nt of the prM/E fragment. Taxon labels include year of isolation, GenBank accession number, strain designation,...

Results of Bayesian tip-significance testing showed statistically significant geographic clustering among Venezuelan YFV strains. The association index, parsimony score, and maximum monophyletic clade statistics provided strong support (p<0.01) that strains from Venezuela cluster by location, suggesting that YFV is maintained for long periods within Venezuela. Similar results have been shown for Peru (6), Brazil, and Trinidad (4). The high posterior probabilities, >0.99, observed at all nodes that delineate Venezuelan strains further support these conclusions (Figure 1).

Figure 2

Figure 2. Midpoint rooted Bayesian Markov chain Monte Carlo phylogeny based on yellow fever virus (YFV) complete open reading frame sequences. Numbers at nodes indicate posterior probabilities >0.9. Eastern and western Venezuelan...

We also found evidence of regionally independent evolution within Venezuela, as indicated by the existence of 2 phylogenetically distinct Venezuelan clades with posterior probabilities >0.99 (Figures 1, 2). The clade containing a 2004 strain (2A) from Guárico and a 2010 strain (10A) from Monagas represented all but one of the sequences from eastern Venezuela (i.e., east of Caracas; Table 1). The 7 other strains were collected on the western side of Venezuela (including Portuguesa, Apure, and Barinas States) (Table 1). Two YFV strains were collected in 2004 from eastern and western Venezuela. Despite their nearly synchronous collections in 2004, these sequences fell into distinct clades in the maximum clade credibility (MCC) phylogeny, indicating population subdivision. Although we cannot rule out sampling bias, these data suggest in situ evolution of YFV in Venezuela and regionally independent evolution in distinct geographic foci within the country. The mechanism promoting this population subdivision among YFV strains is unclear and requires further investigation. Further studies on the ecology of the areas where these viruses were isolated might help explain the observed population subdivision.

Although YFV had been maintained in situ for several years within Venezuela, our phylogeographic results indicated YFV movement between Brazil, Trinidad, and Venezuela. Brazil is the major source of YFV introductions into Venezuela, accounting for introductions of Venezuelan strains sampled in 1959, 1961, 1998, and more recently (i.e., independent introductions in 2004 in eastern and western Venezuela; Figure 1). The basal location of a 2004 sequence from the Brazilian Amazon within the western Venezuelan 2004–2007 clade (Figure 1) suggests that the 2005 Venezuelan epizootic/epidemic was initiated by an imported progenitor from the Brazilian Amazon Basin that later evolved independently within Venezuela. Our estimate of the most recent common ancestor for the node containing the 2004–2007 Venezuelan sequences was 2001 (95% highest posterior density 1997–2003), suggesting that the ancestral lineage existed for ≈3 years in Venezuela before its detection in nonhuman primates in 2004.

The 2004 Venezuelan (1A) intermediary descendent between the 1995 and 2009 Trinidad sequences is noteworthy. This cluster of Trinidad isolates was previously used as evidence of enzootic YFV maintenance within Trinidad during interepizootic periods (4). Incorporating our new sequences now shows movement between Trinidad and eastern Venezuela. In our phylogeny, strain 1A is the sister lineage of the 2008–09 Trinidad epizootic strains, with the 1995 Trinidadian isolate lying basal to these sequences, with high posterior probabilities (Figure 1). The position of the 2004 Venezuelan 1A sequence possibly reflects importation from Trinidad, which implies that enzootic YFV circulation in Trinidad is not isolated epidemiologically but gave rise to exportation of YFV to Venezuela.

In the absence of more dense sampling, whether bidirectional YFV movement occurs between Venezuela and Trinidad is unclear. Given the proximity and boating traffic between these countries, substantial mixing between their YFV populations would not be surprising. Isolation and sequencing of additional YFV isolates from eastern Venezuela are needed to further evaluate movement between Trinidad and Venezuela.

Complete genomes were sequenced for 5 representative Venezuelan YFV strains from eastern and western Venezuela (Table 1). Comparison of nucleotide and amino acid similarities showed a high degree of conservation across YFV genes (Table 2). The 9 strains selected for comparisons represent the full spectrum of known YFV genetic diversity (Table 2: Figure 2). The most diverse genes shared >97% aa sequence identity, with >99.9% aa sequence identity for most proteins, even though these strains were collected >30 years apart (Table 2). The Bayesian Markov chain Monte Carlo phylogeny based on complete open reading frame sequences strongly supported the inferred maximum clade credibility tree, with all Venezuelan sequences grouping together with strong posterior support. The eastern and western Venezuelan strains grouped separately, with strong support in all of our inferred phylogenies. A total of 5 substitutions delineated the Venezuelan sequences (Figure 2).

Our phylogeographic analysis supports in situ evolution of YFV within Venezuela, as well as regionally independent evolution within the country. Brazil was identified as the major source of YFV introductions into Venezuela, and sequence analysis showed that considerable YFV movement may occur between Trinidad and Venezuela. Results of our Bayes factor test for non-zero rates also support the epidemiologic link between Venezuela, Brazil, and Trinidad (Figure 1, panel C). The sequences generated in our study fill a major gap in the geographic sampling of YFV.