Introduction
The intense recent movements of human populations are reflected in the current diffusion and expansion of the HIV epidemic around the world [1]–[3]. Population bottlenecks, genetic recombination, genetic drift, and founder effects are characteristics associated with viral dissemination within these human populations and define the variability and nature of the establishment of the HIV/AIDS pandemic [4], [5]. A single transmission event in an unaffected area may result in the rapid spread of a unique viral form within a group with specific risk behaviors [6], [7], resulting in the establishment of the epidemic in that area [8], [9]. However, because of the rapid evolution of HIV and its global diffusion, the exact pathways of its dissemination are often unclear.
The emergence of HIV-1 resulted from the cross-species transmission of simian immunodeficiency viruses from chimpanzees to humans in West–Central Africa at the beginning of the 20th century [10]. Group M, responsible for the vast majority of HIV infections worldwide, initially spread throughout Africa, and in response to the actions of several genetic forces, has diversified into different subtypes [11]–[14]. The spread of these variants in the human population was not noticed for nearly eight decades. The first records of infection date from 1981 in American patients infected with subtype B viruses, who presented with clinical symptoms of what is today known as AIDS [15], [16].
The spread of subtype B from Africa initially occurred via a single introduction to Haiti in the 1960s, which was probably associated with the return of Haitian professionals from work missions in the Congo [5]. After the expansion of the epidemic in the Caribbean, current evidence points to the dissemination of the virus from there directly into North America. The subsequent transmission and spread of the virus in the United States allowed the epidemic to grow and expand to other parts of the world [5], [17]–[19]. Today, HIV-1 subtype B occupies an important position in the epidemiological profiles of various countries in Europe, Asia, and Africa, and is also the only subtype circulating in several countries in the Americas [19]–[23].
The dissemination of an infectious disease reflects the complex interactions between the infectious agent, its host, and the environment [24]. A strategy widely used in epidemiological research to identify the pathways of dissemination of an infectious agent is to combine the analysis of sociodemographic evidence with that of complementary phylogenetic data [3], [5], [24], [25]. In a recent study of the evolutionary history of HIV-1 subtype B, Gilbert et al. (2007) demonstrates the emergence of this subtype from Africa to American countries (starting in Haiti) by examining 117 subtype B sequences from 19 countries [5]. However, South America was poorly represented in this work, with only nine sequences from four countries. Considering that the Caribbean has close economic, historical, and even social relationships with several countries in South America [26]–[28], it is reasonable to investigate if HIV-1B was directly transmitted from the Caribbean to South American countries. Thus, the present study aimed to investigate by phylogenetic and genetic statistics analyses the role of South American countries in the establishment of the HIV-1 subtype B in the Americas.
Materials and Methods
Dataset Selection
Around 6000 HIV-1 subtype B sequences of protease and portions of reverse transcriptase segments of the pol gene (nucleotides 2253–3233 relative to strain HXB2) were selected from the Los Alamos HIV Sequence Database (http://www.hiv.lanl.gov/) and GenBank (http://www.ncbi.nlm.nih.gov/nucleotide/). To ensure the selection of high-quality data, we selected the sequences that met the following criteria: (a) the samples were isolated from patients living in the Americas; (b) the country of origin was clearly established; (c) only one sequence per patient was included; (d) no report of intersubtype recombination; (e) no evidence of hypermutation; and (f) no occurrence of premature stop codons, frameshift mutations, or ambiguity saturation (excess of undetermined nucleotides). Moreover, sequences from the same country of isolation, describe in the same study and phylogenetically close related were excluded from our dataset. To understand the spread of the virus without compromising the quality of the results, all the sequences were examined for evidence of intersubtype recombination. We selected sequences with a minimum confidence threshold for pure subtype B of 0.95 with a window size of 200 nt, using the RIP tool at the Los Alamos HIV Sequence Database. The dataset was also evaluated using additional reference sequences by constructing a neighbor-joining phylogeny, to guarantee the selection of nonintersubtype recombinants. Although all HIV sequences currently described can be considered intrasubtype recombinants at some level, these evolutionary events are probably insignificant in the context of the origin and geographical grouping of HIV subtypes [29], [30]. After a carefully selection, a dataset of 313 HIV-1B pol publically available sequences retrieved from 27 countries from North America, South America, Central America and Caribbean were used in the following analyses.
Sequence alignments were created using MUSCLE [31] and manually edited to optimize them. Four African sequences of subtype D and two of subtype C were selected as outgroups. Three different alignments were constructed for this study: one set including 313 sequences for ML analysis, one set comprising 263 sequences for Bayesian analysis, and a third set for the genetic structure analysis. These sets are available upon request.
Phylogenetic Reconstruction
The reconstruction of phylogenetic trees was performed with the maximum likelihood (ML) method using a set of 313 subtype B sequences that met our quality criteria (Table S1). The ML analysis was conducted with the program phyML [32] under the GTR model of nucleotide substitution, with a proportion of invariable sites, and substitution rate heterogeneity (GTR+G+I). Nearest-neighbor interchange was used for heuristic tree searches. Support for the internal nodes was obtained with parametric bootstrapping using 1000 replicates.
A Bayesian Markov Chain Monte Carlo (MCMC) approach, implemented in BEAST ver. 1.5.4 [33], was used with a set of 263 sequences (Table S1) to reconstruct the phylogenetic tree and estimate the date of the most-recent common ancestor of the epidemic in the Americas. The evolutionary history was inferred with a Bayesian Skyline (BSP) coalescent tree prior, under an uncorrelated lognormal relaxed clock, and the GTR+I+G model of nucleotide substitution. Three independent runs of 300 million steps sampled every 30,000 generations were performed and the effective sample size was evaluated in TRACER [34]. The maximum sum of clade credibility tree was selected from the posterior tree distribution.
We inferred an ML phylogeny to investigate the role of South America in the HIV subtype B epidemic. This approach supported a relationship between the sequences from the Caribbean and those from South America. Notably, the sequences from Colombia, Venezuela, Brazil, Suriname, and Guyana were intermingled with those from Trinidad and Tobago, the Dominican Republic, and Haiti.
To further explore these relationships, Bayesian trees were inferred using 263 sequences (Table S1) representing North America (n = 71), Central America and the Caribbean (n = 88), and South America (n = 104) under an uncorrelated relaxed clock and with a Bayesian Skyline coalescent tree prior. The effective sample size (ESS) was calculated by combining the outputs from the three runs for each model, and excluding the first 10% of steps as the burn-in for each chain. The Bayesian MCMC-independent runs converged on similar values and all parameter estimates showed ESS values of more than 200.
Genetic Diversity
The population genetic structure of HIV subtype B among the countries of North America, Central America, South America, and the Caribbean was quantified using estimates of the F statistics [35]. A set of 308 sequences was built, excluding countries represented by only one sequence, and any ambiguous nucleotide was changed to “N”. Estimates were calculated using analysis of molecular variance (AMOVA) [36] in Arlequin ver. 3.5.1.2 [37] under the Kimura two-parameter model with 10,000 randomizations. Invariable sites were included and sites with gaps/missing data were considered. A nonmetric multidimensional scaling plot was obtained with SPSS ver. 8 (Inc., Chicago, IL).
Discussion
The emergence and dispersal patterns of HIV-1 subtype B from its epicenter in Africa were major events in the history of the epidemic, which has become a major public health issue. The HIV-1 subtype B pandemic has been the focus of several research groups in distinct disciplines, because it was the first subtype to be isolated in industrialized nations and the first to spread with mobile populations [17], [38]–[40]. In the field of phylogeography, HIV-1 subtype B is the subject of continuing debate. Several suppositions have been made about its temporal and geographical distribution patterns, including its origin and dissemination [5], [41]–[43]. Gilbert et al. (2007) fueled the discussions when they traced the initial spread of the epidemic from Africa in 1966 (1962–1970), showing that the epidemic of subtype B most likely began in Haiti, given the monophyletic cluster of sequences from this nation [5]. The authors suggested that Haiti was the key conduit for the introduction of subtype B into the United States before its global dissemination. Our results add another piece to this epidemiological puzzle, providing evidence that the spread of HIV-1B in the early 1960s in the Americas was not as unidirectional as initially suggested. The phylogenetic and statistical approaches used here point to the significant participation of South American countries in the transmission and evolution of the HIV-1 subtype B epidemic in the Americas.
The Caribbean countries undoubtedly played an important role in the HIV-1 subtype B epidemic. On the pol gene phylogeographic reconstruction, a consistent clade of isolates from the Caribbean arose simultaneously from the same common ancestor of the Pandemic Clade (A+B+C). Although the results inferred from our phylogenetic trees do not elucidate the basal position of the Caribbean Clade with respect to the Pandemic Clade, our analysis of genetic diversity and the results of previous studies that included ancestral sequences point to this conclusion [5], [41]. In addition, we cannot state that Haiti [5] or any other country was the origin of the subsequent dissemination because ancient sequences are unavailable. Within the Caribbean clade, the epidemic in Trinidad and Tobago seems to have primarily derived from two or more effective HIV-1 introduction events, because on all the trees obtained, most of the sequences from this country form two distinct clades (Figure 1). This result differs from the findings of previous studies that, based on the gag and env genes, and on the nearly complete HIV-1 genome, reported a monophyletic epidemic in that country [5], [19], [44]. We also provide evidence for the introduction of non-pandemic subtype B clades from the Caribbean into countries in northern South America, such as Suriname, Guyana, Brazil, Colombia, and Venezuela. Despite the existence of phylogenetic evidence that the United States was the midpoint between Caribbean countries and the global spread of HIV-1 subtype B, our results do not show a direct transmission event of the HIV-1B from the Caribbean to the United States early in the epidemic. The direct introduction of the virus into the United States from the Caribbean would have generated a genetic signature in the viral genome, and when dealing with appropriate genetic markers, such as the pol gene, the phylogenetic pattern expected under such model should effectively group a significant number of sequences from the United States, dating from the beginning of the epidemic, within the Caribbean strains. Our evolutionary analysis of HIV-1B agrees with other studies in the timing of its introduction into the Americas [5], and estimates the date of the most-recent common ancestor to be 1964 (1950–1967). The clustering of the Caribbean strains together in older branches could be the result of founder effects from one or a few introductions to that region in the 1960s.
Both the Bayesian and ML methods show a cluster derived from the same common ancestor of the Caribbean clade that groups isolates collected at different times from South American countries (pandemic clade A), albeit with low probability. This cluster occupies a position basal to the pandemic clades within subtype B (Figure 1), providing evidence for two epidemiological scenarios: (a) the direct introduction of HIV-1B into South America, which seeded a secondary outbreak in the United States; or (b) the concurrent spread of HIV-1B from the Caribbean to South America and North America.
Supporting scenario a, the beginning of the HIV epidemic in the Americas coincided with a boom in oil production by Venezuela [26], [28]. Great changes in its economic situation caused Venezuela to implement policies to attract immigrants from Colombia, Ecuador, Peru, Cuba, Trinidad and Tobago, the Dominican Republic, and the United States [26], [28]. The population of migrants from other Latin American countries tripled between 1970 and 1980 according to Venezuela's censuses [26]. It would not be surprising if this movement of people from the Caribbean also introduced and disseminated the HIV epidemic into South America. Further reinforcing our hypothesis, Leal and VillaNova (2010) used 66 near-full-length genomic sequences (8160 bp) of worldwide HIV-1 subtype B isolates to show that the epidemic in Brazil, a South American country, shares a common ancestor with a “North American–European cluster”, and that Haitian strains occupy the deepest positions in this phylogeny [44]. Together, these various lines of evidence suggest a link between the Caribbean epidemic and the direct introduction of HIV-1B into South America.
However, according to scenario a, after spreading to the countries of South America, the virus was introduced from there into North America. Emigrations from Mexico to the United States have been the largest migratory movements on the planet [26]. Therefore, Mexico could represent the “entrance door” for the epidemic into the United States. Interestingly, of the 10 phylogenetic relationships involving Mexican sequences throughout the whole ML tree, eight are linked to South American strains (80%). The Bayesian trees include only two Mexican sequences because information about the sampling dates was unreliable (Table S1).
Finally, the bidirectional spread of HIV-1B from the Caribbean, as suggested by scenario b, may also have participated in the establishment of the epidemic in the Americas because three distinct clusters, including pandemic cluster B, arise from the Caribbean clade on the Bayesian trees. The historical factors involved in scenario a could also support scenario b. Furthermore, the historical registers of AIDS cases among Haitian individuals in the 1970s in the United States are an indication of the pathway of the epidemic into North America from the Caribbean [45]. It is also possible that the HIV-1B pandemic in the Americas arose from both scenarios acting simultaneously. In this case, the epidemic in the South American countries might have originated from the virus circulating in the Caribbean and from the diffusion of the pandemic from the United States.
The epidemic in Cuba deserves special attention because in addition to its extraordinary genetic diversity [46], Cuba seems to have a remarkable subtype B epidemic. One clade encompassing 12 Cuban sequences is clearly related to samples from South American countries, suggesting a different epidemiological link to that inferred for the Caribbean region. Such a relationship supports the hypothesis that the expansion of the epidemic from the Caribbean countries was not specifically directed towards the United States, as previously suggested [5]. After all, South American countries played an important role historically in the definition of the politics of Cuba [47].
As well as the phylogenetic analysis, we calculated the degree of differentiation between the continent-specific compartments of the epidemic using AMOVA. Human migrations are a confounding factor in the already complex dissemination of HIV. The exchange of people among countries can mix viral subpopulations and mask the historical processes that create patterns of genetic variation and geographic signatures within an epidemic [24]. Despite the high rate of migration in the Americas, the data presented here demonstrate a significant degree of viral population structure within the various regions (Table 1 and Figure 4). Such structuring can originate in the selection effects of the host's immune system, in recombination processes, and in founder effects [19]. However, there is a lack of evidence of such selection [19] and all the sequences analyzed here met the criterion of no intersubtype genetic mixing. Therefore, founder effects that acted on the viral population in the beginning of the epidemic and are still detectable 40 years later seem to better explain the genetic compartmentalization observed.
The ΦST estimates, which test the viral population structure in the Caribbean, North America, and South America, show that the Caribbean epidemic has a closer genetic relationship to that established in South America (Figure 2), suggesting an epidemiological link between the two Americas. However, the ΦST estimates from the South American and North American epidemics point to the circulation of genetically related viruses, attributable to the constant movement of human populations between these two regions. The relationships among the sequences from specific countries are shown in Figure 2 and corroborate some of the results derived from our trees. Again, it seems that the Caribbean strains are more closely genetically related to those from South American countries than to those from North America. Furthermore, apart from Guyana and Suriname, which have epidemics related to that of Trinidad and Tobago, the sequences from all the countries of South America are closely related, suggesting an intricate epidemic for that continent. Finally, the role of Mexico in the dispersal of HIV-1B throughout the Americas should be rethought, because our phylogenetic and genetic statistical analyses point to a connection between the United States, Mexico, and the South American countries.
The low support displayed in our Bayesian tree is probably linked to the extreme genetic conservation of the pol gene, combined with the large number of sequences used in this study. We recognize that the pol gene may not offer sufficient genetic variation to ensure a strong phylogenetic signal and thus confer sufficient statistical support for the branches of our trees. However, a recent study evaluating the evolution of bacterial genes under simulated biological conditions revealed that realistic estimates of the statistics not necessarily estimate how well a reconstructed phylogeny that actually represents cladistic relationships exist in nature [48]. Moreover, our analysis successfully reiterated the date and geographic trace of previous studies based on other genes, such as env and gag [5], [19], [44]. Similarly, it was demonstrated that this gene is useful in the identification of transmission events by phylogenetic means even when codon positions associated with drug resistance are maintained [49]. We used the pol gene because it provided the largest set of dated sequences, sampled across the widest possible number of countries in the Americas. The recovery of ancestral sequences from American countries, especially South American countries, should undoubtedly better trace the spread of HIV-1B, strengthening the support for one or other possible scenario.
Our results do not contradict those of previous studies, but in fact, our statistical genetic and phylogenetic analyses add further pieces to the historical puzzle of HIV-1 subtype B in the Americas, revealing that part of the epidemic in South America derived directly from the Caribbean epidemic. We propose a scenario that began with the introduction and spread of the virus locally in the Caribbean region, followed by its dispersal into northern South America, establishing an epidemic genetically similar to that in the Caribbean. An epidemiological link between South America and North America was easily established by several waves of migration from the various countries of Latin America to the United States [26]. However, a direct link between the Caribbean and North America also contributed to the dissemination of HIV-1B and historical registers confirm this connection [50], [51]. The data presented here offers a new perspective on the epidemic of HIV-1 subtype B in the Americas. This work also highlights the utility of population genetic methods in understanding the evolution and spread of this epidemic, contributing primarily to our understanding of the interactions between the virus and the migration processes governing the diffusion of human populations.