An estimated 1.5 million people were living with the HIV type 1 (HIV-1) in Latin America in 2012, most of them concentrated in Brazil (40%), Mexico (11%), Colombia (10%), Venezuela (7%) and Argentina (6.5%) . The HIV prevalence in the adult population (15–49 years) ranges from 0.2% in Mexico to more than 1.0% in Belize, Guyana and Suriname . Most of the HIV epidemics in this region are concentrated in and around networks of MSM, although heterosexual HIV transmission is increasing in the older epidemics in South America and injecting drug use is another significant route of HIV transmission, especially in the southern cone of South America and in Mexico .
The HIV-1 group M subtype B is the most prevalent clade in Latin America, accounting for about 70% of infections in the region . The spread of HIV-1 subtype B in the Americas probably occurred via a single introduction event from Central Africa into Haiti around the middle 1960s and later dissemination of the virus from Haiti to other Caribbean islands and to the United States . The virus that entered the United States was further disseminated from this country to other countries around the world, establishing a ‘subtype B pandemic’ (BPANDEMIC) clade, whereas other subtype B lineages seem to have remained mostly restricted to the Caribbean [subtype B Caribbean (BCAR) clades] . A recent study conducted by our group analyzed 1042 HIV-1 subtype B pol gene sequences from 14 different Caribbean countries and revealed that nonpandemic BCAR lineages have been widely disseminated through the Caribbean region since the late 1960s, accounting for an important fraction of current HIV-1 infections in several countries including Haiti and the Dominican Republic (∼75%), Jamaica (∼50%) Trinidad and Tobago (∼95%), and other Lesser Antilles (∼40–75%) .
Two previous studies suggest that nonpandemic BCAR lineages may have been also directly disseminated from the Caribbean islands into South  and Central  American countries. The study of Junqueira et al. identified a few HIV-1 subtype B pol sequences from Brazil, Colombia, Guyana, Suriname and Venezuela that were phylogenetically intermixed among basal nonpandemic Caribbean sequences, suggesting a direct epidemiological link between the Caribbean and South American epidemics. Another recent study showed that a minor fraction (5.5%) of Panamanian subtype B pol sequences were also intermixed among nonpandemic BCAR strains and further suggests that some of those BCAR clades were mainly disseminated in Panama by heterosexual transmission . Overall, these results suggest that the BCAR clades have not remained confined to the Caribbean region, but have been also disseminated to continental regions of the Americas. The relative prevalence of the BPANDEMIC and BCAR clades across different Latin American countries, however, remains largely unknown.
The objective of this study was to estimate the current prevalence of the BPANDEMIC and BCAR clades in Latin America and to reconstruct the spatiotemporal dynamics of dissemination of the HIV-1 BCAR clades in the region. For this, we used a comprehensive dataset of HIV-1 subtype B pol sequences (n = 7654) isolated from 18 different Latin American countries between 1989 and 2011. These Latin American sequences were combined with subtype B reference sequences representative of the BPANDEMIC (US/France = 300) and the BCAR (Caribbean/Panama = 316) clades and then subjected to maximum likelihood and Bayesian phylogeographic analyses.
HIV-1 subtype B pol sequence dataset
We downloaded all HIV-1 subtype B pol sequences from Latin America that covered the entire protease and partial reverse transcriptase (PR/RT) regions (nucleotides 2253–3260 relative to HXB2 clone) and were available at the Los Alamos HIV Database (http://www.hiv.lanl.gov) by December 2013. Additional HIV-1 subtype B pol sequences from Latin America covering only part of the reverse transcriptase (nucleotides 2673–3203 relative to the HXB2 clone) were also downloaded for some countries with few PR/RT sequences available (Bolivia, Suriname and French Guyana). The subtype assignment of all sequences included here was confirmed using the REGA HIV subtyping tool v.2  and by performing phylogenetic analyses (see below) with HIV-1 group M subtype reference sequences. Only one sequence per individual was selected and those sequences containing frameshift mutations or with incorrect subtype assignment were removed. This resulted in a final dataset of 7654 subtype B pol sequences isolated from 18 Latin American countries between 1989 and 2011. These sequences were aligned with subtype B pol (PR/RT) sequences from the United States (n = 165), France (n = 135), the Caribbean (n = 279) and Panama (n = 37), representative of the BPANDEMIC and the BCAR clades as described previously [5,7]. Sequences were aligned using the Clustal W program  and all sites associated with major antiretroviral drug resistance in protease (30, 32, 46, 47, 48, 50, 54, 76, 82, 84, 88 and 90) and reverse transcriptase (41, 65, 67, 69, 70, 74, 100, 101, 103, 106, 115, 138, 151, 181, 184, 188, 190, 210, 215, 219 and 230) were excluded. All alignments are available from the authors upon request.
Maximum likelihood phylogenetic trees were inferred under the generalized time reversible (GTR)+I+Γ nucleotide substitution model selected using the jModeltest program . The maximum likelihood trees were reconstructed with the PhyML program  using an online web server . Heuristic tree search was performed using the subtree-pruning-regrafting (SPR) branch-swapping algorithm and the reliability of the obtained topology was estimated with the approximate likelihood-ratio test  based on the Shimodaira–Hasegawa-like procedure. The maximum likelihood trees were visualized using the FigTree v1.4.0 program .
Analysis of the spatiotemporal dispersion pattern
The evolutionary rate, the age of the most recent common ancestor (TMRCA) and the spatial diffusion pattern of nonpandemic HIV-1 subtype B clades circulating in South America were jointly estimated using the Bayesian Markov Chain Monte Carlo approach as implemented in BEAST v1.8 [15,16] with BEAGLE to improve run-time . Analyses were performed using the GTR+I+Г4 nucleotide substitution model, a relaxed uncorrelated lognormal molecular clock model , and a Bayesian skyline coalescent tree prior . The mean evolutionary rates previously estimated for the subtype B pol gene (2.0–2.5 × 10–3 substitutions/site per year) [7,20–22] were incorporated as an informative prior interval. Migration events throughout the phylogenetic history and the most relevant migration pathways were reconstructed using a reversible discrete phylogeography model and the Bayesian stochastic search variable selection approach , with a continuous-time Markov chain (CTMC) rate reference prior . Three Markov Chain Monte Carlo chains were run for 500 × 106 generations and then combined using LogCombiner v1.8. Convergence and uncertainty of parameter estimates were assessed by calculating the effective sample size and 95% highest probability density (HPD) values, respectively, after excluding the initial 10% of each run with Tracer v1.6 . The maximum clade credibility tree was summarized with TreeAnnotator v1.8 and visualized with FigTree v1.4.0. Migratory events were summarized using the cross-platform SPREAD application .
Detection of HIV-1 subtype B Caribbean clades in the majority of Latin American countries
In order to estimate the relative prevalence of pandemic (BPANDEMIC) and nonpandemic (BCAR) subtype B lineages in Latin America, pol (PR/RT) sequences from different Latin American countries were divided into six subsets: Central America (n = 688), Mexico (n = 1677), Argentina (n = 1548), Brazil-I (n = 1329), Brazil-II (n = 1329) and other South American countries (n = 909). A seventh subset containing shorter subtype B pol (reverse transcriptase) sequences from some Latin American countries poorly represented in the PR/RT dataset (Bolivia = 45, French Guyana = 108, Suriname = 21) was also constructed. Each of the seven Latin American subsets was combined with a reference subtype B dataset selected from a previous study  containing 500 sequences representative of the BPANDEMIC (US/France = 300) and the BCAR (Caribbean = 200) clades (Table S1, http://links.lww.com/QAD/A619). The maximum likelihood analyses of all PR/RT (Fig. 1a and Fig. S1, http://links.lww.com/QAD/A619) and reverse transcriptase (Fig. 1b) subsets confirmed the complete segregation of the BPANDEMIC reference sequences in a highly supported (approximate likelihood-ratio test >0.90) monophyletic clade nested within basal BCAR reference sequences. The maximum likelihood analyses also confirmed the circulation of BCAR sequences in most Latin American countries, although with highly variable prevalence (Fig. 2 and Table S2, http://links.lww.com/QAD/A619). The BCAR sequences reach a high prevalence (40–50%) in French Guyana and Suriname; low prevalence (1–10%) in Brazil, Colombia, Ecuador, Mexico, Panama and Venezuela; and very low prevalence (<1%) in Argentina, El Salvador, Honduras and Peru. We found no evidence of circulation of BCAR clades in Bolivia and Chile. The number of PR/RT or reverse transcriptase sequences from Belize, Costa Rica, Guatemala, Guyana, Nicaragua, Paraguay and Uruguay was too small (n <10) to allow any conclusion about the relative prevalence of different subtype B clades circulating in those Latin America countries.
Spatiotemporal dispersal pattern of the HIV-1 subtype B Caribbean clades in Latin America
To reconstruct the origin and spatiotemporal dynamics of nonpandemic subtype B Latin American lineages, the HIV-1 BCAR PR/RT sequences with known sampling date from Latin America here identified (n = 103) were combined with BCAR PR/RT sequences from the most widely sampled (n >10) Caribbean islands [Dominican Republic (n = 123), Jamaica (n = 73), Trinidad and Tobago (n = 50), and Haiti (n = 12)] and from Panama (n = 37), previously identified [5,7]. The BCAR sequences were further aligned with subtype D PR/RT sequences (n = 10) from the Democratic Republic of Congo that was pointed as the most probable source of subtype B strain introduced in the Americas . HIV-1 subtypes B and D sequences were classified into 14 discrete geographic locations (Table S3, http://links.lww.com/QAD/A619) and subjected to Bayesian phylogeographic analysis.
The mean estimated evolutionary rate of the HIV-1 BCAR/D pol dataset was 2.1 × 10−3 substitutions/site per year (95% HPD 2.0 × 10−3 – 2.2 × 10−3 substitutions/site per year), whereas the corresponding median coefficient of rate variation was 0.31 (95% HPD: 0.27–0.35), supporting the selection of a relaxed molecular clock model. The root location of the HIV-1 subtype B ancestor was most probably placed in the island of Hispaniola (Dominican Republic/Haiti) (posterior state probability = 0.92) (Fig. 3), consistent with previous findings [4,5]. The median estimated TMRCA of subtypes B/D (1956), subtype D (1968) and subtype B (1968) were also very similar to that previously obtained using different pol and env datasets [4,5] (Table 1). The close match of major spatiotemporal calibration points across different studies validates the time scale inferred from this analysis and indicates that the overall phylogeographic reconstruction was quite robust to the inclusion of new BCAR sequences from Latin America.
After the introduction of HIV-1 subtype B into Hispaniola around the middle 1960s, nonpandemic BCAR lineages were independently disseminated to other countries from the Caribbean and Latin America from the early 1970s onwards. Some of those viral migrations seeded secondary outbreaks that resulted in the origin of several country-specific BCAR subclades including those previously identified in Trinidad and Tobago (BCAR-TT) [4,5], Jamaica (BCAR-JM-I)  and Panama (BCAR-PA-I, BCAR-PA-II and BCAR-PA-III) , and others here identified in Argentina (BCAR-AR), Brazil (BCAR-BR-I, BCAR-BR-II and BCAR-BR-III), Guyana (BCAR-GY), Mexico (BCAR-MX-I and BCAR-MX-II) and Venezuela (BCAR-VE) (Fig. 3). The nonpandemic clades BCAR-TT, BCAR-JM-I and BCAR-BR-I seem to have originated around the early 1970s, whereas most of the remaining country-specific BCAR clades probably arose between the late 1970s and the middle 1980s (Fig. 3 and Table 1).
Reconstruction of viral migrations across time suggests that Hispaniola was the major hub of dissemination of nonpandemic subtype B clades in the region and further identified a few secondary hubs in the Caribbean (Trinidad and Tobago) and South America (Brazil and Guyana) (Fig. 4a and 4b). The BCAR-TT clade was independently disseminated from Trinidad and Tobago to other Caribbean islands and to several South American countries including Brazil, Guyana (originating the BCAR-GY clade), Suriname and Venezuela. The BCAR-GY clade was disseminated from Guyana to Suriname and the BCAR-BR-I clade was disseminated from Brazil to Argentina at multiple times (originating the BCAR-AR clade). The Bayes factor tests for significant nonzero rates supports epidemiological linkage between Hispaniola and most other Caribbean and Latin American countries included in the study (with exception of Argentina and Guyana) as well as between Trinidad and Tobago and Jamaica/Guyana/Brazil, between Brazil and Argentina, and between Guyana and Suriname (Fig. 4c and 4d and Table S4, http://links.lww.com/QAD/A619).
The HIV-1 subtype B virus was probably originally introduced into Haiti seeded by the epidemic from the Democratic Republic of Congo around the middle 1960s . After a short period of local expansion within the island of Hispaniola (shared by Haiti and the Dominican Republic), the virus seems to have moved out on several independent occasions. The introduction of the virus into the United States around the late 1960s explosively amplified the number of new cases of HIV-1 subtype B infection and originates a BPANDEMIC strain that was disseminated across the world . Other secondary outbreaks simultaneously emerged in the Caribbean  and Latin America [6,7] as the result of short-distance disseminations of nonpandemic BCAR strains out of Hispaniola. This study demonstrates that BCAR strains reached nearly all countries in Latin America, although their prevalence is usually much lower than that estimated for the BPANDEMIC clade (Fig. 2). The only exceptions in the region were Suriname, French Guyana and probably Guyana, where both BPANDEMIC and BCAR clades seem to circulate at roughly similar prevalence.
Our results indicate that Haiti and Dominican Republic, which together are home to about 75% of people living with HIV in the Caribbean , were probably the major sources of BCAR lineages disseminated into the region. Nonpandemic BCAR strains started to spread from Hispaniola in the beginning of the 1970s and would have reached Trinidad and Tobago, Jamaica, Brazil, Colombia, Ecuador, El Salvador, Honduras, Mexico, Panama, Suriname and Venezuela in the following years. Trinidad and Tobago can be viewed as a secondary hub, seeding tertiary BCAR outbreaks in short-distanced countries such as Jamaica, Venezuela, Guyana and Brazil. Jamaica, by contrast, seems to have played a minor role in the regional dispersion of BCAR strains. We also identified short-distance spreading of BCAR lineages from Brazil to Argentina and from Guyana to Suriname, indicating that some South American countries also acted as secondary hubs of dissemination of nonpandemic subtype B lineages in the region.
Although Dominican Republic, Haiti and Trinidad and Tobago were pointed as the most important sources of BCAR lineages disseminated to Latin America, we cannot rule out the possible role of other Caribbean islands with high prevalence of BCAR strains such as Martinique, Guadeloupe and other Lesser Antilles  not included in our phylogeographic analysis because of the very low numbers (n <10) of PR/RT sequences available. This geographical sampling bias may have resulted in an overestimation of the role of Dominican Republic, Haiti and Trinidad and Tobago as source of BCAR lineages in the region. The use of more geographically balanced HIV-1 subtype B Caribbean datasets will be of paramount importance to obtain more precise estimates of the contribution of each Caribbean island in the regional dissemination of nonpandemic subtype B strains.
Several country-specific BCAR clades were detected in Argentina, Brazil, Guyana, Mexico, Panama and Venezuela, suggesting that despite their overall low prevalence, nonpandemic subtype B lineages have been disseminated locally in several Latin American countries. Estimation of the TMRCA of those country-specific BCAR clades further suggests that BCAR lineages started to be disseminated from the Caribbean into Latin America between the early 1970s and the early 1980s. This time scale coincides with the global dissemination of the BPANDEMIC clade from the United States  and with the estimated origin of several BPANDEMIC lineages in Latin America [7,28]. Although the BPANDEMIC and the BCAR clades probably arrived at the same time in Latin America, the BPANDEMIC strain was able to ignite much larger outbreaks and infected a much larger number of individuals than any BCAR strain in most of the countries analyzed.
The different epidemic outcomes of the BPANDEMIC and BCAR lineages in Latin America could be related to virological and/or sociological factors. Notably, the highest HIV prevalence rates (>1%) in Latin America and the Caribbean were detected among countries with a high proportion (≥50%) of BCAR clades like Haiti, Bahamas, Guyana, Jamaica and Trinidad and Tobago , thus arguing against the hypothesis of a low epidemic potential of BCAR lineages. Transmission route is clearly an important factor shaping the HIV dissemination dynamics and major differences in the epidemic outcome of distinct subtype B clades may have appeared as a consequence of differences in the underlying transmission networks. We suggest that in most Latin American countries the BPANDEMIC strain was introduced and initially disseminated within highly connected networks of MSM and injecting drug users, whereas the BCAR clades were mainly disseminated through heterosexual networks with lower rates of partner exchanges, which may explain the more successful dissemination of the BPANDEMIC lineage.
The remarkably successful dissemination of BCAR clades in some northern countries of South America including French Guyana, Suriname and Guyana, probably reflects the high mobility of people between these countries and the Caribbean islands . This is facilitated not only by the geographical proximity of those South American countries to the Caribbean islands, but also by cultural, linguistic and socioeconomic ties. Suriname and Guyana are members of the Caribbean common market, an organization of 15 Caribbean nations and dependencies that also includes Bahamas, Belize, Haiti, Jamaica, Trinidad and Tobago and several other Lesser Antilles islands. The Caribbean common market not only promotes economic integration, but also facilitates the free movement of individuals for tourism or labor among countries. It is noted that a significant proportion (10%) of immigrants residing in Trinidad and Tobago are from Guyana , which may explain the epidemiological link observed between nonpandemic BCAR-TT and BCAR-GY clades circulating in Trinidad and Tobago and Guyana, respectively.
The higher frequency of BCAR clades in Colombia, Panama and Venezuela (4–9% of subtype B infections) when compared with other Latin American countries (<2% of subtype B infections) also probably reflects a more frequent population mobility as a consequence of greater geographical proximity and historical links. It is interesting to note that the first reported Panamanian AIDS case was a Haitian woman diagnosed in September 1984 , which supports a longstanding presence of viruses of Caribbean origin in Panama. This country is also an important commercial hub because of the presence of the Panama Canal that promotes transit of people and goods. Junqueira et al. previously noted that a boom in oil production in Venezuela attracted immigrants from several countries in the region between 1970 and 1980, including people from Trinidad and Tobago and the Dominican Republic, which may have promoted the introduction of BCAR strains into Venezuela during that time. Furthermore, Colombia and Venezuela have been pointed out as the most important source countries in South America for tourists and labor migrants (including female sex workers) to many Caribbean islands (particularly in the Netherlands Antilles) .
In summary, this study demonstrates that several nonpandemic HIV-1 BCAR strains have been disseminated from the Caribbean into Latin America since the early 1970s. The BCAR strains reached nearly all countries from Latin America here analyzed and in some of them were spread locally, establishing secondary outbreaks. Despite the early and widespread dissemination of BCAR strains in the continent, HIV-1 subtype B epidemics in most Latin American countries were mainly driven by the BPANDEMIC clade that accounts for most (>90%) of current HIV-1 subtype B infections in the region. The only exceptions were Suriname, French Guyana and probably Guyana, where both BPANDEMIC and BCAR clades seem to circulate at roughly similar prevalence as observed in many Caribbean islands. Intraregional population mobility combined with chance founder events in populations with high rates of partner exchange were probably the major forces driving the actual distribution of the different subtype B strains in the Americas.
The authors wish to thank Dr Vera Bongertz for critical review of the manuscript.
Contributions: The study was conceived and designed by G.B. Data acquisition was performed by M.C. All authors contributed to the data analysis and final version of the review.
Source of funding: This work was supported by Public Health Service grants E-26/110.439/2014 from the FAPERJ and 472896/2012-1 from the CNPq. M.C. was funded by a fellowship from Instituto Oswaldo Cruz-FIOCRUZ.
Conflicts of interest
There are no conflicts of interest.
3. Hemelaar J, Gouws E, Ghys PD, Osmanov S. Global trends in molecular epidemiology of HIV-1 during 2000–2007
4. Gilbert MT, Rambaut A, Wlasiuk G, Spira TJ, Pitchenik AE, Worobey M. The emergence of HIV/AIDS in the Americas and beyond
. Proc Natl Acad Sci U S A
5. Cabello M, Mendoza Y, Bello G. Spatiotemporal dynamics of dissemination of non-pandemic HIV-1 subtype B clades in the Caribbean region
. PLoS One
6. Junqueira DM, de Medeiros RM, Matte MC, Araujo LA, Chies JA, Ashton-Prolla P, et al. Reviewing the history of HIV-1: spread of subtype B in the Americas
. PLoS One
7. Mendoza Y, Martinez AA, Castillo Mewa J, Gonzalez C, Garcia-Morales C, Avila-Rios S, et al. Human immunodeficiency virus type 1 (HIV-1) subtype B epidemic in Panama is mainly driven by dissemination of country-specific clades
. PLoS One
8. de Oliveira T, Deforche K, Cassol S, Salminen M, Paraskevis D, Seebregts C, et al. An automated genotyping system for analysis of HIV-1 and other microbial sequences
9. Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG. The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools
. Nucleic Acids Res
10. Posada D. jModelTest: phylogenetic model averaging
. Mol Biol Evol
11. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O. New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0
. Syst Biol
12. Guindon S, Lethiec F, Duroux P, Gascuel O. PHYML online: a web server for fast maximum likelihood-based phylogenetic inference
. Nucleic Acids Res
13. Anisimova M, Gascuel O. Approximate likelihood-ratio test for branches: a fast, accurate, and powerful alternative
. Syst Biol
14. Rambaut A. FigTree v1.4: Tree Figure Drawing Tool. http://tree.bio.ed.ac.uk/software/figtree/2009
[accessed 5 October 2014].
15. Drummond AJ, Nicholls GK, Rodrigo AG, Solomon W. Estimating mutation parameters, population history and genealogy simultaneously from temporally spaced sequence data
16. Drummond AJ, Rambaut A. BEAST: Bayesian evolutionary analysis by sampling trees
. BMC Evol Biol
17. Suchard MA, Rambaut A. Many-core algorithms for statistical phylogenetics
18. Drummond AJ, Ho SY, Phillips MJ, Rambaut A. Relaxed phylogenetics and dating with confidence
. PLoS Biol
19. Drummond AJ, Rambaut A, Shapiro B, Pybus OG. Bayesian coalescent inference of past population dynamics from molecular sequences
. Mol Biol Evol
20. Hue S, Pillay D, Clewley JP, Pybus OG. Genetic analysis reveals the complex structure of HIV-1 transmission within defined risk groups
. Proc Natl Acad Sci U S A
21. Zehender G, Ebranati E, Lai A, Santoro MM, Alteri C, Giuliani M, et al. Population dynamics of HIV-1 subtype B in a cohort of men-having-sex-with-men in Rome, Italy
. J Acquir Immune Defic Syndr
22. Chen JH, Wong KH, Chan KC, To SW, Chen Z, Yam WC. Phylodynamics of HIV-1 subtype B among the men-having-sex-with-men (MSM) population in Hong Kong
. PLoS One
23. Lemey P, Rambaut A, Drummond AJ, Suchard MA. Bayesian phylogeography finds its roots
. PLoS Comput Biol
24. Ferreira MAR, Suchard MA. Bayesian analysis of elapsed times in continuous-time Markov chains
. Can J Stat
25. Rambaut A, Drummond A. Tracer v1.6. http://tree.bio.ed.ac.uk/software/tracer/2007
[accessed 5 October 2014].
26. Bielejec F, Rambaut A, Suchard MA, Lemey P. SPREAD: spatial phylogenetic reconstruction of evolutionary dynamics
28. Murillo W, Veras N, Prosperi M, de Rivera IL, Paz-Bailey G, Morales-Miranda S, et al. A single early introduction of HIV-1 subtype B into Central America accounts for most current cases
. J Virol
29. Borland R, Faas L, Marshall D, McLean R, Schroen M, Smit M, et al. HIV/AIDS and mobile populations in the Caribbean: a baseline assessment
. International Organization for Migration; 2004 Available from: http://www.iom.int/jahia/webdav/site/myjahiasite/shared/shared/mainsite/published_docs/books/hiv_mobile_carribbean.pdf
. [accessed 5 October 2014].
30. de Ycaza MM, Rios V, Miranda E, Narvaez E, Sanchez G. [Acquired immune deficiency syndrome. First confirmed case in Panama]
. Rev Med Panama