Fig. 3. Schematic diagrams of phylogenetic patterns expected under various hypotheses for the origin and spread of subtype B. (a) If the virus reached Haiti first, then Haitian HIV-1 sequences are expected to branch off from the root part of the subtype B subtree before sequences from elsewhere. Alternatively, (b) the Haitian epidemic could have been imported from the U.S. or elsewhere; (c) both the U.S. and Haitian epidemics could have begun effectively simultaneously then remained distinct; or (d) high levels of migration could have obscured where the virus arrived first.

Fig. 4. The 50% majority rule consensus tree from the Bayesian MCMC (MrBayes) analysis. Posterior probabilities for each node are shown. The branch lengths represent the mean value observed for that branch among the postburnin sampled trees. This is the full version of the abridged tree depicted in Fig. 1, which collapsed the Trinidad and Tobago clade and dominant clade into blue and yellow triangles, respectively. The same color coding is used here. The archival sequences are labeled in larger bold text. All of the other strains are listed by their LANL designations ("subtype.country.year.strain"). The country codes are as follows: AR = Argentina; AU = Australia; BR = Brazil; CA = Canada; CD = Democratic Republic of the Congo; CO = Colombia; DE = Germany; EC = Ecuador; ES = Estonia; FR = France; GA = Gabon; GB = United Kingdom; HT = Haiti; JP = Japan; KR = South Korea; NL = Netherlands; TH = Thailand; TT = Trinidad and Tobago; UG = Uganda; US = United States of America; and ZA = South Africa. The position of the Haitian archival sequence H6 is unresolved and is consistent with it being basal to the pandemic clade sequences as indicated by the relaxed molecular clock analysis (Fig. 2).

Fig. 5. Corroborative analyses. (a) MrBayes results using the env gene sequences with the African subtype D outgroup sequences replaced by a CRF01 (A/E) outgroup sequence from Thailand. The results are virtually identical to those using African subtype D sequences. Other outgroups (subtype C from India; subtype A from Kenya) produced similar results (not shown). These analyses indicate that the "Haiti first" inference is not dependent on the outgroup used and is not an artifact of convergent evolution between Haitian subtype B and African subtype D sequences. (b) The consensus tree from the MrBayes analysis of partial gag gene sequences. As with env, the Haitian sequences occupy the basal positions in the gag phylogeny, indicating a Haitian origin of subtype B. The pandemic clade forms a single monophyletic cluster, as with env. For the gag gene analysis P_{nonHaitian-origin} = 0.007 and P_{Haitian-origin} = 0.993.