U.S. Human immunodeficiency virus type 1 epidemic: date of origin, population history, and characterization of early strains

Abstract

Human immunodeficiency virus (HIV) type 1 subtype B sequences (whole envelope and the p17 region of gag) were obtained from peripheral blood mononuclear cell samples collected in 1981 from seven HIV-infected U.S. individuals and in 1982 from one infected Canadian resident. Phylogenetic and nucleotide distance analyses were performed by using database sequences representing North American strains collected from 1978 to 1995. The estimated phylogeny was starlike, with early strains represented on different lineages. When sequences were grouped by years of collection, nucleotide distance comparisons demonstrated an increase in diversity over time and indicated that contemporary strains are more closely related to early epidemic strains than to each other. Using a recently developed likelihood ratio reduction procedure, the date of origin of the U.S. epidemic was estimated to be 1968 +/- 1.4 years. A coalescent approach was also used to estimate the population history of the U.S. subtype B epidemic. Our analyses provide new information that implies an exponential growth rate from the beginning of the U.S. HIV epidemic. The dating results suggest a U.S. introduction date (or date of divergence from the most recent common ancestor) that precedes the date of the earliest known AIDS cases in the late 1970s. Furthermore, the estimated epidemic growth curve shows a period of exponential growth that preceded most of the early documented cases and also indicates a leveling of prevalence rates in the recent past.

FIG. 1.

Maximum-likelihood tree of gp160 sequences, depicting the relationship of the early U.S. sequences (bold) to North American subtype B strains collected in later years. The numbers preceding the sequence names refer to the collection year. Strain D.85Z2 (subtype D) is the outgroup, and ConB95 is the consensus sequence for subtype B (HIV database, LANL).

FIG. 2.

Nucleotide distance comparisons within sequences grouped by year of collection, by gene region (a). (b) Between-group (1981-1982 group versus other groups) distances are shown. Numbers within columns are mean distances, and error bars are ± standard errors of the mean.

FIG. 3.

Chart depiction of data from the SSCD analysis, showing estimated dates for the two internal gp160 tree nodes encompassing U.S. strains only (upper) and both U.S. and Haiti strains (lower). Each symbol within the lines represents the respective calculation performed on alignments with a progressive number of sites stripped (x axis). The blue line (series 1) represents the LRT test statistic 2 X (DR-SRDT) calculated for the alignments with sites removed based on their contribution to the overall LRT; the yellow line represents the LRT statistic calculated for alignments where variable sites were progressively removed at random (testing the power of the procedure). The 95% confidence limit according to a χ² distribution for which higher LRT values signify a significant rejection of the molecular clock is drawn as a horizontal line. The more rigorous 99% site-specific cutoff, signifying the number of sites needing stripping (133) for molecular clock rejection, is shown as a vertical line. The chart simultaneously plots the LRT values (blue and yellow lines) versus the number of sites removed, together with the date of the internal node (upper and lower) corresponding to the number of sites removed.

FIG. 4.

Population growth history analysis of U.S. subtype B gp160 sequences using the TipDate and Genie programs. (a) TipDate-generated maximum-likelihood tree (SRDT model). (b) Genie analysis depicting coalescent estimates of effective population size (number of infections). The stepwise plot (shaded line) is the nonparametric estimate. The smooth plot (bold curve) is the maximum-likelihood parametric estimate obtained under the logistic growth model.