Phylogenetic analysis consistent with a clinical history of sexual transmission of HIV-1 from a single donor reveals transmission of highly distinct variants

Abstract

Background: To combat the pandemic of human immunodeficiency virus 1 (HIV-1), a successful vaccine will need to cope with the variability of transmissible viruses. Human hosts infected with HIV-1 potentially harbour many viral variants but very little is known about viruses that are likely to be transmitted, or even if there are viral characteristics that predict enhanced transmission in vivo. We show for the first time that genetic divergence consistent with a single transmission event in vivo can represent several years of pre-transmission evolution.

Results: We describe a highly unusual case consistent with a single donor transmitting highly related but distinct HIV-1 variants to two individuals on the same evening. We confirm that the clustering of viral genetic sequences, present within each recipient, is consistent with the history of a single donor across the viral env, gag and pol genes by maximum likelihood and bayesian Markov Chain Monte Carlo based phylogenetic analyses. Based on an uncorrelated, lognormal relaxed clock of env gene evolution calibrated with other datasets, the time since the most recent common ancestor is estimated as 2.86 years prior to transmission (95% confidence interval 1.28 to 4.54 years).

Conclusion: Our results show that an effective design for a preventative vaccine will need to anticipate extensive HIV-1 diversity within an individual donor as well as diversity at the population level.

Figure 1

Clinical data for P1 compared with that of P2. The a. CD4+ counts (/mm³) and b. log viral loads (copies/ml) for P1 (blue) and P2 (red) are shown. P1 and P2 were exposed to the same third party on day 0. P1 remained off therapy for 930 days post-exposure whilst P2 progressed more rapidly and commenced HAART 249 days post-exposure. Plasma for baseline sequencing was collected on day 63 but the CD4+ count or VL were not recorded. At day 245, P1 was diagnosed with acute HCV infection and had evidence of super-infection in plasma collected at day 235, having been exposed to a fourth person after day 63.

Figure 2

Trees generated for phylogenetic cluster analysis. Phylogenetics cluster analysis was carried out using day 63 viral sequences from P1 and P2. Zoomed-in images of trees are shown in Figure 2 for the env fragment in a. and b., the gag fragment in c. and d., and the pol fragment in e. and f.. Results from two different methods of cluster analysis are shown for each fragment: ML (PhyML) trees in a., c., and e., and Bayesian MCMC based consensus trees in b., d., and f.. Terminal nodes represent sequences sampled from P1 (blue circles) or P2 (red circles), as well as reference sequences. Env sequences for P1 and P2 were sampled by SGA and represent gap-stripped alignments 1305 nucleotides in length. Gag and pol fragments were sampled by bacterial cloning. The full tree images can be viewed in Additional Figures 1 and 2. All scale bars show 0.05, equivalent to 5% divergence. ML bootstrap values or Bayesian MCMC based posterior probabilities for the clustering of P1 and P2 are given as percentages next to the common ancestor node.

Figure 3

Highlighter plot of env gp120 nucleotide sequences. Full-length env gp120 sequences from day 63 were sampled by SGA. The Highlighter plot shows gaps in grey and nucleotide substitutions (A = green, T = red, G = orange, C = light blue), revealing difficult-to-align regions. The master sequence against which all other sequences are compared is the majority-rule P1 consensus sequence at day 63, shown as the top sequence.

Figure 4

Relaxed-clock tree for env. Between-host divergence, in terms of pre-transmission evolution, was quantified as the estimated tMRCA using a Bayesian MCMC based approach. Env C2V5 fragment sequences from P1 and P2, sampled at day 63 by SGA, were calibrated against within-host divergence since the estimated time since transmission as well as the mean rate of substitution from the reference dataset.

Figure 5

Comparison of PNLGSs in inferred env gp120 amino acid sequences. Full-length gp120 amino acid sequences, inferred from day 63 SGA nucleotide sequences, are shown. The proportion of P1 sequences with PNLGS at a particular position are shown as a 'positive' blue bar and the proportion of P2 sequences with a PNLGS is shown as a 'negative' red bar. Positions where 100% of sequences have and PNLGS in both P1 and P2 are indicated by small stars.