Consistent phylogenetic patterns recover HIV epidemiologic relationships and reveal common transmission of multiple variants in known transmission pairs

The growth of HIV sequence databases resulting from drug resistance testing has motivated efforts using phylogenetic methods to assess how HIV spreads(1–4). Such inference is potentially both powerful and useful for tracking the epidemiology of HIV and allocation of resources to prevention campaigns. We recently used simulation and a small number of illustrative cases to show that certain phylogenetic patterns are associated with different types of epidemiological linkage(5); our original approach was later generalized for large NGS datasets and implemented as a free computational pipeline(6). Previous work has claimed that direction and directness of transmission could not be established from phylogeny because one could not be sure there were no intervening or missing links involved(7–9). Here, we address this issue by investigating phylogenetic patterns from 272 previously identified HIV transmission chains with 955 transmission pairs representing diverse geography, risk groups, subtypes, and genomic regions. These HIV transmissions had known linkage based on epidemiological information such as partner studies, mother-to-child transmission, pairs identified by contact tracing, and criminal cases. We show that the resulting phylogeny inferred from real HIV genetic sequences indeed reveals distinct patterns associated with direct transmission contra transmissions from a common source. Thus, our results establish how to interpret phylogenetic trees based on HIV sequences when tracking who-infected-whom, when, and how genetic information can be used for improved tracking of HIV spread. We also investigate limitations that stem from limited sampling and genetic time-trends in the donor and recipient HIV populations.