Study Cohort and HIV Sequences

HIV sequence data was obtained from samples collected in the HPTN 052 clinical trial [34]. This trial enrolled HIV-serodiscordant pairs and assessed the impact of early antiretroviral treatment (ART) on HIV transmission. A full description of the study protocol and institutional review board oversight is available in the original publication [34]. Genetic linkage of most index-partner pairs was based on phylogenetic and Bayesian analysis of HIV pol sequences obtained by bulk Sanger sequencing; in selected cases, linkage was confirmed by neighbor-joining tree analysis of NGS, using the 454-Roche Biotechnology platform [32, 33]. This study only included pairs in which both the index patient and their partner were infected with a single HIV strain.

Samples were obtained from partners at the visit when seroconversion was documented or at the next study visit (hereafter, “SC samples”; median time of collection, 91 days after the last visit during which an HIV-negative test result was obtained [range, 84–588 days); partners were not followed in the HPTN 052 trial after HIV infection was confirmed. Paired SC samples were collected from index patients at the visit closest to the visit when their partner’s seroconversion was documented (for 9 index patients, their SC sample was collected >90 days before seroconversion detection in the partner; for 14, their SC sample was collected 0–90 days before seroconversion detection; and for 10, their SC sample was collected 1–90 days after seroconversion detection; Figure 1). For 31 index-partner pairs, additional samples collected earlier from index patients (hereafter, “index samples”; median time of collection, 362 days before collection of the index patients’ SC samples [range, 84–1174 days]) were also available and were analyzed separately.

Time between collection of samples from human immunodeficiency virus (HIV)–infected index patients and their partners who acquired genetically linked HIV infection during the study. Samples were collected from partners and index patients close to the time when the partner seroconverted (hereafter, “SC samples”); for 31 pairs, samples collected from the index patient at an earlier time point (hereafter, “early index samples”) were also available. Positive values indicate that the SC or early index sample was collected before the partner sample; negative values indicate that the index patient’s SC or early index sample was collected after the partner’s SC sample. The identifier for each index patient and partner pair is shown on the x-axis. Pairs were created on the basis of the timing of collection of SC samples from partners and index patients. Two pairs did not have an early index sample available for analysis.

A total of 450 336 NGS-derived reads from env (nucleotides 7941–8264 relative to HXB2) were obtained from plasma samples from the 33 index-partner pairs [34, 35]. From these reads, 9051 consensus sequences (hereafter, “sequences”) were generated using GS Amplicon Variant Analyzer, version 2.5 (Roche); each sequence represented a cluster of ≥10 individual reads. Each sample had an average of 91 sequences representing 4503 reads. Sequence alignments were manually edited by codon, using AliView [36], and frameshift insertions were removed. Sequences were subtyped using REGA (available at: http://dbpartners.stanford.edu:8080/RegaSubtyping/stanford-hiv/typingtool). Reference sequences were obtained from the Los Alamos HIV Database (available at: https://www.hiv.lanl.gov/).