Background
APOBEC3-mediated G-to-A hypermutation in HIV-1 proviral DNA complicates drug resistance interpretation. Its extent varies across sequences. Available tools can detect but cannot differentiate mutations driven by drug pressure from APOBEC3 editing. Thresholds and criteria for excluding hypermutated sequences are undefined, limiting interpretation and clinical decision-making.
Methods
We sequenced near-full-length (nFL) HIV-1 proviral genomes from virologically suppressed persons with HIV in Rhode Island. Hypermutations were quantified with Hypermut v3 in nFL genomes and by gene (protease-PR, reverse transcriptase-RT, integrase-INT). We examined codons of 18 drug resistance mutations (DRMs) that can arise from APOBEC3 editing (PR D30N, M46I, G48S, G73S; RT D67N, E138K, M184I, G190E, G190S, M230I; INT G118R, E138K, G140R, G140S, G163K, G163R, D232N, R263K); for each codon, assessed DRM presence and hypermutation at the causative nucleotides; and evaluated how excluding sequences with ≥3%, ≥2%, or ≥1% hypermutations affected DRM detection.
Results
Proviral DNA sequences were obtained from 422 persons: 80% male, median age 34 years, 57% men who have sex with men, 93% subtype B, HIV diagnosis 1983-2022. At least one hypermutated site in nFL genomes was present in every participant (84% PR, 100% RT, 86% INT), 21% nFL sequences (14% PR, 20% RT, 16% INT) were identified as hypermutated, and 4% (16/422), 9% (37/422) and 25% (106/422) of nFL sequences exceeded thresholds of ≥3%, ≥2% and ≥1% hypermutation. Sequence distribution by proportion of hypermutated sites was right-skewed across all genes. At least one hypermutated DRM was present in 23% (96/422; 17% PR, 21% RT, 14% INT). Across the 18 DRMs, hypermutation frequencies peaked at 61% (PR), 46% (RT), and 57% (INT). Excluding nFL sequences with ≥3%, ≥2%, and ≥1% hypermutation sites reduced these maxima to 16%, 9%, and 7% (PR); 25%, 11%, and 7% (RT); and 32%, 20%, and 2% (INT), respectively. Reductions varied by DRM and were substantial at the ≥2% threshold, eliminating several hypermutated DRMs and markedly reducing others.
Conclusions
Hypermutation, overall and at DRM sites, is common in HIV-1 proviral DNA. Excluding sequences with ≥2% hypermutated sites, rather than ignoring all, substantially reduced detection of DRMs prone to APOBEC3 editing. Whether selective versus complete filtering of hypermutated sequences has a beneficial clinical impact warrants further study.