The human immunodeficiency virus (HIV) has a genome that is rich in adenine, and its rapid evolution shows an observed bias of guanine (G) to adenine (A) mutations. Two mechanisms have been proposed to explain these properties: (1) an imbalance in dNTP pool concentrations which drives the misincorporation process during reverse transcription, and (2) cytidine deamination by the APOBEC3G/3F restriction factor, causing G to A mutations most notably in specific dinucleotide contexts. Although crucial to understanding HIV evolution, current estimates on misincorporation bias during the replication cycle are based on scarce in vitro measurements. In this work, HIV partial pol sequences obtained for drug resistance testing purposes are analyzed using likelihood methods to estimate various models of HIV misincorporation bias in vivo. The technique is robust to selection on the amino acid sequence and selection against CpG dinucleotides. A model where misincorporations are explained only by an imbalance in dNTP pool concentrations, together with a preference for transitions versus transversions, explained 98% (95% confidence interval [C.I.] 93-100) of the observed variation in freely estimated misincorporation rates. Although dinucleotide context was responsible for variation in misincorporation probabilities, this variation was not specific for G to A mutations implying that the footprint of APOBEC3G/3F editing could not be detected. These results indicate that an imbalance in dNTP pool concentrations explains most of the bias in HIV nucleotide misincorporations, while the effect of editing by APOBEC3G/3F on HIV evolution, based on its dinucleotide specificity, could not be observed in this study.
- sequence analysis
- IMMUNODEFICIENCY-VIRUS TYPE-1