APOBEC3G versus reverse transcriptase in the generation of HIV-1 drug-resistance mutations

HIV-1 is known for its high adaptive capacity, which is most seriously demonstrated by its ability to escape from potent antiviral drugs in treated HIV-infected individuals. Viral escape occurs by the selection of drug-resistant variants with single or multiple amino acid substitutions in the targeted enzymes reverse transcriptase (RT) or protease (PR), and more recently the evelope protein (Env). HIV-1 genetic variation is primarily caused by the error-prone RT enzyme, which incorporates approximately one mismatch nucleotide per genome per replication cycle [1]. The RT-induced mutations show a distinct pattern, with transitions being more frequently observed than transversions, and G-to-A changes being most prevalent [1,2]. It has been argued that this mutational bias has shaped the HIV-1 genome and in fact all lentiviral genomes, which are indeed extremely A-rich [3,4]. We decided to perform a follow-up analysis on this subject for two specific reasons. First, many more antiviral agents have entered the clinic in recent years, coinciding with the description of new drug-resistance mutations in the targeted viral genes [5]. Second, it was recently demonstrated that a cellular enzyme, the CEM15 APOBEC3G cytidine deaminase, can also cause G-to-A changes in the viral genome [6–9]. This enzyme converts cytosine to uracil in nascently reverse transcribed HIV-1 genomes, which results in G-to-A changes in the viral plus-strand RNA genome. An APOBEC3G-introduced variation within the HIV-1 genome could increase the size of the viral quasispecies, and thereby add to HIV-1 pathogenesis [7]. The impact of APOBEC3G versus RT on HIV-1 evolution, and specifically the appearance of drug-resistant virus variants, has not been addressed so far. Here, we present an initial survey of the most common resistance-associated amino acid changes and the underlying codon changes. We used the Los Alamos HIV-1 drug resistance database for this survey (http://resdb.lanl.gov/Resist_DB/default.htm). Because certain codon changes have been reported many times, we scored each codon change only once. Alternative nucleotide substitutions within a single codon in different patients were counted as independent mutational events. In case two or even three nucleotide changes were selected within a codon, we scored all individual nucleotide substitutions. The mutational survey was performed for the PR, RT and Env protein (Table 1). We also calculated the total number of nucleotide substitutions, which was used to make a ranking order. The most frequent substitution was the G-to-A change, which accounts for 21% of the drug-resistance mutations. Other transition-type of mutations were also prevalent (A to G: 17%, T to C: 9%), but the C-to-T transition was seen at a surprisingly low frequency (7%). Great differences were also observed among the transversions, ranging from 8% for T to A to 3% for C to G.Table 1: Resistance conferring mutations in the protease, reverse transcriptase, and Env genes.Given the extreme bias in the nucleotide composition of the HIV-1 genome (36% A, 24% G, 22% T, and 18% C), it seems appropriate to correct for this (last column of Table 1). The results highlight the preferential use of G-to-A changes (22%) in the evolution of drug-resistant HIV-1 variants. After this correction, all transitions cluster as the top group of most prevalent mutations. Changes to A are most prevalent among the transversions (T to A and C to A). This pattern, combined with the most frequent G-to-A mutation, may have shaped the A-rich genome [3,4]. Previously, these changes were thought to be caused exclusively by the error spectrum of the RT polymerase. High rates of G-to-A substitutions among treated patients in our study may result from alterations in the intracellular deoxyribonucleotide triphosphate pool that is caused by certain antiretroviral drugs [10]. However, the preponderance of G-to-A changes in drug-resistance mutations follows the general mutational bias seen in other HIV-1 studies [1,2]. Another complicating factor is that the HIV-1 mutational pattern and possibly the mutation rate may change as a result of certain amino acid substitutions in the RT enzyme itself. Many drug-resistance-related mutations within the RT enzyme have been found to modify polymerase fidelity [11–15]. We already mentioned the cellular APOBEC3G enzyme as an alternative cause of G-to-A hypermutation. Highly mutated HIV-1 genomes have been described with up to 60% of guanine residues substituted within a certain genome segment. Such massively edited genomes will contain many missense and nonsense codon changes, and critical regulatory sequences may be affected. APOBEC3G may be part of a cellular antiretroviral defence mechanism that inactivates the viral genomes. As a counter measure, HIV-1 encodes the Vif protein that neutralizes APOBEC3G action by direct protein–protein interaction, which prevents it from being incorporated into the virion [16]. We now propose that APOBEC3G-mediated hypermutation may be the driving force for rapid virus evolution if the number of G-to-A mutations is restricted. It is therefore too early to dismiss APOBEC3G, and possibly other APOBEC members, as contributors to limited G-to-A mutation and HIV-1 evolution instead of viral death. However, as a considerable number of G-to-A changes can be scored in APOBEC3G-negative cells in combination with Vif-positive viruses [1,2], we favour the idea that RT is more important in the evolution of drug-resistant variants. Sponsorship: This paper was partly supported by EU grant QLRT-2001-01311.