Human APOBEC3 induced mutation of human immunodeficiency virus type-1 contributes to adaptation and evolution in natural infection

Abstract

Human APOBEC3 proteins are cytidine deaminases that contribute broadly to innate immunity through the control of exogenous retrovirus replication and endogenous retroelement retrotransposition. As an intrinsic antiretroviral defense mechanism, APOBEC3 proteins induce extensive guanosine-to-adenosine (G-to-A) mutagenesis and inhibit synthesis of nascent human immunodeficiency virus-type 1 (HIV-1) cDNA. Human APOBEC3 proteins have additionally been proposed to induce infrequent, potentially non-lethal G-to-A mutations that make subtle contributions to sequence diversification of the viral genome and adaptation though acquisition of beneficial mutations. Using single-cycle HIV-1 infections in culture and highly parallel DNA sequencing, we defined trinucleotide contexts of the edited sites for APOBEC3D, APOBEC3F, APOBEC3G, and APOBEC3H. We then compared these APOBEC3 editing contexts with the patterns of G-to-A mutations in HIV-1 DNA in cells obtained sequentially from ten patients with primary HIV-1 infection. Viral substitutions were highest in the preferred trinucleotide contexts of the edited sites for the APOBEC3 deaminases. Consistent with the effects of immune selection, amino acid changes accumulated at the APOBEC3 editing contexts located within human leukocyte antigen (HLA)-appropriate epitopes that are known or predicted to enable peptide binding. Thus, APOBEC3 activity may induce mutations that influence the genetic diversity and adaptation of the HIV-1 population in natural infection.

Figure 1. Site-specific editing frequencies in infected cells from single-cycle APOBEC3 titration transfection experiments.

The single-cycle substitution rate for HIV-1 in the absence of human APOBEC3 was 8.6×10⁻⁴ mutations per nucleotide, whereas the mean single-cycle substitution rate for HIV-1 in the presence of human APOBEC3 ranged from about 1×10⁻³ to 2×10⁻² per nucleotide substitution (A). The frequency of substitutions increased significantly in the region of the Gag gene of HIV-1 we sequenced in accord with increasing concentrations of the APOBEC3 proteins. The maximum single-cycle substitution rate for HIV-1 was 2×10⁻³ substitutions per site in the presence of APOBEC3D, 1.4×10⁻² substitutions per site in the presence of APOBEC3F, 2.7×10⁻² substitutions per site in the presence of APOBEC3G and 1.1×10⁻⁴ substitutions per site in the presence of APOBEC3H. The concentration of APOBEC3 at which we observed half of the estimated maximum substitution rate was 0.02 for APOBEC3D, 2.09 for APOBEC3F, 0.23 for APOBEC3G, and 0.22 µg for APOBEC3H. The single-cycle substitution rate for each mutation type (transition = Ts or transversion = Tv) of HIV-1 in the titration transfection experiments differed by 1.2 order of magnitude (B). For each of the human APOBEC3 proteins, we show the positions in the Gag gene of HIV-1 where the frequency of G-to-A mutation increased with increasing amounts of human APOBEC3 protein (Spearman rank correlation coefficient, P-value<0.05) (C). The G-to-A mutations are shown in a number of contiguous nucleotide sequence editing contexts. We used a sliding window to deduce the base frequency of G-to-A mutations (in each contiguous nucleotide context of the edited sites for each APOBEC3 protein (APOBEC3D, APOBEC3F, APOBEC3G, and APOBEC3H) using the total G-to-A mutation frequency at increasing concentration. Positions with a non-significant increase in G-to-A mutations were excluded from the calculations.

Figure 2. Positions of G-to-A mutations within the APOBEC3 trinucleotide context of the edited sites are shown in the HIV-1 genome.

A heat map of the slope of the logistic regression of deamination frequency with increasing amount of APOBEC3 protein expression is shown at left (Spearman correlation coefficient, P<0.05) (A). The nucleic acid position in HXB2 (GenBank Accession number K03455) is to the right of the G-to-A position. A heat map of G-to-A frequency in APOBEC3 trinucleotide context of the edited sites with increasing amounts of expression of each APOBEC3 protein is shown in the middle panel. The location of the guanosine in the trinucleotide context of the edited sites in the segment of the targeted sequence is shown at right. Mutation frequency comparison with and without guanosines within the defined APOBEC3 trinucleotide context of the edited sites per each enzyme is shown (B). Upon removal of guanosine in the APOBEC3 trinucleotide context of the edited sites, substitution rate lowered to 9.77×10⁻⁴ mutations per nucleotide substitution.

Figure 3. Maximum-likelihood tree for HIV-1 sequences from the patients studied.

Shown are the maximum-likelihood phylogenies of viral nucleotide sequences from the Gag and Vif genes of HIV-1 derived from the ten patients at two time points during infection rooted with the estimated ancestral subtype B sequence . The sequences from each patient are individually colored and labeled accordingly. The maximum-likelihood tree shows viral sequences that belong to a transmission pair (S004 and S005) and a dually infected patient (S007). Branch lengths indicate the number of nucleotide substitutions per site. The scale bar at the bottom refers to the degree of sequence mismatch. Maximum-likelihood bootstraps are presented below each branch.

Figure 4. Overlap of the APOBEC3 edited HIV-1 genomes in cell culture experiments and in patients.

A 96-base pair segment that overlaps a region sequenced in the Gag gene of HIV-1 in cell culture experiments and from patients is shown. APOBEC3 trinucleotide context analysis identified 26 sites that were significantly enriched for G-to-A mutations in more than one patient. The right y-axis corresponds to the number of patients and the left y-axis represents the slope of the logistic regression corresponding to the cell culture G-to-A mutation per concentration where significant correlation between these two values was observed (Spearman's test, P<0.05). The positions are numbered according to HXB2 and the alignment shows the HIV-1 pIIIB/Δvif sequence and the consensus sequence for each patient at the early time point. The blue line shows the number of patients with HIV-1 genomes that have significantly increased numbers of G-to-A mutations in positions that conform to the trinucleotide context of the edited sites and accrue over time (Fisher's exact test P<0.01). The light blue, red, green, and purple areas show the slopes with significant correlation (Spearman's test, P<0.05) for each G-to-A mutation with increasing concentrations of the APOBEC3D, APOBEC3F, APOBEC3G or APOBEC3H proteins, respectively.

Figure 5. Pairwise distance analysis.

Removal of the guanosine from the APOBEC3 trinucleotide context of the edited sites produced a significant decrease in the observed pairwise genetic distances of the sequences compared to the early time point consensus sequence, expressed as number of substitutions per site (P-values<0.05 are considered statistically significant).