Human APOBEC3G-Mediated Editing Can Promote HIV-1 Sequence Diversification and Accelerate Adaptation to Selective Pressure

Abstract

Human apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like 3G (APOBEC3G, hereinafter referred to as A3G) is an innate virus restriction factor that inhibits human immunodeficiency virus type 1 (HIV-1) replication and induces excessive deamination of cytidine residues in nascent reverse transcripts. To test the hypothesis that this enzyme can also help generate viral sequence diversification and the evolution of beneficial viral variants, we have examined the impact of A3G on the acquisition of (−)2′,3′-dideoxy-3′-thiacytidine (3TC) resistance in vitro. That characteristic resistance mutations are rapidly fixed in the presence of A3G and 3TC suggests that A3G-mediated editing can be an important source of genetic variation on which natural selection acts to shape the structure of HIV-1 populations.

HIV-1 is constantly exposed to sequence changes because of a combination of rapid rates of viral replication, poor fidelity of the virus replicating enzymes (reverse transcriptase [RT] and RNA polymerase II), and recombination during simultaneous infection with multiple virus strains (5, 6, 17, 19, 21). In addition, mutation of HIV-1 by human APOBEC3G (A3G), a naturally expressed host nucleic acid-editing enzyme that deaminates deoxycytidine to deoxyuridine in the (mostly) minus strand of newly synthesized HIV-1 DNA, may further contribute to viral sequence change (1, 2, 9, 10, 15, 16, 31). Such transition mutations correspond to deoxyguanosine-to-deoxyadenosine (G-to-A) changes on viral plus strands.

Mutation caused by excessive A3G-mediated deamination of HIV-1 DNA typically has deleterious effects on virus replication through the loss of genetic integrity. The HIV-1-encoded Vif protein, which serves to induce the proteasomal degradation of A3G through recruitment to the cullin5-elongin BC-rbx2 ubiquitin ligase, efficiently counteracts the potent antiviral phenotype of this enzyme (18, 24-26, 30). It has been suggested that low levels of A3G activity that survive inhibition by Vif may promote infrequent mutations that, rather than preventing virus replication, may provide a source of genetic variation. Indeed, extensive sequencing of transmitted HIV-1 strains indicates that sequence variation bearing the hallmark of A3G-mediated mutation is commonplace (13, 22, 28), and results from experiments employing low levels of A3G expression confirm that modest G-to-A mutation frequencies, as opposed to hypermutation, can be recapitulated in vitro (23). Such sequence changes would have the potential to underlie advantageous alterations in the HIV-1 phenotype, such as escape from adaptive immunity or the development of drug resistance.

Acquisition of a drug-resistant phenotype requires the accumulation of mutations at specific amino acid residues. The use of the nucleoside analog RT inhibitor (−)2′,3′-dideoxy-3′-thiacytidine (3TC) imposes strong selection for a single amino acid substitution in the conserved Tyr-Met-Asp-Asp motif of the catalytic site of RT. High-level resistance is typically conferred by amino acid substitutions that change the methionine at position 184 to isoleucine (M184I), valine (M184V), or threonine (M184T) (3, 4). When 3TC is used as a single drug for people infected with HIV-1, drug-resistant virus outcompetes wild-type virus within a few weeks (3), suggesting that the required sequence changes commonly preexist in the virus population. Given that the M184I mutation is caused by a change in the previously reported consensus target for A3G-mediated editing (specifically, AUG-GAT to AUA-GAT) (1, 10, 29), we hypothesized that A3G has the potential to facilitate this particular adaptation to 3TC (9, 20). Using the presence of 3TC as a means of applying selection pressure, we therefore used modified CEM-SS cell lines that express or lack A3G (Fig. 1A) to determine whether this host nucleic acid-editing enzyme can contribute to wild-type HIV-1 sequence diversity and adaptation.

FIG. 1.

Immunoblot analysis of A3G protein in modified CEM-SS cells and relative replication efficiencies of HIV-1 grown in control and A3G-modified CEM-SS cells. (A) The cells were stably transduced with a retroviral vector encoding A3G or a corresponding control parental vector (8). Whole-cell lysates were analyzed by immunoblotting using primary antibodies to A3G (α-A3G) or heat shock protein 90 (HSP90)-α/β (Santa Cruz Biochemicals), followed by horseradish peroxidase-conjugated secondary antibodies and chemiluminescence (14). HSP90 was used as a loading control. The level of A3G protein in these cells closely matched that detected in primary human CD4⁺ T cells that had been activated with anti-CD3 and anti-CD28 antibodies (14). (B and C) Replication efficiencies of HIV-1 (copies of viral RNA per microliter of culture supernatant) in control or A3G-modified CEM-SS cell cultures with 3TC and without 3TC [3TC (−)]. Cells were infected with HIV-1_NL4-3 at a MOI of 0.01. HIV-1 gag in the cell culture supernatants from four independent experiments was measured by real-time quantitative reverse transcription-PCR. The horizontal axis indicates the days after infection with wild-type HIV-1_NL4-3. The vertical axis indicates the amount of viral RNA per microliter of CEM-SS cell culture supernatant. The means and standard deviations of the viral RNA values from four independent experiments are shown.

The control and A3G-modified CEM-SS cell cultures (5 ×10⁵ cells) were challenged with replication-competent HIV-1_NL4-3 at a low multiplicity of infection (MOI) of 0.01 and incubated for 4 h before being washed. The cultures were maintained continuously with all four combinations of A3G and the drug (i.e., no A3G and no 3TC, A3G with no 3TC, no A3G with 3TC, and A3G with 3TC). For these experiments, we used the 90% inhibitory concentration (IC₉₀) of 3TC, which was determined to be 1.48 μM in CEM-SS cells (data not shown). Virus replication was monitored as the amounts of HIV-1 gag RNA in the cell culture supernatants, as judged by real-time quantitative reverse transcription-PCR. Amplification of the HIV-1 gag gene was performed with forward primer HIVgag793f (5′-GGTGCGAGAGCGTCAGTATTAAG-3′, corresponding to positions 796 to 815 [796 → 815] according to HIV-1_HXB2 numbering), reverse primer HIVgag911r (5′-AGCTCCCTGCTTGCCCATA-3′, a 908 → 893 reverse complement), and probe HIVgag835p (5′-6-carboxyfluorescein [FAM]-TGGGAAAAAATTCGGTTAAGGCCAGGG-Black Hole Quencher [BHQ]-3′; 838 → 861). The relative amount of HIV-1 target RNA was normalized to the quantification cycle for a concentration calibrator by using an external standard curve of serial 10-fold dilutions of reference HIV-1 gag RNA. All cell culture infections were carried out in quadruplicate.

As expected, because the virus carries a functional vif gene, it replicated efficiently in the absence of 3TC irrespective of the presence of A3G (Fig. 1B) (24). In contrast, the virus grew very differently in the two cell types in the presence of 3TC (Fig. 1C). For virus grown in A3G-modified cells, HIV-1 RNA levels increased ∼10-, ∼10,000-, and ∼100,000-fold relative to levels for virus grown in control CEM-SS cells at 9, 13, and 16 days after infection, respectively. To confirm that viruses from A3G-modified cultures had acquired a 3TC-resistant phenotype, we tested the virus culture supernatant at day 16 for differences in IC₅₀ and IC₉₀ values of 3TC. Virus titrated to 1,000 50% tissue culture infective doses (TCID₅₀) per 10⁶ cells was used to infect CEM-SS cells containing serial dilutions of 3TC (range, 0 to 25.0 μM). The 3TC sensitivity values were determined from plots of HIV-1 p24 antigen verses 3TC concentrations (11). The fold change in 3TC sensitivity was determined by normalization to the IC₅₀ and IC₉₀ values for the HIV-1_NL4-3 stock. Table 1 shows that virus grown in A3G-modified CEM-SS cells in the presence of 3TC developed high-level resistance to the drug in three of four independent experiments but that viruses recovered from control cells retained typical sensitivity to 3TC. Low levels of measurable virus persisted in control CEM-SS cells in the presence of 3TC, due possibly to some residual input virus or limited viral replication.

TABLE 1.

Fold changes in 3TC susceptibility of virus isolated from control or A3G-modified CEM-SS cell cultures in the presence of 3TC at day 16^a

Expt	Condition		No. of days in cell culture	3TC susceptibility
	A3G modification	Presence of 3TC		IC50 (μM)	Fold change	IC90 (μM)	Fold change
Control	−	−	0	0.06	∼1	1.48	∼1
1	−	+	16	0.02	∼1	0.90	∼1
2	−	+	16	0.05	∼1	1.35	∼1
3	−	+	16	0.04	∼1	0.86	∼1
4	−	+	16	0.07	∼1	1.60	∼1
1	+	+	16	7.12	>100	1.56 × 105	>100,000
2	+	+	16	7.28	>100	435	>300
3	+	+	16	0.05	∼1	1.1	∼1
4	+	+	16	17.99	>300	4,067	>2,700

^aDose-response experiments were used to determine IC₅₀ and IC₉₀ values for viruses propagated in control and A3G-modified CEM-SS cells for 16 days. The first row shows the IC₅₀ and IC₉₀ values for the HIV-1_NL4-3 stock. Data from four replicate experiments are shown.

To determine the genetic basis of their resistance, we studied sequential viral isolates obtained from the control and A3G-modified cultures in the presence or absence of 3TC. We used reverse transcription-PCR amplification of culture supernatant viral RNA followed by bulk sequencing to screen for mixed bases at positions in the RT region of the pol gene known to be associated with 3TC drug resistance (Table 2). The sense primer (5′-CAGAGCAGACCAGAGCCAAC-3′; 2140 → 2156) and antisense primer (5′-CTGCTATTAAGTCTTTTGATGGGTC-3′; 3508 ← 3528) were designed to avoid the consensus A3G target motif. In the sequence chromatograms, mixed bases generally represent polymorphisms present in greater than 10% of the virus population. Comparative DNA sequence analysis showed a clear association between drug resistance and the presence of the G-to-A mutation that underlies the predicted M184I substitution or the A-to-G mutation that causes the predicted M184V substitution in virus grown in A3G-modified CEM-SS cells in the presence of 3TC (Tables 1 and 2). As expected, these mutations were not detected in analyses of residual viral sequences from CEM-SS cell cultures that lacked A3G (data not shown).

TABLE 2.

Analysis of population-based sequencing of virus grown in A3G-modified CEM-SS cells in the presence of 3TC^a

No. of days in cell culture	3TC concn (μM)	Nucleotide sequence in the RT region of pol and predicted amino acid encoded at position 184 in expt:
		1	2	3	4
3	1.48	ATG	ATG	ATG	ATG
		M	M	M	M
7	1.48	A/GTG	ATG	ATG	ATG/A
		M/V	M	M	M/I
9	1.48	GTG	ATG/A	ATG	ATG/A
		V	M/I	M	M/I
13	3	GTG	ATA	ATG	ATA
		V	I	M	I
16	3	GTG	ATA	ATG/A	ATA
		V	I	M/I	I

^aData for the virus population on days 3, 7, 9, 13, and 16 following infection of A3G-modified CEM-SS cells with wild-type HIV-1_NL4-3 in the presence of 3TC (at the IC₉₀) are shown. The concentration of 3TC was escalated from 1.48 to 3.0 μM upon observation of syncytium formation. All viral sequences were aligned with the wild-type HIV-1_NL4-3 sequence. Sequences began at nucleotide position 468 in the RT gene (nucleotide position 3027 in HXB2) and extended to position 670 (nucleotide position 3229 in HXB2). Shown are the sequence chromatogram peak-under-peak reads for the predicted amino acid encoded at position 184 in the RT region of pol at day 16.

To establish the appearance and frequency of the mutations that confer 3TC resistance in the virus population with much greater precision, we performed in-depth analysis of the virus populations by highly parallel DNA sequencing. The HIV_NL4-3 RT-specific forward primer (5′-ACCAGCAATATTCCAGTGTAGCA-3′; 3017 → 3039) contained the fusion primer A and multiplex identifier (MID) sequence (454 Life Sciences). The reverse primer (5′-AATGGAGGTTCTTTCTGATGTTTT-3′; 3206 ← 3229) contained the fusion primer B and MID sequence (454 Life Sciences). Assays were carried out using the 454 genome sequencer FLX system (Roche). The samples were mixed together with identifying tags, and an average of about 10,000 sequence reads per tag was obtained. The distribution of the sequence reads among the different samples, however, had a large variance: after the quality of reads was controlled for, the actual number of sequences analyzed for each of the various tags varied between about 5,000 and 18,000.

In agreement with the direct DNA sequencing results, we found mutations at position 184 in RT that confer 3TC resistance in virus grown in A3G-modified CEM-SS cells in the presence of 3TC (IC₉₀) but not in virus grown in the absence of A3G. The A3G-mediated G-to-A mutation (M184I) accumulated over time according to the sigmoidal logistic curve expected under a constant selective advantage in two experiments (Fig. 2A, exp. 2 and 4) and was identified in a mixed population at day 13 in a third experiment (Fig. 2A, exp. 3) (27). The RT-induced A-to-G mutation (M184V) also accumulated logistically over time in a fourth experiment (Fig. 2A, exp. 1). The predicted amino acid substitutions common to these resistant viruses were not identified in viruses grown in the absence of A3G. Interestingly, in virus grown in A3G-modified CEM-SS cells in the absence of 3TC, the predicted M184I substitution accumulated linearly at low frequency, suggesting that the rapid emergence of drug resistance is due to a specific and predictable mutation (with a large effect) attributable to A3G (Fig. 2B). Even though the mutation may have a detrimental effect on viral fitness in the absence of the drug (5, 6, 7, 19), the overall rates of virus replication did not differ appreciably between the control and A3G-modified CEM-SS cell cultures (Fig. 1A). Thus, even with the high error rate of the virus replicating enzymes, A3G can further augment the genetic complexity of the virus population such that mutations that confer drug resistance can occur in the absence of selection pressure.

FIG. 2.

Frequencies of mutations that confer 3TC resistance in A3G-modified CEM-SS cells grown in the presence and absence of the drug. (A) The frequencies of mutations at position 184 in the RT region of pol in the virus population at days 3, 7, 9, and 13 following infection of A3G-modified CEM-SS cells with wild-type HIV-1_NL4-3 in the presence of 3TC (IC₉₀) are shown. The virus population was subjected to in-depth analysis by highly parallel DNA sequencing. To reduce the number of pyrosequencing artifacts, we screened the processed reads for the pattern of light intensity by using quality filters. DNA sequences were removed from the analysis because of more than one mismatch (insertion, deletion, or substitution) in the MID region, a quality score of <25, or mismatches (except for G-to-A changes) in the two flanking amino acids on either side. The pyrosequencing data were analyzed with the assumption that the observations were drawn independently and randomly from the population of viruses. Exact binomial 95% confidence intervals were calculated using the statistical package R on the underlying frequency under this assumption. A background mutation rate determined in the absence of A3G and 3TC was 0.001. The dominant base replacement errors were transitions (G/A interchange and C/T interchange) at a level of 10⁻³; transversions were rare (unpublished data). In three of four independent experiments, a mutation at position 184 in the RT region of pol emerged by day 9 and reached fixation soon thereafter. In one of four experiments, the M184I mutation that conferred 3TC resistance arose only at day 16. (B) The frequencies of mutations in the RT region of pol in the virus population at days 3, 7, 9, and 13 following infection of control [A3G (−)] and A3G-modified [A3G (+)] CEM-SS cells with wild-type HIV-1_NL4-3 in the absence of 3TC are shown. In A3G-modified CEM-SS cells, the frequency of G-to-A changes increased over time.

Our results provide direct experimental evidence that A3G-mediated editing is a source of adaptive mutation on which selection acts to shape the structure of the virus population. These data further support a model wherein physiological concentrations of A3G contribute to the extent of diversity and the rate of evolution of HIV-1 by enhancing the generation of mutations that enable the virus to evolve and adapt to changing conditions during the course of infection (5, 6, 19). Though difficult to prove unambiguously, it is plausible that A3G contributes to the rapid emergence of mutations in HIV-1 RT that confer 3TC resistance in people on therapy with 3TC alone (12, 20). Thus, the high rates of sequence change that characterize HIV-1 infection can be influenced not only by the poor fidelity of the virus replicating enzymes, but also by the host-encoded antiviral factor A3G. Our study therefore suggests that A3G is an important determinant in explaining the mutational dynamic that underlies the cadence of evolutionary change in HIV-1.