Influence of Reverse Transcriptase Variants, Drugs, and Vpr on Human Immunodeficiency Virus Type 1 Mutant Frequencies

Abstract

The evolution of drug resistance is a major complication of human immunodeficiency virus type 1 (HIV-1) chemotherapy. HIV-1 reverse transcriptase (RT) is a major target of antiretroviral therapy and ultimately the target of drug resistance mutations. Previous studies have indicated that drug-resistant HIV-1 RTs can alter HIV-1 mutant frequencies. In this study, we have tested a panel of HIV-1 RT variants for their ability to influence virus mutant frequencies. The RT variants tested included drug-resistant RT variants as well as other variants analyzed in enzyme fidelity studies with the lacZα gene as a mutation target and/or implicated as being important for enzyme fidelity by structural studies. Combinations of mutations that alone had a statistically significant influence on virus mutant frequencies resulted in different mutant frequency phenotypes. Furthermore, when virus replication occurred in the presence of drugs [e.g., 3′-azido-3′-deoxythymidine, (−)2/,3′-dideoxy-3′-thiacytidine, hydroxyurea, thymidine, or thioguanine] with selected RT variants, virus mutant frequencies increased. Similarly, Vpr variants deficient for binding to the uracil DNA glycosylase repair enzyme were observed to influence HIV-1 virus mutant frequencies when tested alone or in combination with RT variants. In summary, these observations indicate that HIV-1 mutant frequencies can significantly change by single amino acid substitutions in RT and that these effects can be altered by additional mutations in RT, by drugs, and/or by expression of Vpr variants. Such altered virus mutant frequencies could impact HIV-1 dynamics and evolution in small population sizes.

The continuous treatment of human immunodeficiency virus type 1 (HIV-1) infection with the combination of three or more antiretroviral drugs has significantly reduced morbidity and mortality (22, 49). Combination antiretroviral therapy typically contains at least two reverse transcriptase (RT) inhibitors as well as a protease inhibitor. The risk in the emergence of drug-resistant HIV-1 increases when there is poor patient compliance to prescribed drug regimens. This risk may also be of great concern during structured treatment interruptions (1).

The viral polymerase encoded by HIV-1 and other retroviruses, i.e., RT, is highly error prone due to the lack of proofreading ability. The high rates of mutation and recombination that occur during the conversion of the single-stranded viral RNA to a double-stranded DNA is thought to play an important role in generating diversity in HIV-1 and other retrovirus populations (41). RNA polymerase II is thought to also contribute to HIV-1 mutagenesis but to a smaller degree (48). Analysis of the fidelity of HIV-1 RT in cell-free reactions and the HIV-1 mutation rate in cells indicate differences that may be due to protein factors not present in the cell-free reactions or altered in vitro assay conditions (45). The general mechanism used by RT to maintain fidelity likely includes the proper positioning of the template-primer complex, the local geometry of the active site, and the global influence of different conformational states of the enzyme.

HIV-1 RT is a heterodimer that is composed of two subunits, p66 and p51 (32, 35, 64). The p66 subunit contains both polymerase and RNase H catalytic sites while the p51 subunit is a proteolytic cleavage product of p66 that lacks the RNase H domain. The overall folding of both subunits is similar, but the spatial arrangements of their subdomains are very different, which prevents p51 from having polymerase activity. The native structure of the p66 subunit folds into a conformation that resembles the right hand (24, 26, 31). The subdomains of p66 are referred to as the fingers, palm, thumb, and connection. Structural studies in conjunction with phylogenetic analyses have implicated conserved amino acid residues and motifs in HIV-1 RT as being important for enzyme fidelity. The RT active site is located in the palm subdomain within conserved residues 185, 186, and 110. Highly conserved amino acid residues in two alpha-helices of the thumb subdomain along with the fingers and palm subdomains of the 66-kDa subunit act as a clamp to position the template-primer complex relative to the polymerase active site (24). In general, the interactions between RT and its substrates (i.e., the template-primer complex and deoxynucleoside triphosphates [dNTPs]) are important determinants of enzyme fidelity.

Additional factors, including the viral protein R (Vpr), can also influence HIV-1 mutant frequencies. HIV-1 Vpr is a 96-amino-acid nonstructural protein that is associated with virus particles and can accumulate in the nuclei of infected cells (9, 38, 50). The HIV-1 Vpr protein has been found to interact with several cellular partners (53, 68), including the DNA repair enzyme uracil DNA glycosylase (UNG) (4). Vpr has been found to recruit the nuclear form of UNG into HIV-1 virions (44). This recruitment is required for Vpr to modulate the in vivo mutation rate (44). Vpr was found to influence the mutation rate of HIV-1, and the Vpr-UNG interaction was involved in modulating the mutation rate (39, 44, 56).

The goal of this study was to investigate the impact of single amino acid substitutions in RT and to test the potential interplay of RT variants, drugs, and Vpr on HIV-1 mutant frequencies. HIV-1 mutant frequencies were measured with an HIV-1 vector, containing the lacZ gene as a reporter gene, which undergoes one round of replication. The RT variants analyzed included those conferring drug resistance to various drugs as well as variants implicated from structural studies and cell-free fidelity studies as being important in enzyme fidelity. One drug-resistant RT analyzed was observed to significantly alter virus mutant frequencies. In addition, significant changes in virus mutant frequencies were observed with many RT variants that significantly altered the cell-free fidelity of purified RT with the lacZα gene as a mutation target. These mutant frequencies were influenced by additional mutations in RT, by various drugs, and by the expression of Vpr variants.

MATERIALS AND METHODS

Retroviral vectors and expression plasmids.

The HIV-1 vector used in these studies is shown in Fig. 1A. The vector cassette containing the lacZ gene, an internal ribosomal entry site (IRES) element, and the neomycin phosphotransferase gene (neo) was introduced into pGEM-NL4-3 (full-length molecular clone of NL4-3) to create pHIVLacZ-IRES-neo. In order to produce vector virus, the HIV vector was complemented in trans with pSVgagpol-rre-MPMV (kindly provided by David Rekosh, University of Virginia), the amphotropic murine leukemia virus env expression plasmid, pSV-A-MLV-env (34), and a Vpr expression plasmid derived from pAS1B (44). The RT variants analyzed in these experiments were constructed by a primary combinatorial, two-step PCR protocol (25, 39) or by using the Quickchange XL mutagenesis kit (Stratagene). All RT variants made were sequenced to verify the proper introduction of mutations. Vectors for expression of the hemagglutinin (HA)-tagged forms of wild-type (wt) or mutated Vpr were constructed in the pAS1B plasmid as described previously (44).

FIG. 1.

Assay system used to analyze HIV-1 mutant frequencies in vivo during one round of HIV-1 replication with the lacZ gene as a reporter gene. (A) HIV-1 vector used to measure virus mutant frequencies. The proviral DNA form of the vector is shown. The large black rectangular boxes are the long terminal repeats. The small gray box is the simian virus 40 promoter. The lacZ gene, the IRES sequence, and the neo gene are indicated. (B) Single-cycle replication assay for mutant frequencies. HeLa cell clones with single integrated vector proviruses were transiently transfected with helper plasmids, and the produced virus was used to infect fresh HeLa cells. G418-resistant cells resulting from virus infection of fresh HeLa cells were selected, and cells were then stained with X-Gal. The ratio of white- plus light blue-stained colonies to total colonies was used to determine the forward mutant frequency.

Transfections, infections, and cocultivations.

The COS-1 and HeLa cell lines used were obtained from the American Type Culture Collection (Manassas, Va.) and were maintained in Dulbecco's modified Eagle's medium containing 10% calf serum or 10% fetal bovine serum, respectively. HIV-1 vector and expression plasmids were transfected into HeLa cells by using the Superfect reagent (Qiagen). HeLa cells were infected in the presence of Polybrene (23). Infection of HeLa target cells was also done by cocultivation of virus-producing cells with target cells (40, 46).

The influence of the antiretroviral drugs on HIV-1 mutant frequencies was determined by postinfection treatment of cells with drug. Postinfection treatment refers to maintaining HeLa target cells in medium supplemented with drug for 2 h before cocultivation and continued until 24 h after cocultivation. Postinfection treatment with drug influences the HIV-1 mutant frequency only during reverse transcription (42).

Experimental protocol for generating a single round of HIV-1 vector replication.

The experimental protocol developed to generate a single round of HIV-1 vector replication is shown in Fig. 1B. In this protocol, HeLa cells containing the HIV vector provirus was created by transiently transfecting the HIV vector and helper plasmids into COS-1 cells, harvesting virus 2 days posttransfection, and infecting HeLa cells. G418-resistant clones were isolated and characterized for the presence of single proviruses. The cells were also stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) to ensure that no mutations had occurred in the lacZ gene. Selected clones were then used to generate a single round of HIV-1 vector replication. HeLa cell clones with single integrated proviruses were transiently transfected with helper plasmids, treated with mitomycin C at 48 h posttransfection, and then mixed with fresh HeLa cells. G418-resistant cells resulting from virus infection of fresh HeLa cells were selected, and cells were then stained with X-Gal. The ratio of white- plus light blue-stained colonies to total colonies observed provided a forward mutant frequency. In each experiment, similar numbers of colonies were screened for control and experimental samples. Titers for control experiments (wt RT in the absence of drug) were typically 500 to 1,000 CFU per 5 × 10⁵ target cells. The vector system and protocols for analysis of mutant frequencies with the lacZα peptide gene as a mutational target have been previously described (45).

Assay for Vpr virion incorporation into HIV-1 particles.

Incorporation of the Vpr variants was analyzed by using a packaging assay in which HA-tagged Vpr was expressed in trans and incorporated into virions (57). Briefly, 293T cells were cotransfected with 10 μg of the HIV-1-based packaging vector pCMVΔR8.9 (lacking the env and auxiliary genes), 5 μg of the pMD.G plasmid for expression of the vesicular stomatitis virus G protein, and 10 μg of pAS1B-Vpr (wt or mutated). Cell culture supernatants were collected 48 h after transfection and filtered through 0.45-μm-pore-size filters, and virions were collected by ultracentrifugation and suspended in ice-cold lysis buffer (10 mM Tris [pH 7.6], 150 mM NaCl, 2 mM EDTA, 0.5% Triton X-100). For preparation of cell lysates, cells were trypsinized, collected by centrifugation, and suspended in ice-cold lysis buffer. Cell lysates were incubated for 5 min and clarified by centrifugation. Proteins from cell and virion lysates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analyzed by Western blotting with anti-HA 3F10 (Boehringer) or anti-CA-p24 antibodies (63).

RESULTS

Analysis of HIV-1 RT single-amino-acid variants on virus mutant frequencies.

Previous studies have shown that some drug-resistant RT enzymes can influence virus mutant frequencies and the virus mutation rate (42, 43). In particular, two zidovudine (AZT)-resistant enzymes with multiple mutations, i.e., M41L/T215Y and M41L/D67N/K70R/T215Y, were found to increase the mutation rate by factors of 3.3 and 4.3, respectively (42). We thus analyzed a series of drug-resistant enzymes for their influence on virus mutant frequencies. The residues mutated were located in the finger and palm subdomains. The mutant enzymes K65R and L74V confer resistance to the drugs dideoxyinosine, dideoxycytosine, and lamivudine (3TC); D67N and K70R confer resistance to AZT; and E89G confers resistance to multiple drugs (18, 19, 51, 62). K65R also confers resistance to 9-R-2-phosphonomethoxypropyl adenine (tenofovir) and abacavir (1592U89) (21, 61). K65R has been previously reported to increase the fidelity of purified RT eightfold while L74V had no affect on cell-free fidelity when the lacZα gene was employed to detect ct mutations (58). The cell-free fidelity of the E89G RT variant increased the dNTP insertion fidelity but did not decrease the cell-free fidelity of the enzyme (11, 12). Analysis of these RT variants for their influence on virus mutant frequencies indicated that the D67N, K70R, L74V, and E89G RT variants did not significantly alter virus mutant frequencies, whereas the K65R RT variant decreased virus mutant frequencies (Table 1). These data indicate that drug resistance conferred by single-amino-acid substitutions in RT can significantly alter HIV-1 mutant frequencies.

TABLE 1.

Influence of HIV-1 RT variants on virus mutant frequencies in one round of replication

RT variant	Mutation location	Avg mutant frequency ± SD (10−1 mutants/cycle)a	Fold difference (P value)c
wt		1.50 ± 0.05
K65R	Finger subdomain	0.45 ± 0.03	0.3 (<0.0001)
D67N	Finger subdomain	1.49 ± 0.07	1.0 (>0.9)
K70R	Finger subdomain	1.47 ± 0.05	1.0 (>0.9)
L74V	Finger subdomain	1.12 ± 0.04	0.7 (>0.05)
D76V	Finger subdomain	0.59 ± 0.04	0.4 (<0.0005)
R78A	Finger subdomain	0.42 ± 0.02	0.3 (<0.0001)
E89G	Palm subdomain	1.20 ± 0.03	0.8 (>0.15)
Y115A	dNTP binding site	3.40 ± 0.09	2.3 (<0.0001)
Q151N	dNTP binding site	0.25 ± 0.01	0.2 (<0.0001)
K154A	dNTP binding site	1.52 ± 0.05	1.0 (>0.9)
F227A	Primer grip	0.93 ± 0.04	0.6 (>0.05)
W229A	Primer grip	0.72 ± 0.06	0.5 (<0.0005)
G262A	Thumb subdomain	NDb	ND
W266A	Thumb subdomain	ND	ND
Y501W	RNase H primer grip	4.03 ± 0.09	2.7 (<0.0001)
I505A	RNase H primer grip	1.41 ± 0.12	0.9 (>0.25)

^aThe average mutant frequencies ± standard deviations were determined from three independent experiments.

^bMutant frequency and standard deviation were not determined (ND) due to low virus titers.

^cP values were determined by chi-square analysis. Differences (n-fold) are given relative to the wt RT.

Recent reports have implicated residues of the finger subdomain (that are not associated with drug resistance) in altering RT fidelity in cell-free systems using the lacZα gene. In particular, the D76V and R78A substitutions have each been found to increase the cell-free fidelity of purified HIV-1 RT by a factor of 9 (29, 30). D76 has been found in structural studies to interact with the template nucleotide that base pairs with the incoming dNTP while R78 interacts with the template nucleotide that base pairs with the nucleotide at the 3′ primer terminus (24). In our experiments, D76V and R78A were observed to decrease HIV-1 mutant frequencies (Table 1).

The amino acid residues that interact with the incoming dNTP and form the dNTP binding site have been identified in structural studies. Two mutations in the dNTP binding site of HIV-1 RT, Q151N and K154A, have been reported to increase fidelity by factors of 13 and 2.1, respectively, using lacZα (65). Interestingly, the Q151 residue in RT has been implicated in drug resistance to multiple drugs (i.e., the Q151M RT variant) (59). The Q151N RT variant significantly decreased virus mutant frequencies while the K154A RT variant was not found to have a significant influence on virus mutant frequencies (Table 1). Intriguingly, another substitution in the dNTP binding site, Y115A, has been previously reported to decrease fidelity by a factor of 4 using the lacZα gene (27). However, the Y115F and Y115V RT variants were found in lacZα cell-free fidelity assays to have slightly lower error rates than that of wt RT (5). The Y115F mutation confers resistance to abacavir (21). Nonconservative changes at the Y115 residue have been found to result in a dramatic reduction in the ability of purified HIV-1 RT to discriminate against ribonucleotides in the presence of both magnesium and manganese cations (6). In our experiments, we observed that the Y115A RT variant significantly increased (2.3-fold) virus mutant frequencies (Table 1).

The αH helix motif of the HIV-1 RT thumb subdomain binds to the minor groove of the template-primer complex and is associated with alterations in RT processivity and fidelity. The 3′ end of the primer is positioned near the RT active site by conserved residues 224 to 335 on the β11b loop and the β12 and β13 hairpins. The region bounded by residues 227 to 235 is referred to as the primer grip (26). The F227A and W229A mutations in the αH helix have been reported to alter misincorporation and mispair extension frequencies, presumedly through an increase in the rate of RT dissociation from the template DNA or through an increase in strand slippage resulting in lower enzyme fidelity (66). These mutations in RT were therefore tested to determine what influence they would have on virus mutant frequencies by using an assay for measuring the HIV-1 mutant frequency in one round of replication (Fig. 1). As indicated in Table 1, the F227A mutation did not significantly alter the virus mutant frequencies, but the W229A RT variant did influence virus mutant frequencies. The G262 and W266 residues are also located in the thumb domain and have been previously found to increase HIV-1 RT cell-free fidelity in the lacZα reporter assay (2). In particular, the G262A and W266A variants have reduced fidelity for template-primer slippage errors, such as frameshift and deletion mutations. However, these RT variants lowered virus infectivity to a level at which their influence on virus mutant frequencies could not be readily determined (Table 1).

The RNase H domain is responsible for the RNase H activity required for the degradation of the viral RNA present in the RNA-DNA replication intermediate, allowing the DNA to act as a template for the DNA-dependent DNA polymerase activity of RT to complete synthesis of the double-stranded viral DNA. The RNase H activity of RT is crucial for the strand transfers that occur during reverse transcription. Recent structural studies of HIV-1 RT complexed with a polypurine track-containing RNA-DNA hybrid has suggested that an RNase H primer grip exists that interacts with the DNA primer strand and could influence the trajectory of the RNA template relative to the RNase H catalytic center (55). Amino acid residues in the RNase H domain that are involved in the RNase H primer grip include residues 473 to 476, 501, and 505. To test whether residues in the RNase H primer grip could influence virus mutant frequencies, mutations at Y501 and I505 (i.e., Y501W and I505A) were analyzed. Analysis of these RT variants on virus mutant frequencies indicated that Y501W, but not I505A, significantly influenced virus mutant frequencies compared to wt RT (Table 1). Interestingly, a recent study of a murine leukemia virus RT variant, Y586F (the equivalent of Y501 in HIV-1 RT), showed a 5-fold increase in the in vivo mutation rate and a 17-fold increase in the frequency of substitution mutations within 18 nucleotides of adenine-thymine tracts (which are known to induce DNA bending) (67). This observation suggests a mechanism for how the Y501W mutant, a residue in the RNase H domain, may influence HIV-1 mutant frequencies. In summary, several RT variants with single-amino-acid changes have been characterized in both the polymerase and RNase H domains that influence virus mutant frequencies.

Analysis of combined mutations in HIV-1 RT on virus mutant frequencies.

The results from Table 1 identified several amino acid residues in RT which, when mutated, can significantly alter virus mutant frequencies. To test whether mutated residues could act together to alter virus mutant frequencies, a series of RT variants with combined mutations were analyzed (Table 2). The variants tested included the D76V/R78A, R78A/Q151N, Q151N/W266A, and Y115A/Q151N mutations, which corresponded to mutations in the finger subdomain, the finger subdomain and the dNTP binding site, the dNTP binding site and the thumb subdomain, and the dNTP binding site, respectively.

TABLE 2.

Analysis of RT variants with combined mutations on HIV-1 mutant frequencies

RT variant	Avg mutant frequency ± SD (10−1 mutants/cycle)a	Fold difference (P value)c relative to:
		wt RT	D76V RT	R78A RT	Y115A RT	Q151N RT
wt	1.51 ± 0.06
D76V	0.60 ± 0.03	0.4 (<0.0005)
R78A	0.43 ± 0.02	0.3 (<0.0001)
Y115A	3.48 ± 0.11	2.3 (<0.0001)
Q151N	0.28 ± 0.01	0.2 (<0.0001)
D76V/R78A	0.15 ± 0.05	0.1 (<0.0001)	0.3 (<0.0001)	0.6 (>0.05)
R78A/Q151N	0.11 ± 0.07	0.07 (<0.0001)		0.3 (<0.0001)		0.4 (<0.0005)
Q151N/W266A	NDb	ND
Y115A/Q151N	1.05 ± 0.10	0.7 (>0.05)			0.3 (<0.0001)	3.8 (<0.0001)

^aThe average mutant frequencies ± standard deviations were determined from three independent experiments.

^bMutant frequency and standard deviation were not determined (ND) due to low virus titers.

^cP values were determined by chi-square analysis.

The HIV-1 RT variants containing either the D76V/R78A or the R78A/Q151N double mutation led to mutant frequencies that were significantly lower than the mutant frequencies of wt HIV-1 (Table 2). The mutant frequency of R78A/Q151N was significantly lower than the variants with each mutation alone while the mutant frequency of D76V/R78A was significantly lower than that of the R76V RT variant but not that of the R78A RT variant. The RT variant with the Q151N/W266A double mutation did not produce high enough levels of infectious virus in order to determine the effect of the two mutations on the virus mutant frequency. Finally, the Y115A/Q151N double mutation in RT led to a mutant frequency that was significantly different than the mutant frequencies after one round of HIV-1 vector replication with RT variants containing the Y115A mutation, the Q151N mutation, or wt RT (Table 2). A recent study with cell-free RT found that the M230I/Y115W RT variant had an altered phenotype compared to that of the Y115W RT variant in misinsertion fidelity assays, which may be associated with our observations here (20). In summary, the data support the conclusion that residues in RT can act together in altering virus mutant frequencies.

Influence of drugs and HIV-1 RT variants on virus mutant frequencies.

Previous studies have indicated that both RT variants and drugs could increase virus mutant frequencies (42, 43). Several of the RT variants identified in Table 1 were found to significantly decrease virus mutant frequencies. To determine if virus mutant frequencies during virus replication with RT variants is influenced by the presence of drugs, selected RT variants (i.e., R78A and Q151N) were used for virus replication in the presence of several different drugs. First, AZT and 3TC were used because of previous work showing that these drugs could increase virus mutant frequencies (42). Virus replication with the R78A or Q151N RT variant in the presence of 0.4 μM AZT increased virus mutant frequencies to those observed during virus replication with each of the HIV-1 RT variants in the absence of drug (Table 3). HIV-1 replication with the R78A or Q151N RT variant in the presence of 0.3 μM 3TC also led to higher mutant frequencies than that observed during virus replication with the RT variants in the absence of drug (Table 3).

TABLE 3.

Influence of various drugs and selected RT variants on HIV-1 mutant frequencies

RT variant	Drug dose	Avg mutant frequency ± SD (10−1 mutants/cycle)a	Fold difference (P value)b
wt	None	1.60 ± 0.07
R78A	None	0.55 ± 0.03	0.3 (<0.0001)
Q151N	None	0.35 ± 0.02	0.2 (<0.0001)
wt	0.4 μM AZT	10.1 ± 0.11
R78A	0.4 μM AZT	4.12 ± 0.04	0.4 (<0.0005)
Q151N	0.4 μM AZT	3.03 ± 0.03	0.3 (<0.0001)
wt	0.3 μM 3TC	4.97 ± 0.04
R78A	0.3 μM 3TC	0.98 ± 0.03	0.2 (<0.0001)
Q151N	0.3 μM 3TC	0.75 ± 0.02	0.2 (<0.0001)
wt	2 mM HU	4.88 ± 0.03
R78A	2 mM HU	0.96 ± 0.03	0.2 (<0.0001)
Q151N	2 mM HU	0.81 ± 0.02	0.2 (<0.0001)
wt	50 μM Thy	5.03 ± 0.04
R78A	50 μM Thy	1.04 ± 0.02	0.2 (<0.0001)
Q151N	50 μM Thy	0.77 ± 0.02	0.2 (<0.0001)

^aThe average mutant frequencies ± standard deviations were determined from three independent experiments.

^bP values were determined by chi-square analysis. Differences are relative to the wt RT plus drug.

Hydroxyurea (HU) and thymidine (Thy) have been previously shown to alter nucleotide pools and to increase the virus mutant frequencies of different retroviruses, including HIV-1 (3, 8, 15-17, 28, 37, 43, 47, 54, 60). We hypothesized that virus replication with the R78A and Q151N RT variants in the presence of either HU or Thy would increase virus mutant frequencies compared to those in the absence of drug because of the ability of HU and Thy to alter nucleotide pools. HU (2 mM) treatment during HIV-1 replication with the R78A RT resulted in an increased mutant frequency compared to that of the R78A RT in the absence of drug. Similarly, an increased mutant frequency was also observed with HIV-1 replication with the Q151N RT along with HU treatment of cells compared to virus replication in the absence of drug. Treatment of cells with 50 μM Thy during virus replication with the R78A or Q151N RT variant led to significant increases compared to those of virus replication with these RT variants in the absence of drug. The increased mutant frequencies observed during virus replication in the presence of HU or Thy were presumedly due to the alteration of nucleotide pools. These data support the conclusion that RT variants with single-amino-acid substitutions and drugs can act together in modulating virus mutant frequencies.

Effect of TG on HIV-1 mutant frequencies.

Thioguanine (TG) has been widely used as an antileukemic agent for many years (13). Recently, it has been reported that the active metabolite of TG, 2′-deoxy-6-thioguanosine 5′-triphosphate, can inhibit HIV-1 replication by preventing RNase H activity (33). We were interested in testing whether TG could influence virus mutant frequencies. First, we tested various concentrations of TG, ranging from 0.5 to 10 μM, on virus mutant frequencies by using the same postinfection treatment strategy described earlier for other drugs (see Materials and Methods). We observed that there was a corresponding increase in virus mutant frequencies (Fig. 2). This indicates that TG influences the virus mutant frequency in a dose-dependent manner. The maximum increase in virus mutant frequency was three times that of the mutant frequency in the absence of the drug. We next tested whether TG treatment could influence virus mutant frequencies during replication with selected HIV-1 RT variants. Interestingly, 5 μM TG treatment was found to significantly increase the mutant frequencies when virus replication occurred with either the R78A or the Q151N variant compared to that of wt RT (Table 4). This indicates that TG and RT variants can act together to modulate virus mutant frequencies.

FIG. 2.

Dose-dependent effect of TG on HIV-1 mutant frequencies. Various concentrations of TG were analyzed for their influence on HIV-1 replication in one round of replication. Mutant frequencies were determined as described in the legend to Fig. 1. The data were collected from three replicate experiments and are presented as average mutant frequencies ± standard deviations.

TABLE 4.

Influence of TG and selected RT variants on HIV-1 mutant frequencies

RT variant	Drug dose	Avg mutant frequency ± SD (10−1 mutants/cycle)a	Fold difference (P value)b
wt	None	1.58 ± 0.05
R78A	None	0.59 ± 0.03	0.4 (<0.0005)
Q151N	None	0.30 ± 0.03	0.2 (<0.0001)
wt	5 μM TG	3.02 ± 0.11
R78A	5 μM TG	1.16 ± 0.04	0.4 (<0.0005)
Q151N	5 μM TG	0.87 ± 0.03	0.3 (<0.0001)

^aThe average mutant frequencies ± standard deviations were determined from three independent experiments.

^bP values were determined by Chi-square analysis. Differences are relative to the wt RT plus drug.

Combined influence of Vpr and HIV-1 RT variants on virus mutant frequencies.

Previous studies have indicated that HIV-1 Vpr can influence the in vivo virus mutation rate (39, 44). This influence is related to the property of Vpr to recruit the nuclear form of UNG (UNG2) into virus particles. One Vpr variant, Vpr^*W54R, which failed to recruit UNG2 into HIV-1 virions, was not able to complement a vpr-null mutant HIV-1 in the mutation rate assay (44, 56). To further characterize how Vpr may influence virus mutant frequencies, other Vpr variants that did not efficiently interact with UNG were analyzed. Specifically, two Vpr variants with single substitutions of H71 and H78 (i.e., Vpr^*H71R or Vpr^*H78R), which bound poorly to UNG2 (56), were analyzed. Another mutant (Vpr^*S79A), which interacted more efficiently with UNG2 than wt Vpr, was also included in the analysis (56). We first checked that these three Vpr variants were efficiently incorporated into HIV-1 particles. As shown in Fig. 3, each Vpr variant was readily detected in both cells and virus particles, indicating that each variant was efficiently expressed and incorporated into HIV-1 virions as efficiently as the wt Vpr protein.

FIG. 3.

Virion incorporation of Vpr variants. 293T cells were cotransfected with pCMVΔR8.9 (lacking the vpr gene) and pMD.G and either with or without a mutated HA-tagged Vpr expression vector. Proteins from cell and virion lysates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analyzed by Western blotting with anti-HA or anti-CA-p24.

We next tested how these Vpr variants influenced virus mutant frequencies during virus replication with either wt RT or an RT variant, R78A. As expected, there was a significant increase in mutant frequency during virus replication with wt RT and expression of either Vpr^*H71R or Vpr^*H78R than when wt Vpr was expressed (Table 5). In addition, a significant increase in virus mutant frequency was observed with the R78A RT in combination with either Vpr^*H71R or Vpr^*H78R compared to that found when wt Vpr was expressed (Table 5). These data indicate that the influence of these Vpr variants on virus mutant frequencies followed similar trends when virus replication was with wt RT or the R78A RT variant. The average mutant frequency observed during virus replication with wt RT and expression of the previously reported UNG2 binding-deficient Vpr^*W54R variant was not significantly different from that of either Vpr^*H71R or Vpr^*H78R. Similarly, when the R78A RT variant was used in virus replication, the mutant frequency with Vpr^*W54R was not significantly different from that of Vpr^*H71R or Vpr^*H78R. This indicates that Vpr^*H71R and Vpr^*H78R influence virus mutant frequencies in the same manner as Vpr^*W54R. Since Vpr^*H71R and Vpr^*H78R do not efficiently interact with UNG2, this confirms that the poor interaction with UNG2 correlates with the altered mutant frequencies. The average mutant frequency observed during virus replication with wt RT and the expression of Vpr^*S79A was not significantly different from that with wt Vpr, and the mutant frequencies with the R78A variant and Vpr^*S79A were not significantly different from those with the R78A RT and wt Vpr. Since Vpr^*S79A interacts efficiently with UNG2, these data further support the observation that the Vpr-UNG interaction can modulate virus mutant frequencies. Furthermore, this indicates that the Vpr-UNG interaction has a similar impact on virus mutant frequencies with either an RT variant (R78A) or the wt RT. In particular, the ability of the R78A RT variant to increase virus mutant frequencies was counteracted by the Vpr effect, indicating that the mechanism responsible for the Vpr effects on the virus mutation rate is independent of the altered mutant frequencies associated with the RT variant.

TABLE 5.

Influence of Vpr variants and a selected RT variant on HIV-1 mutant frequencies

RT variant	Vpr variant	Avg mutant frequency ± SD (10−1 mutants/cycle)a	Fold difference (P value)b of:
			R78A RT relative to wt RT	wt RT Vpr variants relative to wt Vpr	R78A RT Vpr variants relative to wt Vpr
wt	wt	1.61 ± 0.06
R78A	wt	0.58 ± 0.03	0.4 (<0.0005)
wt	Vpr*W54R	5.33 ± 0.06		3.3 (<0.0001)
R78A	Vpr*W54R	1.54 ± 0.03	0.3 (<0.0001)		2.7 (<0.0001)
wt	Vpr*H71R	4.71 ± 0.11		2.9 (<0.0001)
R78A	Vpr*H71R	1.16 ± 0.04	0.2 (<0.0001)		2.0 (<0.001)
wt	Vpr*H78R	4.67 ± 0.09		2.9 (<0.0001)
R78A	Vpr*H78R	1.46 ± 0.04	0.3 (<0.0001)		2.5 (<0.0001)
wt	Vpr*S79A	1.58 ± 0.05		1.0 (>0.9)
R78A	Vpr*S79A	0.53 ± 0.04	0.3 (<0.0001)		1.0 (>0.9)

^aThe average mutant frequencies ± standard deviations were determined from three independent experiments.

^bP values were determined by chi-square analysis.

Finally, we tested whether the Vpr^*W54R variant could act along with an RT variant (Y115A RT), which increased the mutant frequency, to alter virus mutant frequencies in the presence of drug (AZT) compared to wt RT in the absence of drug. Since the potential impact on the virus mutant frequency was anticipated to be significant, the lacZα peptide gene was used as a mutational target (44). The lacZα peptide gene is a smaller target than the lacZ gene and can more easily detect larger changes in mutant frequencies. The virus mutant frequency of the HIV-1 vector containing the lacZα peptide gene with Y115A RT in the presence of AZT was 18 times higher than that observed when wt RT was used in virus replication in the absence of drug. The increase in virus mutant frequency follows a multiplicative model (43). When Y115A RT was used with W54R^*Vpr in the presence of 0.4 μM AZT, the virus mutant frequency was 25 times higher than during replication with wt RT, wt Vpr, and no drug (Table 6). The increase in mutant frequency due to Vpr^*W54R was additive. These observations suggest that Vpr influences virus mutant frequencies in a manner that is mechanistically different from how the Y115A RT variant or AZT influences virus mutant frequencies. These observations extend previous studies on the role of Vpr on the influence of virus mutant frequencies (39, 44). In summary, these data indicate an interplay between RT variants, drugs, and Vpr that can significantly alter HIV-1 mutant frequencies.

TABLE 6.

Influence of RT and Vpr variants on HIV-1 mutant frequencies during replication in the presence of AZT with the lacZα peptide gene as a mutation target

RT variant	Vpr variant	AZT concn (μM)	Mutant frequency (mutants/cycle)a	Fold difference (P value)b
wt	wt	0	0.005 ± 0.002
Y115A	wt	0.4	0.090 ± 0.005	18 (<0.0001)
Y115A	W54R	0.4	0.124 ± 0.007	25 (<0.0001)

^aMutant frequencies are averages from three independent experiments ± standard deviations.

^bP values were determined by chi-square analysis.

DISCUSSION

In this study, we have analyzed HIV-1 mutant frequencies in a single round of replication with an HIV-1 vector containing the lacZ gene. A series of amino acid substitutions were created in RT to determine their influence on virus mutant frequencies. The amino acid residues that were targets for site-directed mutagenesis were chosen based upon their association with drug resistance or based upon structural and/or biochemical studies that implicated their interaction with the primer-template complex or with the incoming dNTP and implicated their ability to influence enzyme fidelity. One drug resistance mutation in RT (i.e., K65R) was observed to increase virus mutant frequencies. This indicates that at least some drug resistance mutations conferred by single-amino-acid substitutions can significantly alter HIV-1 mutant frequencies.

In many instances, RT variants analyzed in cell-free fidelity studies with the lacZα gene had a milder effect on virus mutant frequencies than that predicted from the cell-free studies. These differences could be due, in part, to the different mutation targets used. In this study, the entire lacZ gene was used as a mutation target, whereas in many of the cell-free fidelity studies, the lacZα gene was used as a target (2, 5, 11, 27, 29, 30, 58, 65). In cell-free systems, synthesis of one DNA strand is used to determine these rates. In contrast, the assay described here includes both minus-strand (with an RNA template) and plus-strand (with a DNA template) DNA synthesis. In addition, other differences between the cell-free and mutant frequency assays (including the presence or absence of protein factors, different RTs, and different experimental conditions) may play a role in discrepancies between the two systems.

Combining mutations in RT which alone were found to reduce virus mutant frequencies led to further reductions in the virus mutant frequency, indicating that the mutations can act together to modulate virus mutant frequencies. The impact of selected RT variants that reduced virus mutant frequencies were significantly increased with AZT, while more subtle yet significant increases in virus mutant frequencies were observed with 3TC, HU, and Thy. The mechanism(s) whereby RT variants and drugs alter virus mutant frequencies is under current investigation.

The active metabolite of the antileukemic agent TG was recently shown to inhibit HIV-1 replication by preventing RNase H activity (33). We tested whether TG postinfection treatment could alter HIV-1 mutant frequencies and found that TG increased (threefold maximum) virus mutant frequencies in a dose-dependent manner. Though the mechanism(s) for how TG influences HIV-1 mutant frequencies is unknown, previous studies suggest that inhibition of RNase H activity may increase HIV-1 mutant frequencies. Treatment of cells with TG was also found to increase the mutant frequencies of viruses with the R78A and Q151N RT variants. Mutation of the RNase H domain can also alter HIV-1 mutant frequencies, as indicated by the Y501W RT variant. These data indicate that the RNase H domain can alter virus mutant frequencies.

The two Vpr variants that were previously shown to have altered binding to UNG, Vpr^*H71R and Vpr^*H78R, were found to be efficiently incorporated into virus particles and increased the virus mutant frequency when wt RT and the R78A RT variant were used for virus replication. Therefore, these two new Vpr variants had a mutant frequency phenotype similar to that of the previously reported UNG2 binding-deficient Vpr^*W54R variant (44), providing further evidence for a role for Vpr-UNG2 association in modulating HIV-1 mutant frequencies. These data also suggest that the mechanism(s) used by Vpr to alter HIV-1 mutant frequencies is distinct from the mechanism by which the R78A mutation in RT modulates fidelity.

We have shown that single-amino-acid substitutions in RT can change HIV-1 mutant frequencies and that these effects are influenced by additional mutations in RT, by drugs, and by Vpr variants deficient for binding to UNG2. The magnitude of these changes were all within 30-fold of the mutant frequency with wt virus in the absence of drug. The observations made in this study provide the basis for two lines of continued investigation. First, these data will allow for mechanistic studies of the determinants of HIV-1 mutant frequencies and the HIV-1 mutation rate. Second, these observations will allow for the analysis of altered rates of mutation on HIV-1 replication as well as on HIV-1 pathogenesis and drug therapy. Little is known about changes in fidelity during the natural course of HIV-1 infection. However, studies with simian immunodeficiency virus (SIV) have indicated that, during the course of disease progression, changes in fidelity occur (10). In particular, an SIV clone called SIVMNE 170 was isolated from a macaque during the late symptomatic phase of infection with the parental strain SIVMNE CL8. A misincorporation assay indicated that the SIVMNE 170 RT showed much higher fidelity than SIVMNE CL8, suggesting that the fidelity of lentiviral RTs may increase during the course of viral infection. Previous studies have indicated that drugs and drug-resistant RT can significantly increase virus mutant frequencies and the HIV-1 mutation rate (42, 43). These increased mutant frequencies could have important implications for HIV-1 evolution, population dynamics, and drug therapy regimens. However, the impact of an altered HIV-1 mutation rate is dependent on the dynamics of the virus population. Deterministic computer modeling has been used to predict the effects of mutation and selection on virus populations, particularly with HIV-1 (7, 52), while stochastic models have also been proposed (36). A metapopulation model for HIV-1 replication has been recently reported (14). This model found that the combination of founder effects and subpopulation turnover can result in an effective population size much lower than the actual population size and may contribute to the importance of genetic drift in HIV-1 evolution despite a large number of infected cells. The impact of changes in virus mutant frequencies due to RT, drugs, or Vpr variants on HIV-1 population dynamics and evolution could be particularly important in small population sizes.