Insights into the Molecular Mechanism of Mitochondrial Toxicity by AIDS Drugs*

Several of the nucleoside analogs used in the treatment of AIDS exhibit a delayed clinical toxicity limiting their usefulness. The toxicity of nucleoside analogs may be related to their effects on the human mitochondrial DNA polymerase (Pol γ), the polymerase responsible for mitochondrial DNA replication. Among the AIDS drugs approved by the FDA for clinical use, two are modified cytosine analogs, Zalcitabine (2′,3′-dideoxycytidine (ddC)) and Lamivudine (β-d-(+)-2′,3′-dideoxy-3′-thiacytidine ((−)3TC])). (−)3TC is the only analog containing an unnaturall(−) nucleoside configuration and is well tolerated by patients even after long term administration. In cell culture (−)3TC is less toxic than its d(+) isomer, (+)3TC, containing the natural nucleoside configuration, and both are considerably less toxic than ddC. We have investigated the mechanistic basis for the differential toxicity of these three cytosine analogs by comparing the effects of dideoxy-CTP), (+)3TC-triphosphate (TP), and (−)3TC-TP on the polymerase and exonuclease activities of recombinant human Pol γ. This analysis reveals that Pol γ incorporates (−)3TC-triphosphate 16-fold less efficiently than the corresponding (+)isomer and 1140-fold less efficiently than dideoxy-CTP, showing a good correlation between incorporation rate and toxicity. The rates of excision of the incorporated analogs from the chain-terminated 3′-end of the DNA primer by the 3′-5′-exonuclease activity of Pol γ were similar (0.01 s− 1) for both 3TC analogs. In marked contrast, the rate of exonuclease removal of a ddC chain-terminated DNA occurs at least 2 orders of magnitude slower, suggesting that the failure of the exonuclease to remove ddC may play a major role in its greater toxicity. This study demonstrates that direct analysis of the mitochondrial DNA polymerase structure/function relationships may provide valuable insights leading to the design of less toxic inhibitors.

Several of the nucleoside analogs used in the treatment of AIDS exhibit a delayed clinical toxicity limiting their usefulness. The toxicity of nucleoside analogs may be related to their effects on the human mitochondrial DNA polymerase (Pol ␥), the polymerase responsible for mitochondrial DNA replication. Among the AIDS drugs approved by the FDA for clinical use, two are modified cytosine analogs, Zalcitabine (

2,3-dideoxycytidine (ddC)) and Lamivudine (␤-D-(؉)-2,3-dideoxy-3-thiacytidine ((؊)3TC])). (؊)3TC is the only analog containing an unnatural L(؊) nucleoside configuration and is well tolerated by patients even after long term administration.
In cell culture (؊)3TC is less toxic than its D(؉) isomer, (؉)3TC, containing the natural nucleoside configuration, and both are considerably less toxic than ddC. We have investigated the mechanistic basis for the differential toxicity of these three cytosine analogs by comparing the effects of dideoxy-CTP), (؉)3TC-triphosphate (TP), and (؊)3TC-TP on the polymerase and exonuclease activities of recombinant human Pol ␥. This analysis reveals that Pol ␥ incorporates (؊)3TC-triphosphate 16fold less efficiently than the corresponding (؉)isomer and 1140-fold less efficiently than dideoxy-CTP, showing a good correlation between incorporation rate and toxicity. The rates of excision of the incorporated analogs from the chain-terminated 3-end of the DNA primer by the 3-5-exonuclease activity of Pol ␥ were similar (0.01 s ؊1 ) for both 3TC analogs. In marked contrast, the rate of exonuclease removal of a ddC chain-terminated DNA occurs at least 2 orders of magnitude slower, suggesting that the failure of the exonuclease to remove ddC may play a major role in its greater toxicity. This study demonstrates that direct analysis of the mitochondrial DNA polymerase structure/function relationships may provide valuable insights leading to the design of less toxic inhibitors.
The in vitro and in vivo mechanisms of cytotoxicity of the nucleoside analogs used as anti-HIV 1 drugs are not well un-derstood. Various forms of toxicity, including peripheral neuropathy, myopathy, and pancreatitis, are observed after long term use of nucleoside analogs 3Ј-azido-3Ј-deoxythymidine (AZT) (1), 2Ј,3Ј-dideoxyinosine (ddI) (2), 2Ј,3Ј-dideoxycytidine (ddC) (3,4), and 2Ј,3Ј-didehydro-3Ј-deoxythymidine (5). The profile for toxicity varies with each analog. For instance, whereas all of the three, AZT, ddI, and ddC, exert cytotoxic effects on human muscle cells and induce functional alternation of mitochondria, only AZT, but not ddI and ddC, can induce a myopathy in HIV-1-infected patients (6). The fact that toxicity of different drugs is cell-, tissue-, and organ-specific (6) indicates that the differences in subcellular bioavailability and pharmacokinetic profile involving the activation and deactivation of the drugs can all contribute to the clinical toxicity. Many mechanisms have been proposed for the origins of nucleoside analog toxicity, but the most obvious sites of action are the enzymes responsible for host cell DNA replication and repair: DNA polymerases ␣, ␦, and ⑀, which are responsible for chromosomal DNA replication; DNA polymerase ␤, which is involved in DNA repair; and DNA polymerase ␥ (Pol ␥), which carries out mitochondrial DNA synthesis (7)(8)(9). The toxicity toward proliferating tissues, such as the well documented bone marrow toxicity of AZT, is believed to be due to the inhibition of one or more of the chromosomal DNA polymerases (10). The toxicity manifested in many nonproliferating tissues is thought to result more specifically from the effects of inhibition of DNA Pol ␥ by nucleoside analogs (10 -14). In support of this postulate, the recognizable clinical syndromes associated with nucleoside analog toxicity are reminiscent of mitochondrial dysfunctions associated with mitochondrial genetic disorders (12,(15)(16)(17)(18); moreover, prolonged exposure of the cell to nucleoside analogs results in decreased mitochondrial DNA synthesis, decreased cell viability, and changes in the mitochondrial ultrastructure (19). Some in vitro studies have shown that for many of the nucleoside analogs tested, Pol ␥ is more sensitive than any of the other DNA polymerases (20), but there are conflicting reports (19,21). Earlier studies to assess the interaction of nucleoside analogs have used partially purified preparations of native enzyme. However, recently, highly purified homogenous preparations of recombinant human Pol ␥ have become available through cloning and overexpression in a baculovirus expression system (22,(23)(24)(25)(26). Human DNA Pol ␥ is composed of a catalytic subunit of 140 kDa and an accessory subunit of ϳ54 kDa (22). The large catalytic subunit provides both polymerase and 3Ј-5Ј-exonuclease activities (23,24), whereas the smaller accessory subunit (25) facilitates substrate binding and dNTP incorporation, thereby improving pro-cessivity (26). The cloning, overexpression, purification, and reconstitution of the human Pol ␥ has now afforded a highly purified enzyme preparation, allowing a more rigorous analysis of the reactions governing nucleotide selectivity during incorporation and removal by the proofreading exonuclease (27).
Among the anti-HIV nucleoside analogs used clinically, the ␤-L-(Ϫ)-2Ј,3Ј-dideoxy-3Ј-thiacytidine ((Ϫ)3TC, lamivudine) is the only one with the unnatural L configuration, and most interestingly, it has been shown to be more potent and less toxic than the D isomer, (ϩ)3TC (Fig. 1). Effective doses of (Ϫ)3TC are well tolerated during therapy (28 -30), and there is little or no evidence of mitochondrial injury (13,30). In contrast, ddC is one of the more toxic of the nucleoside analogs approved for AIDS therapy by the Federal Drug Administration. The mechanistic basis for observed stereochemical differential toxicity for the two 3TC isomers and enhanced toxicity of ddC are not understood, and a number of factors may come into play. The differences in toxicity may be a combination of factors including cellular uptake (31,32), transport (10), metabolic activation (31)(32)(33), incorporation (31,33,34), and removal or degradation from the system (10,27,33,35,36). A quantitative analysis of each factor is clearly required.
The mechanism of inhibition of DNA Pol ␥ by the nucleoside analogs involves in vivo phosphorylation to form the triphosphate and subsequent incorporation into the DNA primer strand, resulting in chain termination. Thus in vitro studies have focused on evaluating the triphosphate form of analogues. Earlier steady-state kinetic studies have examined the effect of the (ϩ) and (Ϫ)3TC-TP incorporation into DNA by Pol ␥ (21, 31) and suggest a 5-fold difference in K i values. However, these steady-state kinetic studies were conducted under conditions that employed a low concentration of Pol ␥, and according the to the K d value of 35 nM for formation of the heterodimer (26), the polymerase would not have existed as the holoenzyme complex. Indeed, the significance of earlier studies using native purified enzyme is unclear due to the variable activity resulting from the loss of the small subunit during purification.
Similar to many of the DNA polymerases, Pol ␥ has 3Ј-5Јexonuclease activity, which is believed to perform a proofreading function (22,37). It is possible that the failure of 3Ј-5Јexonuclease of Pol ␥ to efficiently remove chain terminators may play a role in mitochondrial toxicity of nucleoside analogs. However, discrepancies exist in the literature concerning the efficiency of the exonuclease in removal of these analogs and correlation to drug toxicity. It has been proposed that efficient removal of 3Ј-terminal (Ϫ)3TC-monophosphates (MPs) by the 3Ј-5Ј-exonuclease activity of Pol ␥ could be responsible for the low toxicity of (Ϫ)3TC (33); however, similar rates of exonuclease removal were reported for the excision of dCMP, ddC-MP, and (Ϫ)3TC-MP from the 3Ј-terminus of a DNA primer, even though there are significant differences in mitochondrial toxicity (33). This assessment was made with a partially purified preparation of Pol ␥ that may not represent the holoenzyme complex due to the loss of the accessory subunit. Moreover, a novel cytosolic exonuclease was isolated from H9 cells that can remove chain-terminated dNMP and may be a contaminating activity in partially purified Pol ␥ (27). Finally, in cell culture studies, ddC showed higher inhibition of mitochondrial DNA synthesis than both the (ϩ) and (Ϫ)3TC (31,39). Thus, a detailed study using pure human Pol ␥ is needed.
Previously, we have studied the stereochemical selectivity between the (ϩ) and (Ϫ)3TC triphosphate for inhibition of HIV-1 reverse transcriptase (RT) (40). We have also determined the detailed kinetic properties of human mitochondrial DNA polymerase for both the catalytic subunit and the holoenzyme (23,26). We now describe our transient kinetic analysis of incorporation and excision of dideoxycytidine triphosphate (ddCTP), (ϩ)3TC, and (Ϫ)3TC, with the goal of understanding how these clinically important modified cytosine analogs interact with the Pol ␥ holoenzyme complex in comparison to dCTP. An in-depth understanding of the mechanism of inhibition of the DNA ␥ polymerase is important in overcoming problems associated with cytotoxicity with nucleoside analogs. Knowledge of the mechanistic similarities and differences may identify key features important in selectivity for the interaction with the viral DNA polymerase (HIV-1 reverse transcriptase) over the DNA ␥ polymerase and thereby assist in the design of better AIDS drugs.

EXPERIMENTAL PROCEDURES
The large and small subunits of Pol ␥ were purified as described (23,26). The protein concentration was determined spectrophotometrically at 280 nm, with the extinction coefficients 234,420 and 71,894 M Ϫ1 cm Ϫ1 for the large and small subunits, respectively. A synthetic DNA duplex, 23⅐45 bp (40), was used in which the next correct base for incorporation was dCTP. Concentrations of the oligonucleotides were estimated spectrophotometrically at 260 nm using the calculated extinction coefficients 226,750 M Ϫ1 cm Ϫ1 and 491,960 M Ϫ1 cm Ϫ1 for the 23-and 45-mer, respectively. Before annealing, the primer was 5Ј-labeled with T4 polynucleotide kinase and [␥-32 P]ATP. The dCTP and ddCTP were purchased from Sigma, and the (ϩ) and (Ϫ)3TC-TP isomers were the kind gift of Dr. Raymond Schinazi at Emory University.
Experiments to measure polymerization kinetics were performed as described (40). Rapid quench experiments were done using KinTek Instruments Model RQF-3 rapid quench-flow apparatus (41). Unless mentioned otherwise, all concentrations refer to concentrations during reactions after mixing on the quench apparatus. Typically, the experiments were done by first incubating the duplex DNA (250 nM) with pre-mixed large (67 nM) and small subunits (440 nM) in reaction buffer (50 mM Tris-HCl, 100 mM NaCl, pH 7.8). A sample of this solution (15 l) was rapidly mixed with an equal volume of a solution of dCTP analogs (at variable concentrations) and MgCl 2 (2.5 mM) in reaction buffer. The reactions were quenched by ϳ67 l of 0.5 M EDTA at pH 8.0. All reactions were performed at 37°C. Single nucleotide incorporation was monitored by extension of 5Ј-32 P-labeled 23-24-bp oligonucleotides. All reaction mixtures were analyzed on 20% denaturing polyacrylamide sequencing gels (8 M urea), imaged on a Bio-Rad GS-525 Molecular Imager System, and quantified using Molecular Analysis (Bio-Rad). Data were fitted by nonlinear regression using the program Kaleida-Graph (Synergy Software). The pre-steady-state burst experiments ( Fig. 3) were fitted to a burst equation: where A represents the amplitude of the burst that correlates with the concentration of enzyme in active form, k obsd is the observed firstorder rate constant for dNTP incorporation, and k ss is the observed steady-state rate constant.
The dissociation constant, K d , for dNTP binding to the complex of Pol ␥ and 23⅐45-bp DNA duplex is calculated by fitting the data into the following hyperbolic equation: where k pol is the maximum rate of dNTP incorporation, [dNTP] is the corresponding concentration of dNTP, and K d is the equilibrium dissociation constant for the interaction of dNTP with the enzyme-DNA complex.
The exonuclease activity was studied by measuring the rate of formation of the cleavage products in the absence of dCTP or dCTP analogs. Products formed from the early time points were plotted as a function of time. The slope of the line was divided by the active enzyme concentration in the reaction, and a k exo for exonuclease activity was obtained. The reaction was initiated by adding MgCl 2 (2.5 mM) into a solution of catalytic subunit (40 nM), accessory subunit (270 nM), and 1500 nM 24⅐45-bp DNA duplex in reaction buffer (50 mM Tris-HCl, 100 mM NaCl, pH 7.8) and quenched with 0.3 M EDTA. For exonuclease activity of the catalytic subunit, 16 nM catalytic subunit and 1000 nM 24⅐45-bp DNA duplex were used.

RESULTS
The underlying mechanisms of toxicity were investigated by examining the kinetics of the two different reactions catalyzed by human mitochondrial DNA Pol ␥: DNA-dependent DNA synthesis and 3Ј-5Ј-exonuclease cleavage. This analysis was carried out by examining the single nucleotide incorporation using synthetically prepared dCTP, ddCTP, and two modified deoxycytidine analogs, (ϩ)3TC-TP and (Ϫ)3TC-TP, opposite a template deoxyguanosine for DNA-dependent DNA polymerization by the recombinant Pol ␥ holoenzyme complex. The exonuclease cleavage was investigated by evaluating the excision of dCMP, ddCMP, (ϩ)3TC-MP, (Ϫ)3TC-MP, and dTMP from the 3Ј-terminus of the corresponding DNA 23-or 24-mer primer by the catalytic subunit alone and the holoenzyme complex of Pol ␥. All of the incorporation studies were carried out at protein concentrations high enough to maintain the holoenzyme Pol ␥ complex and on a time scale short enough to follow a single turnover of enzyme.
Using a transient kinetic approach, we determined the rate of polymerization (k pol ), the dissociation constant for the ground state nucleotide binding (K d ), and the incorporation efficiency or specificity constant (k pol /K d ) for the natural substrate, dCTP, as well as for ddCTP, (ϩ)3TC-TP, and (Ϫ)3TC-TP (shown in Fig. 1). Enzyme specificity is defined by the ratio k cat /K m , representing the apparent second-order rate constant for substrate binding. In the case of DNA polymerases, our previous work (42) has shown that the ground state binding of nucleoside triphosphates is in a rapid equilibrium, and accordingly, the K d measured under single turnover conditions would equal the K m , and the k pol would equal k cat . We employed a synthetic 23/45-mer DNA/DNA primer-template (Fig. 2) to perform a series of pre-steady-state burst and single turnover experiments.
In each case a pre-steady-state burst of product formation was observed, indicating that at saturating concentrations of dNTP, a step after the chemical step is rate-limiting (Fig. 3, A  and B). The rate-limiting step after incorporation of a single nucleotide is the release of the DNA from the enzyme (26). 2 In each panel, the solid line represents the best fit of the data to a burst equation, from which the burst amplitude, A, the observed rate of incorporation, k obsd , and the observed steadystate rate, k ss , were obtained.
A series of experiments were conducted using varying con- observed because the burst rate was comparable with the steady-state rate. Therefore, to measure the rate of incorporation, the reactions were studied under single turnover conditions with enzyme in excess of DNA as described in Johnson et al. (26) and Feng and Anderson (40).
The interaction of dCTP with the enzyme-DNA complex was assessed by fitting the concentration dependence of the burst rates to the hyperbolic equation k obsd ϭ (k pol ϫ [dNTP])/(K d ϩ [dNTP]), where k pol is the maximum rate of incorporation, [dNTP] is the concentration of the nucleoside triphosphate, and K d is the equilibrium dissociation constant for the ground state binding of dNTP with the enzyme-DNA complex. The concentration dependence of the observed polymerization rate for the incorporation of dCTP, ddCTP, (ϩ)3TC-TP, and (Ϫ)3TC-TP into a DNA/DNA 23/45-mer primer-template by Pol ␥ holoenzyme is shown in Fig. 4, A-C.
A complete summary of k pol , K d , and k pol /K d values for each dCTP analog is shown in Table I. The k pol and the K d value for dCTP are all in good agreement with previous studies (26).
A comparison of the analogs reveals that the natural substrate, dCTP, serves as the best substrate for human mitochondrial DNA polymerase, as illustrated by the observed tight binding and fast rate of polymerization. In contrast, ddCTP is incorporated 70-fold more slowly than dCTP, but binds 27-fold more tightly, so the overall incorporation efficiency is only 2.6-fold lower. The (Ϫ)3TC-TP and (ϩ)3TC-TP analogs are incorporated 125-to 350-fold more slowly than dCTP, respectively. The dCTP and (ϩ)3TC-TP share a similar high affinity for the enzyme-DNA complex, indicating chemistry or the conformational change has become rate-limiting for (ϩ)3TC-TP incorporation. In contrast, the affinity of (Ϫ)3TC-TP for the enzyme-DNA complex is 6 -8-fold weaker than that of the analog that of (ϩ)3TC-TP, containing the natural nucleoside configuration. Accordingly, Pol ␥ incorporates the (Ϫ) isomer at least 16-fold less efficiently than the (ϩ) isomer, as defined by the specificity constants summarized in Table I. This difference in incorporation efficiency is in good agreement with cell culture studies with the nucleoside form of the compounds where the (ϩ)3TC isomer has been reported to be 25-and 12-fold more toxic than the (Ϫ)3TC in terms of inhibition of cell growth and mitochondrial DNA synthesis, respectively (36).
A gel illustrating the time course for product formation during the exonuclease reaction is shown in Fig. 5. For clarity, only representative time points have been selected for dCMP (D24-ddC), ddCMP (D24-ddC), (ϩ)3TC-MP (D24-(ϩ)3TC), or (Ϫ)3TC-MP (D24-(Ϫ)3TC). A summary of the rates of exonuclease activity (k exo ) for either the catalytic subunit or the holoenzyme complex of Pol ␥ using different D24-mer substrates is shown in Table II.
The unmodified correctly base-paired primers, D23-dT and D24-dC, exhibited the fastest excision rates; however, these rates are still small and consistent with the "cost of proofreading" due to the removal of ϳ1% of the correctly paired bases at the 3Ј-end of the primers (42). Excision of the (ϩ) and (Ϫ) isomers of 3TC was 2-fold slower than removal of dCMP, and the presence of the accessory protein had minimal effect on this excision rate. The presence of the small accessory subunit of Pol ␥ decreased the exonuclease activity 5-6-fold for the DNA substrates that contained natural nucleotides (DNA 23-dT or 24-dC) at their 3Ј-ends. These observations are in contrast to a recent study indicating that the accessory subunit may stimulate the exonuclease activity (25). Despite the slow excision rate, the exonuclease should still effectively remove the chain terminators with a half-life of ϳ1 min. Although this will slow the rate of mitochondrial DNA replication, the low rate of 3TC-TP incorporation will minimize the toxic effects.
In contrast, little or no DNA cleavage was observed for D24-ddC after either 4 h with the catalytic subunit or 12 h with the holoenzyme. The insusceptibility of the 3Ј-terminal ddCMP to base excision was also reported for a 3Ј-5Ј-exonuclease with porcine liver DNA polymerase ␥ (43). Similar results have been reported in studies on yeast mitochondrial DNA polymerases, where the terminally incorporated ddTMP, AZTMP, and ddCMP were not removed by the 3Ј-5Ј-exonuclease activity of Pol ␥ (44). This striking observation may explain why inhibition of mitochondrial DNA synthesis by ddC is so profound.

DISCUSSION
Our studies have shown that (Ϫ)3TC-TP is less efficiently incorporated than (ϩ)3TC-TP into DNA by Pol ␥, and both are incorporated less efficiently than ddCTP. Moreover, the efficiency of incorporation by Pol ␥ is directly correlated with toxicity of each nucleoside analog. These observations point to incorporation by Pol ␥ into mitochondrial DNA as a major site of toxicity of nucleoside analogs. The availability of reliable preparations of Pol ␥ now affords a rapid method to screen drugs for toxicity and opens the path toward further structure/ function studies to aid in the design of less toxic analogs.
3TC is well tolerated in the clinic (28,29), and no subclinical signs of mitochondrial toxicity resulting from (Ϫ)3TC therapy for 6 months have been observed (30). Apparently to date (Ϫ)3TC has shown little evidence of mitochondrial injury (13). The unnatural L-(Ϫ) configuration of (Ϫ)3TC seems to play an important role in preserving the anti-viral activity while decreasing the inhibitory effect on the host Pol ␥. This example suggests other unnatural nucleoside analogs may show similar effectiveness combined with low toxicity.
In contrast, the effects of inhibition on mitochondrial DNA synthesis in vivo and in vitro by ddC are profound (38,45). The IC 50 value for ddC relative to the (ϩ) and (Ϫ) isomers of 3TC indicated that ddC is 180-and 2300-fold more toxic than the (ϩ) and (Ϫ) isomer, respectively (36). A number of factors may clearly come into play including uptake, transport, metabolic    activation, degradation, and removal from the system, but our studies have shown a lack of exonucleolytic removal of a 3Ј-ddC that may significantly contribute to drug toxicity. Conversely, analogs that activate the exonuclease removal by Pol ␥ may exhibit even lower toxicity, and this may provide a new avenue for exploration of less toxic drugs.
The accessory subunit appears to enhance discrimination by the proofreading exonuclease since slower rates of excision by the holoenzyme were observed for correctly base-paired DNA 23-dT-and 24dC-mers. When the 3Ј-terminus is occupied by a correctly base-paired natural dNMP with a 3Ј-OH, the accessory subunit may slow down the excision of this dNMP by facilitating the retention of the 3Ј-end at the polymerase site. Clearly, additional studies are required to identify the structural features of nucleoside analogs that may be utilized to obtain selectivity for HIV-1 RT over DNA Pol ␥.
In summary, our studies have demonstrated that the (ϩ)3TC-TP is a more potent inhibitor for human mitochondrial DNA Pol ␥ than the (Ϫ) isomer through tighter binding and faster incorporation rate into a DNA duplex, leading to greater toxicity. Using highly purified Pol ␥, there is no significant difference in the excision of the incorporated (ϩ)-and (Ϫ)3TC-MP by the 3Ј-5Ј-exonuclease activity of Pol ␥ from the 3Јterminus of a DNA primer, indicating this step likely does not contribute to the low toxicity of (Ϫ)3TC. Additionally, our studies have shown that a terminally incorporated ddCMP is not efficiently removed by the 3Ј-5Ј-exonuclease activity of Pol ␥, suggesting this may play a role in the high level of inhibition of ddC on mitochondrial DNA synthesis observed in vitro and in vivo. The overall toxicity of a nucleoside analog toward replication will be a function of the frequency of incorporation of the chain terminator and the kinetics of excision by the proofreading exonuclease relative to the rate of DNA replication required to sustain growth (Table III). We can now quantify these parameters and begin to assess the structural/functional determinants that govern nucleotide selectivity for Pol ␥ in each step.
An in-depth understanding of the mechanism of inhibition of DNA polymerase ␥ is important in defining the cytotoxicity of anti-HIV drugs toward mitochondrial replication. A knowledge of the mechanistic and structural similarities and differences between Pol ␥ and HIV-1 RT may identify key features important for selective drug incorporation by the viral HIV-1 RT but not by Pol ␥ and, therefore, facilitate the design of less toxic, viral-specific AIDS drugs.