The 3′-Azido Group Is Not the Primary Determinant of 3′-Azido-3′-deoxythymidine (AZT) Responsible for the Excision Phenotype of AZT-resistant HIV-1*

The mechanism of human immunodeficiency virus (HIV) 1 resistance to 3′-azido-3′-deoxythymidine (AZT) involves reverse transcriptase (RT)-catalyzed phosphorolytic excision of the chain-terminating AZT-5′-monophosphate (AZTMP). Primers terminated with AZTMP are generally better substrates for this reaction than those terminated with 2′,3′-dideoxynucleoside-5′-monophosphate (2′,3′-ddNMP) analogs that lack a 3′-azido moiety. This led to the hypothesis that the 3′-azido group is a major structural determinant for maintaining the primer terminus in the appropriate site for phosphorolytic excision of AZTMP by AZT-resistant (AZTR) RT. To test this hypothesis, we evaluated the incorporation, phosphorolytic excision, and antiviral activity of a panel of 3′-azido-2′,3′-ddN including 3′-azido-2′,3′-ddA (AZddA), 3′-azido-2′,3′-ddC (AZddC), 3′-azido-2′,3′-ddG (AZddG), AZT, and 3′-azido-2′,3′-ddU (AZddU). The results indicate that mutations correlated with resistance to AZT (D67N/K70R/T215F/K219Q) confer resistance to the 3′-azidopyrimidine nucleosides (AZddC, AZT, and AZddU) but not to the 3′-azidopurine nucleosides (AZddA and AZddG). The data suggest that the presence of a 3′-azido group on the 3′-terminal nucleotide of the primer does not confer increased phosphorolytic excision by AZTR RT for all 3′-azido-ddNMP analogs. Thus, the 3′-azido group cannot be the only structural determinant important for the enhanced phosphorolytic excision of AZTMP associated with HIV resistance to AZT. Other structural components, such as the base, must play a role in defining the specificity of the excision phenotype arising from AZT resistance mutations.

The replication of human immunodeficiency virus (HIV) 1 1 is dependent on the enzymatic activities of reverse transcriptase (RT), an RNA-and DNA-dependent DNA polymerase encoded by the viral pol gene. RT synthesizes the double-stranded proviral DNA precursor from the viral (ϩ) RNA genome (1). Because of its essential role in HIV-1 replication, RT is a major target for antiretroviral drug development and two structurally dissimilar classes of RT inhibitors, termed nucleoside RT inhibitors (NRTI) and nonnucleoside RT inhibitors, are routinely used for the clinical treatment of HIV-1-infected individuals (2). NRTI are 2Јdeoxyribonucleoside analogs that usually lack a 3Ј-OH group on the ribose moiety. After intracellular conversion to the active 5Ј-triphosphate form, NRTI-TP inhibit DNA synthesis by competing with the natural nucleotides both for recognition by RT as a substrate and by incorporation into the nascent viral DNA chain (3). Incorporation of an NRTI into the nascent viral DNA chain by RT results in termination of DNA synthesis.
As is the case with all antiretroviral agents, the emergence of drug-resistant HIV-1 variants limits the efficacy of NRTI. Most NRTI-resistant viruses isolated from patients treated with nucleoside analogs have mutations in the pol gene (4). To date, two major phenotypic mechanisms have been proposed to account for HIV-1 resistance to NRTIs (5,6). One mechanism is NRTI discrimination in which the mutant RT can preferentially incorporate the natural dNTP over the NRTI-TP analog. The second mechanism involves pyrophosphate-or ATP-dependent phosphorolytic excision of the chain-terminating NRTI from the 3Ј-end of the primer by HIV-1 RT (7)(8)(9). This excision phenotype has primarily been associated with AZT resistance. In this regard, HIV-1 RT having various combinations of AZT resistance mutations (M41L, D67N, K70R, L210W, T215(F/Y), and K219Q) shows a significantly enhanced rate of excision of AZT-5Ј-monophosphate (AZTMP) compared with the wild-type (WT) enzyme (7,9,10).
For HIV-1 RT to effectively excise AZT from the 3Ј-end of a primer, the chain-terminating AZTMP must reside in the nucleotide-binding site (N-site) of the active site of the enzyme (10,11). Under physiological conditions, the binding of the next correct dNTP can drive the terminating nucleotide into the primer-binding site (P-site) resulting in the formation of a dead-end complex (8,9,10). Formation of this complex prevents the excision reaction from occurring. Several studies have shown that the excision of AZTMP by AZT-resistant (AZT R ) RT is much less sensitive (Ͼ50-fold) to inhibition by the next correct dNTP than other NRTI analogs that lack a 3Ј-azido group (10,(12)(13)(14). These data suggest that the 3Ј-azido group might interfere either with the ability of the primer 3Ј-terminal AZTMP to translocate from the N-to P-site, or with the ability of the AZTMP moiety to occupy the P-site. It has thus been proposed that the 3Ј-azido group of the AZTMP-terminated primer is the primary structural determinant of AZT responsible for the excision phenotype of AZT R RT or HIV-1 (10,11). To evaluate this model, we examined the incorporation, phosphorolytic excision, and antiviral activity of a panel of 3Ј-azido-2Ј,3Ј-ddN compounds including 3Ј-azido-2Ј,3Ј-ddA (AZddA), 3Јazido-2Ј,3Ј-ddC (AZddC), 3Ј-azido-2Ј,3Ј-ddG (AZddG), 3Ј-azido-2Ј,3Ј-ddU (AZddU), and AZT. Our results indicate that AZT resistance mutations do not confer significant cross-resistance of HIV-1 to 3Ј-azidopurine nucleosides (AZddA and AZddG). Furthermore, a 3Ј-azido group on a 3Ј-terminal purine nucleotide does not enhance phosphorolytic excision by AZT R RT, suggesting that other structural factors must play a role in defining the specificity of the excision phenotype arising from AZT resistance mutations.

MATERIALS AND METHODS
Reagents-WT RT and AZT R HIV-1 RT with mutations D67N/K70R/ T215F/K219Q were overexpressed and purified to homogeneity as previously described (15). The protein concentration of the purified enzymes were determined spectrophotometrically at 280 nm using an extinction co-efficient (⑀ 280 ) of 260,450 M Ϫ1 cm Ϫ1 . The active site concentrations of RT were calculated from pre-steady-state burst experiments (16). Burst amplitudes of 59 and 52% were calculated for the WT and AZT R RT enzymes, respectively. All experiments described below were performed using corrected RT active site concentrations. The triphosphate forms of all 3Ј-azido-2Ј,3Ј-ddN were purchased from Tri-Link Biotechnologies (San Diego, CA). The 3Ј-azido-2Ј,3Ј-ddN were also purchased from TriLink Biotechnologies, except for 3Ј-azido-2Ј,3Ј-ddC, which was synthesized as previously described (17). ATP was from Roche Diagnostics.
DNA Substrates-DNA oligomers were obtained from Invitrogen. A 20-nucleotide (nt) DNA primer (5Ј-TCGGGCGCCACTGCTAGAGA-3Ј) and a 57-nt DNA template (5Ј-CTCAGACCCTTTTAGTCAGAATG-GAAANTCTCTAGCAGTGGCGCCCGAACAGGGACA-3Ј) were used in all experiments. Four DNA templates were synthesized, each of which contained a different nucleotide at position 30 (indicated by N in the 57-nt DNA template sequence). This strategy allowed us to evaluate the kinetics of single nucleotide incorporation or excision for all 3Ј-azido-2Ј,3Ј-ddNTP in essentially the same sequence context and using the same 20-nt primer.
Pre-steady-state Kinetic Experiments and Product Analysis-The 20-nt DNA primer was 5Ј-radiolabeled with [␥-32 P]ATP (PerkinElmer Life Sciences) and T4 polynucleotide kinase (Amersham Biosciences) according to the supplier's instructions. The labeled primer was then resolved by denaturing polyacrylamide gel electrophoresis using 7 M urea, 16% polyacrylamide gels, followed by elution of the 21-nt primer from the gel. The 5Ј-32 P-labeled primer was then annealed to the 57-nt DNA template (1.5:1 molar ratio of primer:template) by heating to 90°C, followed by slow cooling to ambient temperature.
A rapid quench instrument (Bio-Logic SFM-400, Claix, France) was used for pre-steady-state kinetic experiments with reaction times ranging from 10 ms to 7 s. A typical experiment was performed at 37°C in 50 mM Tris-HCl (pH 8.0) containing 50 mM KCl, 10 mM MgCl 2 and varying concentrations of nucleotide (dNTP). HIV-1 RT was preincubated with the DNA substrate prior to rapid mixing with dNTP and divalent metal ions to begin the reaction. The final concentrations in the reaction mixture were 100 nM RT active sites and 25 nM of the template/ primer (T/P). After varying times the reactions were quenched with 0.5 M EDTA. Quenched samples were mixed with an equal volume of gel loading buffer (98% deionized formamide, 10 mM EDTA, and 1 mg/ml each of bromphenol blue and xylene cyanol), denatured by heating at 85°C for 5 min, followed by electrophoretic resolution of the product on a 7 M urea, 16% polyacrylamide gel. The disappearance of the 20-nt substrate primer and the formation of the 21-nt product were analyzed using a Bio-Rad GS525 Molecular Imager (Bio-Rad).
Data Analysis-Kinetic data were analyzed by nonlinear regression using Sigma Plot software (Jandel Scientific) with the appropriate equations as described below. The apparent burst rate constant (k obs ) for each particular concentration of dNTP was determined by fitting the time courses for the formation of the 21-nt product with the equation, where A represents the burst amplitude. The turnover number (k pol ) and apparent dissociation constant for dNTP (K d ) were then obtained by plotting the apparent catalytic rates, k obs , against dNTP concentrations and fitting the data to the hyperbolic Equation 2, Assay of RT-catalyzed Phosphorolysis-The 20-nt DNA primer was 5Ј-end labeled with [␥-32 P]ATP and then annealed to one of the four DNA templates described above. 3Ј-Azido-NRTI was added to the primer 3Ј terminus by incubation with WT RT and 100 M of the appropriate 3Ј-azido-2Ј,3Ј-ddNTP for 30 min at 37°C. The 32 P-labeled chain-terminated 21-nt primer was purified by elution of the appropriate band following resolution by 7 M urea, 16% acrylamide denaturing gel electrophoresis. The purified chain-terminated primer was then annealed to the appropriate DNA primer for use in phosphorolysis experiments.
The phosphorolytic removal of 3Ј-azido-2Ј,3Ј-ddNMP was assayed by incubating WT or AZT R RT with the chain-terminated T/P of interest (4:1 ratio of RT:T/P) in 50 mM Tris-HCl (pH 8.0), 50 mM KCl. The reaction was initiated by the addition of 3.0 mM ATP and 10 mM MgCl 2 . Inorganic pyrophosphatase (0.01 unit; Sigma) was included. Aliquots were removed at the defined times and quenched with equal volumes of gel loading dye (98% deionized formamide, 10 mM EDTA, and 1 mg/ml each of bromphenol blue and xylene cyanol). The 21-nt substrate and 20-nt products were separated by denaturing gel electrophoresis and the reduction in substrate and formation of product were analyzed by phosphorimaging. Data were fit to the following single-exponential equation to determine the apparent rate of ATP-mediated excision, where A represents the amplitude for product formation.

Electrophoretic Analysis of RT Polymerization Products Formed under Continuous DNA Polymerization Conditions-A heteropolymeric
RNA-dependent DNA polymerase T/P was prepared as described previously (18) using the T7 polymerase RNA transcript from AccI-linearized plasmid pHIV-PBS as template and the synthetic 20-nt DNA oligonucleotide primer described above.
DNA polymerization reactions were carried out by incubating 200 nM (active site) WT and AZT R RT with 50 nM heteropolymeric T/P in 50 mM Tris-HCl (pH 8.0), 50 mM KCl for 5 min prior to the addition of 5 M of each dNTP, 1 M of the desired 3Ј-azido-2Ј,3ЈddNTP, 3.0 mM ATP, 10 mM MgCl 2 , and 0.01 unit of inorganic pyrophosphatase. After defined incubation periods, aliquots were removed from the reaction tube and quenched with equal volumes of gel loading dye. Products were separated by denaturing gel electrophoresis and quantified by phosphorimaging.
Cells, Viruses, and Antiviral Susceptibility Assays-MT-2 cells (obtained from the AIDS Research and Reference Reagent Program, NIAID, National Institutes of Health) were maintained in RPMI 1640 supplemented with 10% fetal bovine serum, 10 mM HEPES buffer, 50 IU of penicillin/ml, and 50 g of streptomycin/ml. WT and AZT R virus stocks, generated from plasmid DNA encoding an infectious proviral clone of HIV-1 LAI (19,20), were prepared in MT-2 cells and subjected to titer determination as previously described (21). Antiviral activity was determined by inoculating MT-2 cells with WT or AZT R virus stocks (multiplicity of infection, 0.01 50% tissue culture infective dose/cell) followed by incubation in the presence of 2-or 3-fold serial dilutions of 3Ј-azido-2Ј,3Ј-ddN. The amount of HIV-1 p24 antigen released into the culture supernatant was determined 7 days postinfection and used to determine the concentration of each compound required to reduce p24 antigen production by 50% (IC 50 ).

RESULTS
Pre-steady-state Incorporation of 3Ј-Azido-2Ј,3Ј-ddNTPs by WT and AZT R HIV-1 RT-Pre-steady-state kinetic analyses of 3Ј-azido-2Ј,3Ј-ddNTP incorporation by WT and AZT R HIV-1 RT were undertaken to elucidate the detailed interactions of these nucleotide analogs with the RT polymerase active site. Fig. 1 illustrates the results of a typical experiment. These experiments were used to define the maximum rate of incorporation (k pol ), dissociation constant (K d ), and catalytic efficiency of incorporation (k pol /K d ) for each nucleotide analog for both the WT and AZT R enzymes (Table I). Both WT and AZT R RT could readily incorporate all the 3Ј-azido-2Ј,3Ј-ddNTP. The relative catalytic efficiencies of incorporation (k pol /K d ) were AZdATP Ͼ AZTTP Х AZddUTP Ͼ AZddCTP Ͼ AZddGTP. Consistent with previously published data (22,23), both the WT and AZT R enzymes incorporated AZTTP with near equivalent catalytic efficiencies. Similarly, the catalytic efficiencies of the WT and AZT R enzymes were nearly equivalent for the other 3Ј-azido-2Ј,3Ј-ddNTPs used in this study (Table I).
Pre-steady-state Excision of 3Ј-Azido-2Ј,3Ј-ddNMP by WT and AZT R HIV-1 RT-Transient kinetic analyses of ATP-mediated excision by WT and AZT R HIV-1 RT were carried out to determine the apparent rates for phosphorolytic removal (k excision ) of the 3Ј-azido-2Ј,3Ј-ddNMP analogs (Table II). The relative rates of ATP-mediated phosphorolysis by WT and AZT R RT (in the absence of dNTP) were AZddAMP Ͼ AZdd-UMP Ͼ AZTMP Ͼ AZddCMP Ͼ AZddGMP. In contrast to the incorporation data, there were significant differences in the rates of ATP-mediated phosphorolytic excision of the various 3Ј-azido-2Ј,3Ј-ddNMP by WT and AZT R RT. Consistent with previously published data (9, 10, 13, 24), AZT R RT was more effective then the WT enzyme in removing AZTMP from a chain-terminated primer. The AZT R enzyme was also significantly more efficient (9-fold) in removing AZddUMP than the WT enzyme. However, AZT R RT showed only a moderately improved ability (ϳ3-fold) to remove AZddAMP and AZddCMP compared with the WT enzyme. AZT R and WT RT were equally effective at removing AZddGMP (Table II).
Previous studies have shown that the next complementary nucleotide can inhibit ATP-mediated chain terminator removal (8 -10, 12). Furthermore, it has been suggested that the relative insensitivity of AZTMP excision to inhibition by the next nucleotide compared with other NRTI-MP may account for the greater degree of resistance to AZT conferred by TAMs than to other NRTI (10,12). To determine the effect of AZT resistance mutations on the magnitude and nature of inhibition by the next complementary nucleotide during the phosphorolytic removal of 3Ј-azido-2Ј,3Ј-ddNMP analogs, 50% inhibitory concentrations (IC 50 ) of the next complementary nucleotide for the ATP-dependent removal of the primer 3Ј-terminal NRTI-MP were determined for each of the 3Ј-azido-2Ј,3Ј-ddNMP analogs (Table II). For both the WT and AZT R enzymes, the inhibitory concentrations of the next complementary dNTP were substantially higher than putative intracellular dNTP concentrations. IC 50 values for the AZT R enzyme were ϳ3-fold higher than for the WT enzyme.
Comparison of the Incorporation and Excision Efficiencies of 3Ј-Azido-2Ј,3Ј-ddNTP by WT and AZT R RT-During HIV-1 replication there are multiple opportunities for RT to incorporate and excise NRTI-MP. The capacity for WT or mutant RT to excise any given NRTI is therefore best considered as the ratio of the efficiency of RT-catalyzed NRTI excision to the catalytic efficiency of RT-catalyzed NRTI incorporation. In this context, analysis of the data from Tables I and II shows that AZT R RT exhibits a preference to incorporate and then excise AZd-dUTP Ͼ AZTTP Ͼ AZddCTP Ͼ AZddATP Ͼ AZddGTP (Fig. 2). Of note, AZT R RT showed no increased ability to excise AZdd-GMP compared with the WT enzyme. It should be noted that the catalytic efficiency of ATP-mediated 3Ј-azido-2Ј,3Ј-ddNMP excision was calculated by dividing the observed k excision (Table  II) by the concentration of ATP used in the assay (3000 M), rather than the K D for ATP. We used this approach because difficulty in obtaining saturating ATP concentrations in these assays precludes accurate estimation of the K D for ATP. However, we feel our approach provides a reasonable comparative estimation of the catalytic efficiency of nucleotide excision because previous studies have suggested that there is no significant difference in the affinity for binding of ATP by WT and AZT R reverse transcriptase (25).
To further evaluate the differences in incorporation and excision of the various 3Ј-azido-2Ј,3Ј-ddNTP, we analyzed steadystate DNA synthesis by both WT and AZT R RT in the presence of each of the 3Ј-azido-2Ј,3Ј-ddNTP and 3 mM ATP, using a heteropolymeric RNA template corresponding to the HIV-1 sequence used for (Ϫ) strong stop DNA synthesis, primed with a DNA oligonucleotide (7). The 173-nt incorporation events needed to produce the full-length DNA product in this assay system allow multiple 3Ј-azido-2Ј,3Ј-ddNTP incorporation and excision events during the formation of the full-length final product. Fig. 3 shows that in the presence of 3 mM ATP, AZT R RT synthesized significantly greater amounts of full-length DNA than WT RT in reactions containing AZTTP (Fig. 3E) or AZdd-UTP (Fig. 3F), and lesser but still significant amounts of fulllength DNA in reactions containing AZddCTP (Fig. 3D). Similarly, much greater levels of small termination products were noted in reactions catalyzed by WT RT in the presence of the 3Ј-azidopyrimidine compounds compared with AZT R RT. In contrast, there were essentially no differences in either the amount of full-length final product or premature termination products produced by WT and AZT R RT in the presence of AZddATP (Fig. 3B) or AZddGTP (Fig. 3C).
Antiviral Activity of 3Ј-Azido-2Ј,3Ј-ddN-WT or AZT R virus (encoding the mutations D67N/K70R/T215F/K219Q in RT) were tested for susceptibility to inhibition by 3Ј-azido-2Ј,3Ј-ddN in virus spread assays in lymphocytoid MT2 cells. The antiviral activity of the various 3Ј-azido-2Ј,3Ј-ddN was consistent with the enzyme studies, in that AZT R virus showed demonstrable resistance to the 3Ј-azidopyrimidine nucleosides AZT and AZddC, but not to the 3Ј-azidopurine nucleosides AZddA or AZddG (Table III). In our experiments, AZddU did not exhibit antiviral activity in the MT2 cell line at concentrations up to 200 M. Previous studies have shown that AzddU is not phosphorylated in certain cells, and hence is not effective against HIV in these cell systems (26).

DISCUSSION
Antiretroviral therapy with thymidine nucleoside analogs such as AZT or d4T selects for several mutations (generally referred to as thymidine analog mutations or TAMs) in the RT gene that confer high-level resistance to AZT, and moderate levels of resistance to d4T (27)(28)(29). It is increasingly evident that TAMs are associated with varying degrees of cross-resistance to other "non-thymidine" NRTI including abacavir, ddI, ddC, and tenofovir (30 -33). TAMs in RT do not enable discrimination between the natural dNTP substrate and the analogous NRTI-TP during mutant RT-catalyzed DNA synthesis (22,23). Instead, TAMs increase the ability of RT to remove or excise the incorporated NRTI by a phosphorolytic enzymatic reaction (7-10).
The 3Ј-terminal nucleotide of the growing viral DNA strand resides in two different positions in the polymerase active site of the RT-nucleic acid complex during polymerization (10,11,33). The N-site, or the nucleotide site, is that where the primer 3Ј-terminal nucleotide is located immediately after addition of a dNTP or NRTI-TP. The P-site is that in which the primer 3Ј terminus is moved by a single register from the N-site. This translocation event positions the next template base to allow hydrogen bond stabilization with the base of the incoming cognate dNTP, thereby facilitating formation of the appropriate ternary complex for additional dNTP incorporation.
RT-catalyzed phosphorolytic excision of an NRTI-MP from the primer 3Ј terminus requires that the terminal NRTI reside in the N-site (10,11). Binding of the next complementary dNTP seems to promote translocation of the primer 3Ј terminus into the P-site (34). If the 3Ј-terminal nucleotide is an NRTI, this translocation effectively forms a "dead-end complex" in which phosphorolytic excision cannot occur (8,9), because phosphodiester bond formation is precluded. Excision of AZTMP by HIV-1 AZT R RT is much less sensitive to inhibition by the next complementary dNTP
The inability of AZT R RT to efficiently excise 3Ј-azido-2Ј,3Ј-ddAMP and 3Ј-azido-2Ј,3Ј-ddGMP in the steady-state DNA synthesis experiments was not a result of these analogs being more sensitive to dead-end complex formation by the next complementary nucleotide. AZT R RT showed ϳ3-5-fold increases in IC 50 relative to WT RT for inhibition of all of the 3Ј-azido-dNMPs evaluated (Table II). In all cases, the levels of the next complementary dNTP needed to inhibit AZT R RTmediated excision were much greater than putative intracellular dNTP concentrations (35). HIV-1 susceptibility to the 3Ј-azido-2Ј,3Ј-ddN (Table III) was consistent with biochemical data. AZT R virus exhibited at least 10-fold resistance to AZT and AZddC. AZddU was inactive in the MT2 cell line used in these studies, possibly because of inadequate intracellular conversion to the triphosphate (25). Interestingly, AZT R HIV showed no cross-resistance to either AzddA or AZddG. This result suggests that 3Ј-azidopurine nucleoside analogs that exhibit good antiviral activity and cytotoxicity profiles may be useful for the treatment of AZTresistant HIV.

3Ј-Azido Group and RT-catalyzed Nucleotide Excision
Our data clearly show that the nucleotide base component, not just the sugar 3Ј-azido moiety, is important for the excision phenotype. In this regard, TAMs in RT enable the enzyme to efficiently excise chain-terminating 3Ј-azidopyrimidines, but are less effective for excision of 3Ј-azidopurines. These data suggest that careful consideration be given to the choice of NRTI in experiments characterizing the phosphorolysis phenotype conferred by specific combinations of TAMs.