Non-nucleoside Inhibitors of HIV-1 Reverse Transcriptase Inhibit Phosphorolysis and Resensitize the 3 (cid:1) -Azido-3 (cid:1) -deoxythymidine (AZT)-resistant Polymerase to AZT-5 (cid:1) -triphosphate*

Removal of 3 (cid:1) -azido-3 (cid:1) deoxythymidine (AZT) 3 (cid:1) -azido-3 (cid:1) -deoxythymidine 5 (cid:1) -monophosphate (AZTMP) from the terminated primer mediated by the human HIV-1 reverse transcriptase (RT) has been proposed as a relevant mechanism for the resistance of HIV to AZT. Here we compared wild type and AZT-resistant (D67N/K70R/ T215Y/K219Q) RTs for their ability to unblock the AZTMP-terminated primer by phosphorolysis in the presence of physiological concentrations of pyrophosphate or ATP. The AZT-resistant enzyme, as it has been previously described, showed an increased ability to unblock the AZTMP-terminated primer by an ATP-de-pendent mechanism. We found that only mutations in the p66 subunit were responsible for this ability. We also found that three structurally divergent non-nucleoside reverse transcriptase inhibitor (NNRTI), nevirapine, TIBO, and a 4-arylmethylpyridinone derivative, were able to inhibit the phosphorolytic activity of the enzyme, rendering the AZT-resistant RT sensitive to AZTTP. The 4-arylmethylpyridinone derivative proved to be about 1000-fold more potent in inhibiting phosphorolysis than nevirapine or TIBO. Moreover, combinations of AZTTP with NNRTIs exhibited an exceptionally high degree of synergy in the inhibition of AZT-resistant enzyme only when ATP or PP i were present, indicating that inhibition of phosphorolysis was responsible for the synergy found in the combination. Our results not only demonstrate the importance of phosphorolysis con-cerning HIV-1 RT resistance to AZT but also point to the implication of this activity in the strong synergy found in some combinations of NNRTIs with AZT. The effect of NNRTIs on ATP binding is difficult to deter-mine. Direct binding assays using labeled ATP are not possible, because millimolar concentrations of the nucleotide are needed, whereas the RT concentration is in the nanomolar to low mi-cromolar range. We tested by an indirect approach whether the concentration of ATP or PP i affects the inhibition caused by 4-arylmethylpyridinone. Useful ATP concentrations were 1–9 m M , because higher concentrations result in substrate inhibition of the phosphorolytic reaction. In the range 0–9 m M ATP, the IC 50 for 4-arylmethylpyridinone in the RT-catalyzed phosphorolysis reaction only increased about 1.5-fold. The inhibition of DNA polymerization by NNRTI in the absence of AZTTP was also slightly sensitive to the presence of ATP, with the IC 50 values increasing about 1.5–2-fold. These results suggest that NNRTIs do not exert their by impacting on ATP binding, although some interference with ATP may be present. carried out as described under Procedures” in the absence or presence of 3 m M ATP, using AZT enzyme. Four concentrations of 4-arylmethylpyridinone (0.025, 0.05, 0.1, and 0.2 (cid:3) M ) were combined over a range of AZTTP concentrations, providing a matrix of data. The reverse of the relative activity of each combination was then plotted against the concentration of 4-arylmethylpyridinone (24). The concentrations of AZTTP were selected to obtain total inhibition values in the range 10–90%. A , no ATP. The reactions were incubated in buffer A containing 3 n M poly(rA)-dT 20 , 20 (cid:3) M dTTP, 10 n M AZT r -RT at 37 °C and started by adding 10 m M MgCl 2 . Afte r 1 h ofincubation at 37 °C, the reactions were quenched by the addition of EDTA, and 15 (cid:3) l of the mixture were spotted onto a DE81 paper, washed, and counted. The concentrations of AZTTP were 0 ( E ), 0.0125 ( (cid:1) ), 0.025 ( (cid:2) ), 0.05 ( f ), and 0.1 (cid:3) M ( ‚ ). B , 3 m M ATP. The experiments were performed as before, but 3 m M ATP was present during the incubation. The concentrations of AZTTP were 0 ( E ), 0.125 ( (cid:1) ), 0.25 ( (cid:2) ), 0.5 ( f ), and 1 (cid:3) M ( ‚ ).

Human immunodeficiency virus, type 1 (HIV-1) 1 reverse transcriptase (RT) is responsible for the conversion of singlestranded viral RNA into double-stranded DNA prior to integration into the genome of the human host. Numerous compounds that inhibit the DNA polymerase activity of RT have been described. They can be divided into two broad classes. The first group, that of nucleoside analogs, includes dideoxynucleoside compounds, such as 2Ј,3Ј-dideoxycytidine and AZT, that inhibit viral replication by acting as chain terminators of DNA synthesis (1). The second group, the non-nucleoside reverse transcriptase inhibitors (NNRTI), includes a large number of structurally dissimilar hydrophobic compounds that bind to a site on the RT palm subdomain adjacent to but distinct from the polymerase active site (2).
The FDA-approved HIV-1 therapies involve drugs that inhibit two viral enzymes, reverse transcriptase and protease. AZT was the first drug approved against HIV-1 and is still widely used in combination with other antiretroviral drugs. The prolonged clinical use of this nucleoside analog in monotherapy gives rise to highly resistant viruses containing mutations in the RT enzyme, D67N, K70R, T215F/Y, K219E/Q (3), and, in some cases, M41L and L210W. Viruses carrying at least four mutations are more than 100-fold less sensitive to AZT than wild type virus in cell culture. Although the genotype for AZT resistance is well characterized, it has been impossible to detect resistance to AZT 5Ј-triphosphate (AZTTP) in polymerization assays using recombinant RTs carrying these mutations. Recently, it has been demonstrated that HIV-1 RT can remove some 3Ј-terminal chain-terminating residues from blocked DNA using either pyrophosphate or ATP as substrate (4,5). The prevailing hypothesis nowadays is that mutations conferring resistance to AZT do not interfere with the incorporation of the inhibitor but increase the excision rate of AZTMP from the 3Ј-end of the nascent DNA (4,6,7).
It has been reported that the thiocarboxanilide non-nucleoside inhibitor UC781, in combination with AZTTP, synergistically inhibits the replication of AZT-resistant HIV-1 virus (8). Because UC781 was also able to block the pyrophosphorolytic reaction, it has been suggested that this inhibition could enhance the effect of AZT and might be a relevant mechanism for the delayed development of resistance to combinations of AZT plus UC781 in antiviral assays. Our work was aimed to find the biochemical mechanism mediating the synergy between AZT and NNRTIs obtained with the AZT-resistant RT. At present it is not known whether the inhibition of the pyrophosphorolytic activity is specific for UC781, because no other NNRTI has been reported to block this activity. Moreover, the effect of NNRTIs on ATP-dependent phosphorolysis has not been tested. We therefore analyzed the effect of two structurally dissimilar compounds, nevirapine and TIBO 82913 (9, 10), along with a NNRTI related to both HEPT and pyridinone, the 3-dimethylamino-4-(3,5-dimethylbenzyl)-5-ethyl-6-methylpyridin-2(1H)-one (compound 1) called 4-arylmethylpyridinone for short (Scheme 1) (11), on ATP and PP i -dependent phosphorolysis, using wild type and AZT-resistant enzymes. We found that all NNRTI were able to inhibit the phosphorolytic activity of the enzyme, although with different potencies.
Furthermore, we analyzed the interaction between AZTTP and NNRTIs with the AZT-resistant enzyme. Because both groups of inhibitors bind to different sites on RT in a nonexclusive manner (12), combinations of nucleoside analogs and non-nucleoside inhibitors might have a potential synergistic inhibitory effect on HIV-1 RT. Synergistic inhibition of HIV replication in cell culture has been reported for many combinations of nucleoside and non-nucleoside RT inhibitors (13)(14)(15)(16). Nevertheless, several other studies have shown that the same combinations showed no synergy in inhibiting RT activity in vitro (17)(18)(19)(20)(21). In this report we show that combinations of AZTTP with NNRTIs show an exceptionally high degree of synergy in inhibiting AZT-resistant enzyme only when ATP or PP i are present, showing that inhibition of phosphorolysis may be implicated in the mechanism of this interaction in vivo. The relevance of these findings in combined anti-HIV chemotherapy is discussed.

EXPERIMENTAL PROCEDURES
Preparation of HIV-1 RT-Recombinant p66/p51 wild type RT was obtained as follows. The p66 plasmid cloned in Escherichia coli JM109 (22) was used to generate p51 by introducing two stop codons in amino acid positions 441 and 442 using a mutagenesis kit from Stratagene. Both clones were grown and induced by isopropyl-␤-D-thiogalactopyranoside separately, and the cell pellets were mixed maintaining an excess of the p51 subunit. The cells were disrupted in a French press and purified by immobilized metal affinity chromatography as described (20). Excess of monomeric p51 was removed by gel filtration through a Superdex 200 column (Amersham Biosciences) (23). AZTresistant RT (AZT r -RT) contained four amino acid substitutions, D67N, K70R, T215Y, and K219Q, introduced in the genes of both subunits using the same mutagenesis kit. The presence of expected mutations was verified by the complete sequencing of all RT clones. All of the enzyme preparations were homogeneous as judged by gel filtration and SDS-PAGE.
PAGE Analysis of Polymerization Assay-The DNA oligonucleotide termed d21 (5Ј-GGGGATCCTCTAGAGTCGACC-3Ј) was labeled with [␥-32 P]ATP at the 5Ј-end and then annealed to a 36-nucleotide RNA template called r36 (5Ј-AAAAAAAAAAAAAAAGGUCGACUCUA-GAGGAUCCCC-3Ј). This primer-template was incubated with the enzyme in the presence of 10 M dTTP in buffer A (100 mM NaCl, 50 mM Tris-HCl, 1.25 mM EGTA, 0.5 mM EDTA, 0.05% Nonidet P-40 at pH 8) and started by adding 10 mM MgCl 2 (pH 8) in a final volume of 25 l. After 1 h of incubation at 37°C, the reactions were quenched by the addition of an equal volume of 25 l of loading buffer (90% formamide, 10 mM EDTA, 0.025% bromphenol blue, and 0.025% xylene cyanol). The samples were analyzed by denaturing PAGE using 12% polyacrylamide gels containing 7 M urea. The electrophoretically resolved products were visualized by autoradiography, and the amount of total product obtained was calculated by densitometric quantification of all oligonucleotides longer than 35 nucleotides using Imagemaster software (Amersham Biosciences).
Preparation of AZTMP-terminated Primer-Oligonucleotide d21 (1 nmol) annealed with r36 (2 nmol) was incubated with 10 nM RT and 100 M AZTTP in 100 mM NaCl, 50 mM Tris-HCl, 0.05% Nonidet P-40, and 4 mM MgCl 2 (pH 8) for 4 h at 37°C in a final volume of 200 l. The chain-terminated primer-template was precipitated with ethanol, resuspended in 90% formamide and 10 mM EDTA, and finally purified by denaturing PAGE using a 12% polyacrylamide gel containing 7 M urea. The bands were visualized on a TLC silica gel plate under UV light, and the band corresponding to the terminated primer was cut out. The oligonucleotide was eluted from the gel in 0.3 M sodium acetate and 2 mM EDTA (pH 7.5), precipitated with ethanol, and dissolved in 25 l of 10 mM Tris-HCl, 1 mM EDTA (pH 8).
Phosphorolysis Assay-The d21-AZTMP oligonucleotide was labeled with [␥-32 P]ATP at the 5Ј-end and then annealed to a 39-nucleotide RNA template called r39 (5Ј-AAAAAAAAUAAAAGAACAGGUCGACU-CUAGAGGAUCCCC-3Ј). This primer-template was incubated with the enzyme in the presence of the indicated concentrations of PP i or ATP in Buffer B (50 mM NaCl, 50 mM Tris-HCl, 0.05% Nonidet P-40, pH 8), and the reaction was started by adding 10 mM MgCl 2 (pH 8) in a final volume of 25 l. After incubation at 37°C, the reactions were stopped by the addition of the same volume of loading buffer (90% formamide, 10 mM EDTA, 0.025% bromphenol blue, and 0.025% xylene cyanol). The samples were analyzed by denaturing PAGE using 12% polyacrylamide gels containing 7 M urea. The bands were visualized by autoradiography and quantified by densitometry as before. Because RT eliminates several nucleotides from the terminated primer, the amount of total product obtained was calculated by densitometric quantification of all oligonucleotides smaller than 22 nucleotides.
Combination Assays-The enzyme was incubated in Buffer A with 3 nM poly(rA)-dT 20 , 10 M [␣-33 P]dTTP in the presence or absence of 3 mM ATP or 250 M PP i in a final volume of 50 l. The reactions were started with 10 mM MgCl 2 , and after 1 h of incubation at 37°C, the reactions were quenched by adding 5 l of EDTA 0.5 M. Fifteen l of the mixture were spotted onto a DE81 (Whatman), washed three times with 0.5 M Na 2 HPO 4 (pH 7.5), dried, and counted. The interaction index was calculated as explained previously (20,24). Briefly, dose-response curves for each inhibitor alone were obtained within a wide range of effects by fitting experimental data to Equation 1 by unweighted nonlinear regression.
where f is the fractional inhibition, D represents the concentration (dose) of the inhibitor when tested alone, IC 50 is the concentration of the inhibitor giving 50% of inhibition, and m is a parameter giving the sigmoidicity of the dose-response curve, using the commercial available fitting program Grafit (Erithacus software). Interaction between inhibitors was evaluated by means of the interaction index (I).
where D 1 and D 2 are the concentrations of inhibitors 1 and 2 individually producing the same effect as the combination (d 1 ϩ d 2 ) (24, 25), calculated from Equation 1. When I ϭ 1, agents in the combination do not interact; if I Ͼ 1, the combination is antagonistic, and if I Ͻ 1 the combination is synergistic.

Rate of Phosphorolysis by wt RT and Mutant
AZT r -RT-AZT resistance has been related to an increased phosphorolytic activity of the mutant RT, leading to the removal of AZTMP from a chain-terminated primer (4, 5, 7). We measured the rate of phosphorolysis for wt and resistant RTs (Fig. 1A). At saturating RT concentrations, excision of AZT followed apparent first order kinetics. The addition of 150 M PP i resulted in a similar rate of pyrophosphorolysis for both wt and resistant enzymes, with rate constants of 0.081 Ϯ 0.009 and 0.074 Ϯ 0.007 min Ϫ1 , respectively (Fig. 1B). Interestingly, when 3 mM ATP was used as substrate instead of PP i , wt RT showed a very slow removal of the chain terminator, with a rate constant of 0.001 Ϯ 0.0002 min Ϫ1 . Under these conditions, this reaction was more efficient for resistant RT, which showed a rate constant of 0.011 Ϯ 0.001 min Ϫ1 .
The experimental procedure employed to purify the heterodimeric RT (see "Experimental Procedures") allowed us to obtain a nonphysiological RT chimera, containing the four mu-SCHEME 1. Structure of the 4-arylmethylpyridinone derivative.
tations (D67N, K70R, T215Y, and K219Q) only in the p66 subunit, whereas the p51 subunit contained the wild type sequence. This hybrid RT showed a rate of pyrophosphorolysis similar to the wild type and AZT-resistant enzymes (k app ϭ 0.065 Ϯ 0.004 min Ϫ1 ). Unlike wild type enzyme, the chimeric RT efficiently catalyzed ATP-dependent phosphorolysis (k app ϭ 0.020 Ϯ 0.005 min Ϫ1 ). These data clearly show that only mutations in the p66 subunit were responsible for the increased ability of the mutant enzyme to unblock the AZTMP-terminated primer by an ATP-dependent mechanism.
Inhibition of Phosphorolysis by NNRTIs-Non-nucleoside compounds are well known inhibitors of HIV-1 RT DNA polymerase. We tested whether these compounds could also inhibit the phosphorolysis catalyzed by wt and AZT-resistant RTs, using a primer blocked at its 3Ј-end by AZTMP at saturating RT concentrations. As shown in Fig. 2 and Table I, nevirapine, 9-Cl-TIBO, and 4-arylmethylpyridinone inhibit the phosphorolytic activity, although with a different efficiency. 9-Cl-TIBO was the weakest inhibitor, whereas the 4-arylmethylpyridinone was the most efficient. The latter compound was about 1000-fold more potent to block pyrophosphorolysis than nevirapine and inhibited both ATP-and PP i -dependent phosphorolysis in the nanomolar range.
We analyzed whether the inhibition by 4-arylmethylpyridinone was due to the impaired binding of the template-primer or ATP upon binding of the NNRTI. In previous work, we found that 4-arylmethylpyridinone was a linear noncompetitive inhibitor against dGTP in polymerization reactions, with a K i of 35 nM (11). Steady-state kinetics analysis of inhibition by 4-arylmethylpyridinone against template-primer was examined by varying the concentration of poly(rA)-dT 20 while maintaining [␣-33 P]dTTP at a fixed concentration of 10 M. 4-Arylmethylpyridinone displayed a complex inhibition pattern with respect to template-primer. At low concentrations of templateprimer (Ͻ10 nM), the pattern was essentially uncompetitive, indicating that this compound preferentially binds to the RT/ template-primer complex. Above this value the pattern changed to mixed noncompetitive (data not shown).
To confirm that this compound preferentially binds to the DNA-RT complex, 2 nM poly(rA)-dT 20 was incubated for 2 min with 10 M [␣-33 P]dTTP, variable concentrations of 4-arylmethylpyridinone and 2, 5, 10, or 25 nM of AZT r -RT. Under these conditions, IC 50 values for 4-arylmethylpyridinone were 72, 59, 47, and 24 nM, respectively. In the same experiment, IC 50 values for nevirapine were 0.87, 0.85, 2.2, and 2.8 M. As a control, inhibition by 4-arylmethylpyridinone was measured under the same conditions except than a fixed concentration of AZT r -RT was used (50 nM), and increasing concentrations of poly(rA)-dT 20 were added (2-25 nM). In this case, the IC 50 values for 4-arylmethylpyridinone increased from 87 to 250 nM. Taken together, these data support the hypothesis that 4-arylmethylpyridinone preferentially binds to the RT-templateprimer complex rather than to the free enzyme. In fact, the concentration of this compound needed to half-inhibit the ATPdependent phosphorolysis reaction using 2 nM r39-d21-AZTMP, 3 mM ATP, and 5-50 nM of RT is about 1-2 nM, irrespective of the total enzyme concentration added. Phosphorolysis is revealed by the disappearance of the AZTMP-terminated primer and the appearance of shorter products. B, rates of phosphorolytic removal of AZT by wt RT (E) and AZT r -RT (q). The gels presented in A were quantified by densitometry and experimental data fitted to a burst equation. The solid lines represent the best fits obtained by nonlinear regression of the data to equation A ϫ e Ϫkt ϩ C, where A is the amplitude of the burst, k is the apparent first order rate constant of AZT excision, and C is a constant.

FIG. 2.
Inhibition of phosphorolytic activity of wt RT and AZT r -RT by NNRTIs. 25 nM RT were added to a mixture containing 2 nM ( 32 P-labeled) d21-AZT annealed to the r39 template, the indicated amount of the NNRTI, and 150 M PP i or 3 mM ATP in buffer B and 10 mM MgCl 2 at 37°C. The reactions containing PP i were incubated for 60 min, and those containing ATP were incubated for 120 min. Under these conditions, no ATP-dependent phosphorolysis was detectable for the wt RT. The reactions were quenched at the indicated time, and the products were analyzed by denaturing PAGE. Because RT eliminated several nucleotides from the terminated primer, phosphorolysis was calculated by quantifying all the oligonucleotides with lengths of 21 nucleotides or less by densitometry.
The effect of NNRTIs on ATP binding is difficult to determine. Direct binding assays using labeled ATP are not possible, because millimolar concentrations of the nucleotide are needed, whereas the RT concentration is in the nanomolar to low micromolar range. We tested by an indirect approach whether the concentration of ATP or PP i affects the inhibition caused by 4-arylmethylpyridinone. Useful ATP concentrations were 1-9 mM, because higher concentrations result in substrate inhibition of the phosphorolytic reaction. In the range 0 -9 mM ATP, the IC 50 for 4-arylmethylpyridinone in the RT-catalyzed phosphorolysis reaction only increased about 1.5-fold. The inhibition of DNA polymerization by NNRTI in the absence of AZTTP was also slightly sensitive to the presence of ATP, with the IC 50 values increasing about 1.5-2-fold. These results suggest that NNRTIs do not exert their activity by impacting on ATP binding, although some interference with ATP may be present.
Effect of ATP and PP i on the Rescue of AZTMP-terminated Primer-Because ATP lysis is sensitive to the presence of the next incoming nucleotide (6, 26), we measured the sensitivity of RT to AZTTP in the polymerization reaction. These assays were performed with an adenine-rich template to allow the incorporation of AZTTP and the subsequent excision of the incorporated AZT during reaction. When AZTTP was not added to the reaction medium, RT was able to add bases in the absence of template once the primer was completely polymerized. (Fig. 3A). In agreement with earlier observations (3), in the absence of pyrophosphorolysis, AZTTP was equally potent in the inhibition of DNA synthesis catalyzed by both wt and AZT r -RT enzymes, with IC 50 values of 0.20 and 0.26 M, respectively. Inhibition by AZTTP was lower in the presence of PP i . This effect was similar for both the wt and the mutant RTs (IC 50 values were 0.93 and 1.1 M, respectively). Inhibition by AZTTP was also reduced in the presence of ATP. The reduction was less pronounced for the wt RT than for the AZT r -RT (Fig.  3B). These results were consistent with the observed rates of phosphorolysis for the enzymes with ATP and PP i .
Effect of the NNRTIs on the Sensitivity of Resistant RT to AZTTP in the Presence of ATP or PP i -We have demonstrated that the presence of physiological concentrations of PP i or ATP reduces the sensitivity of RT to AZTTP. This decrease was due to the phosphorolysis-mediated rescue of the AZTMP-terminated chain. On the other hand, we also found that NNRTIs can inhibit the phosphorolysis catalyzed by AZT-resistant RT. Because resistance toward AZTTP is due to the increased unblocking capacity of AZT-terminated primer by mutant RT, it should be expected that inhibition of phosphorolysis by a NNRTI would resensitize the AZT r -RT to AZTTP.
To test this hypothesis we measured the effect of several combinations of AZTTP and different NNRTIs in the presence and absence of PP i or ATP. Fig. 4 (A and B) shows the combination of AZTTP with nevirapine in the presence of PP i . When different concentrations of nevirapine were added, the IC 50 values decreased (Fig. 4C). Under these conditions, the inhibition of AZTTP was similar to that obtained in the absence of PP i and nevirapine (0.20 Ϯ 0.01 M). It should be noted that this reduction was effectively due to a resensitization of the AZT r -RT, because the inhibition values of the polymerase activity obtained with 0.1 and 1 M nevirapine in the presence of PP i were 1 and 15%, respectively. Furthermore, the addition of 0.1 M nevirapine in the absence of PP i had no effect on the sensitivity of AZT r -RT to AZTTP (0.24 Ϯ 0.01 M), whereas 1 M nevirapine only decreased the IC 50 value to 0.16 Ϯ 0.01 M. These results confirm that the enhancement of the effect of AZTTP by nevirapine was not due to the additive inhibition of  DNA synthesis but to the fact that nevirapine resensitizes the mutant enzyme by inhibiting the phosphorolytic reaction. This effect was not exclusive of nevirapine, because similar results were obtained with other NNRTIs, although the concentration needed to obtain this effect varied considerably. For example, in the presence of PP i , 50 nM 4-arylmethylpyridinone decreased the IC 50 of AZTTP for the mutant RT from 2.3 Ϯ 0.2 to 0.20 Ϯ 0.02 M. In the absence of pyrophosphate, the addition of 50 nM 4-arylmethylpyridinone had no effect on the IC 50 of AZTTP (0.2 Ϯ 0.02 M). On the other hand, 9-Cl-TIBO was a weak inhibitor because 20 M was needed to increase 5-fold the effect of AZT. This concentration of 9-Cl-TIBO inhibited the polymerization reaction by 15% in the presence of PP i .
The Synergy of AZTTP and NNRTIs in the Inhibition of the AZT-resistant Enzyme-Because NNRTIs were able to resensitize the AZT-resistant RT to AZTTP, it was expected that their combination with AZTTP would be synergistic in the presence of ATP or PP i . To analyze the interaction of these inhibitors, we measured the inhibition caused by each NNRTI alone or in combination with AZTTP in the presence and absence of either ATP or pyrophosphate. With ATP the combination of AZTTP with a NNRTI is far more effective than without ATP (Table II). This synergy is more obvious in the case of 4-arylmethylpyridinone, because in the presence of ATP, the combination was even more efficient than when each inhibitor was tested alone. The interaction index at 50% inhibition for the combination of AZTTP and nevirapine with ATP was 0.27, showing an exceptionally high degree of synergy (Fig. 5). Con-sequently, the interaction index for 4-arylmethylpyridinone in the presence of ATP was also very low, 0.39 at 50% of inhibition and 0.21 at 90% of inhibition. The same combinations were also synergistic when tested in the presence of pyrophosphate, resulting in interaction indexes of 0.30 at 50% of inhibition for the combination of nevirapine with AZTTP and 0.36 for the combination of 4-arylmethylpyridinone with AZTTP. Only additive effects (no interaction) were found when these combinations were tested in the absence of pyrophosphate or ATP (Fig.  5). These results show that the presence of a substrate of the phosphorolytic reaction is essential to mediate the synergy found in the combinations we tested.
We also analyzed the combination of AZTTP with 4-arylmethylpyridinone using a checkerboard design, i.e. varying AZTTP concentrations over a range of NNRTI concentrations (Fig. 6). In this plot, parallel lines are found if inhibitors do no interact, whereas intersecting lines are found if both inhibitors act synergistically (24). From this plot it can be concluded that in the presence of ATP the interaction is at least 10-fold greater than in its absence, as judged by the abscissa intersection points of both graphs. It should be noted that the IC 50 for AZTTP in the absence and presence of 3 mM ATP in this experiment increased 13-fold, so the high synergy found can be attributed to the inhibition of the phosphorolytical activity of the resistant enzyme, in agreement with the hypothesis. It is noteworthy that the concentrations of 4-arylmethylpyridinone needed to detect this interaction are in fact very low. For example, the concentration of 4-arylmethylpyridinone needed to half-inhibit the polymerase activity of AZT r -RT in the presence of ATP 3 mM was 0.33 M, whereas synergy is easily detected at 0.025 M of 4-arylmethylpyridinone (Fig. 6B). As expected, both methods of evaluating the synergy between RT inhibitors are fully consistent (24). In fact, the combination of 0.2 M of 4-arylmethylpyridinone with 1 M on the AZT r -RT showed a 90% of inhibition (Fig. 6B), and the interaction index calculated for this combination is 0.26. This value is very close to the interaction index of 0.21 reported for the combination of AZTTP and 4-arylmethylpyridinone for the same inhibition level when both inhibitors were mixed at a 1:0.78 ratio (Table  II and Fig. 5).

DISCUSSION
In this work we show that NNRTIs, besides inhibiting the polymerase activity of HIV-1 RT, can also block the phosphorolytic reaction. The fact that three unrelated NNRTIs, nevirapine, TIBO, and 4-arylmethylpyridinone, were able to affect the excision of the incorporated AZT suggests that most NNR-TIs may have the potential to inhibit the phosphorolytic activity to some extent. Our results suggest that NNRTIs do not exert their activity by impacting template-primer or ATP binding, although some interference with ATP may be present. Taking into account that pyrophosphorolysis is the reverse of the polymerization reaction, it seems to us more reasonable that the mode of inhibition of phosphorolysis was similar to that of inhibition of DNA synthesis. The inhibition of PP i -dependent phosphorolysis by NNRTIs also gives new clues in understanding the influence of NNRTI on the active site. Presteady-state kinetic analysis of inhibition of the polymerization reaction shows that NNRTIs block the chemical reaction but do not interfere with dNTP binding or with nucleotide-induced conformational changes (12). In contrast to dNTPs, PP i lacks the sugar ring and the nitrogenated base. Thus, inhibition of the pyrophosphorolytic activity by NNRTIs support the hypothesis that these compounds modify the conformation of the catalytic site, where the crucial Mg 2ϩ ions are not anymore in the proper alignment with the carboxyl groups for efficient catalysis, as previously suggested (12). These changes at the active site of RT would prevent the catalysis in both the forward and the reverse reactions. Moreover, a chimeric enzyme, carrying the AZT-resistant mutations only in the p66 subunit, behaved as the AZT-resistant RT. This fact confirms that only mutations on the p66 are relevant for the resistance of the enzyme, as suggested by previous structural studies (27,28).
We consistently found that the phosphorolysis catalyzed by RT is more sensitive to NNRTI than the polymerization reaction. In addition, synergy is easily detected at concentrations of the NNRTI that barely inhibit the polymerase activity but effectively block the phosphorolysis catalyzed by the AZT r -RT (Figs. 4 and 6). The fact that phosphorolysis catalyzed by RT is more sensitive to NNRTI than DNA polymerization, is probably related to the absolute rate of both reactions. Depending on the conditions used, DNA polymerase activity catalyzed by RT may be about 5,000 -20,000-fold faster than ATP-dependent phosphorolysis. For example, typical k cat values for RT catalyzed DNA polymerization are about 3 s Ϫ1 , whereas apparent first order constant for ATP-dependent phosphorolysis for the mutant enzyme can reach about 0.012 min Ϫ1 ϭ 0.0002 s Ϫ1 .
Pyrophosphorolysis of the AZTMP-terminated primer is a more efficient process than ATP lysis, because higher concentrations are needed to catalyze the reaction in the presence of the nucleotide. However, wild type RT is almost unable to catalyze ATP-dependent phosphorolysis, whereas the enzyme carrying the four mutations shows an increase in this activity by at least 10-fold. The difference found between wt and resistant RTs points at ATP as the possible substrate responsible for the rescue of the terminated primer in vivo (5). Many details of   Table II were  the role of pyrophosphate in the AZT resistance mechanism remains to be clarified, but in our hands physiological concentrations of pyrophosphate can severely decrease the sensitivity of the wild type RT to AZTTP. Arion et al. (4) observed that the presence of 150 M pyrophosphate decreased by 3-fold the sensitivity of the AZT-resistant RT for the chain terminator. However, no significant changes were found with the wild type RT. In our study, the presence of PP i at physiological concentrations not only decreased the inhibition of the RT-catalyzed DNA synthesis by AZTTP, but this reduction was similar with both enzymes. These results are consistent with the phosphorolytical activity of wild type and AZT r RTs.
This work was intended to analyze the biochemical mechanism mediating the synergy between NNRTIs and AZT on resistant strains. Despite many studies devoted to this topic, the molecular mechanism underlying the antiviral synergy of combinations of reverse transcriptase inhibitors is in most cases unknown. The synergistic inhibition of HIV replication in cell culture has been reported for many combinations of nucleosidic and NNRTI inhibitors including, among others, bis(heteroaryl)piperazine derivates (29), pyridinone derivates (30), nevirapine (13), HEPT derivates (14), TIBO derivates (15,31), or canalolide A (16). However, other studies have shown that the same combinations showed no synergy in inhibiting RT activity in vitro (17)(18)(19)(20)(21). In previous studies, the relevance of the phosphorolysis on the interaction between inhibitors has not been considered. We found that these combinations were highly synergistic only when tested in the presence of PP i or ATP. Figs. 4 and 6 clearly shows that this high level of synergy is due to the fact that complete inhibition of the phosphorolytic activity of AZT r -RT is attained at concentrations of NNRTI that barely affect its polymerase activity. According to our results, the synergy found is more related to the inhibition of the phosphorolytical activity by the NNRTI than to the combined effect on the polymerase activity. Consequently it is clear that, along with its ability to block the polymerase reaction, the inhibition of the phosphorolytically mediated AZT excision activity by NNTIs should also be taken into account.
Because AZT has become the most accessible anti-HIV agent, a better understanding of the role played by other inhibitors in the resistance to AZT should help to design more efficient therapies against HIV. Nowadays, anti-HIV chemotherapy is based in the use of highly active antiretroviral treatments, where three or more drugs are combined to inhibit virus replication. Because resistance to AZTTP is due to an increased rate in the phosphorolytic reaction by AZT r -RT, the use of a NNRTI blocking efficiently this activity would resensitize resistant viruses to AZT. Borkow et al. (8) have suggested that this inhibition might also be a relevant mechanism for the delayed development of resistance to combinations of AZT plus UC781 in antiviral assays. This proposal deserves further attention and can be tested by analyzing the long term antiviral effect of combinations of AZT and a NNRTI able to effectively inhibit the phosphorolytic activity of HIV-RT. In this context, it would be interesting to know whether the reported observations that combinations of two nucleoside inhibitors (including AZT) with efavirenz are clearly superior to combinations including only nucleoside analogs, could be related to the inhibition of the phosphorolytical activity by this NNRTI.
Our results indicate that 4-arylmethylpyridinone is a very good candidate to be used with AZT in the multitherapy strategy. Although nevirapine is slightly more synergistic than 4-arylmethylpyridinone when combined with AZTTP, 4-arylmethylpyridinone is far more effective in inhibiting DNA polymerase (K i ϭ 20 nM) (11), and it is about 1,000-fold more effective in inhibiting ATP and PP i -dependent phosphorolysis (Table I).
In addition, 4-arylmethylpyridinone can also block the DNA polymerase activity of RT carrying K103E or Y181C mutations, in agreement with the demonstrated ability of this compound to inhibit a nevirapine-resistant virus with an IC 50 of 40 nM (11). Taken together, our data support the importance of screening the ability of the currently used NNRTIs as inhibitors of RT-catalyzed phosphorolysis, leading to the design of new compounds based on their capacity to inhibit the phosphorolytic-mediated AZT excision.