A loss of viral replicative capacity correlates with altered DNA polymerization kinetics by the human immunodeficiency virus reverse transcriptase bearing the K65R and L74V dideoxynucleoside resistance substitutions.

Mechanisms governing viral replicative capacity are poorly understood at the biochemical level. Human immunodeficiency virus, type 1 reverse transcriptase (HIV-1 RT) K65R or L74V substitutions confer viral resistance to 2',3'-dideoxyinosine (ddI) in vivo. The two substitutions never occur together, and L74V is frequently found in patients receiving ddI, while K65R is not. Here we show that recombinant viruses carrying K65R and K65R/L74V display the same resistance level to ddI (about 9.5-fold) relative to wild type. Consistent with this result, purified HIV-1 RT carrying K65R RT or K65R/L74V substitutions exhibits an 8-fold resistance to ddATP as judged by pre-steady state kinetics of incorporation of a single nucleotide into DNA. Resistance is due to a selective decrease of the catalytic rate constant k(pol): 22-fold (from 7.2 to 0.33 s(-1)) for K65R RT and 84-fold (from 7.2 to 0.086 s(-1)) for K65R/L74V RT. However, the K65R/L74V virus replication capacity is severely impaired relative to that of wild-type virus. This loss of viral fitness is correlated to a poor ability of K65R/L74V RT to use natural nucleotides relative to wild-type RT: 15% that of wild-type RT for dATP, 36% for dGTP, 50% for dTTP, and 25% for dCTP. The order of incorporation efficiency is wild-type RT > L74V RT > K65R RT > K65R/L74V RT. Processivity of DNA synthesis remains unaffected. These results explain why the two mutations do not combine in the clinic and might give a mechanism for a decreased viral fitness at the molecular level.

The efficacy of current anti-HIV 1 treatments with nucleoside analogues (nucleoside reverse transcriptase inhibitors) is lim-ited by the emergence of drug-resistant variants. Because antiretroviral therapy does not completely suppress viral replication, a resulting selection pressure favors the emergence of mutant viruses that partially or completely lose sensitivity to one or several inhibitors (1,2). Such an example is given by the progressive appearance of primary mutations in the viral pol gene encoding the reverse transcriptase (RT) after prolonged treatments that make use of dideoxynucleosides. K65R and L74V substitutions in RT are typically associated with a moderate resistance to ddI and ddC, and these substitutions are directly responsible for a loss of efficacy of these drugs in patients carrying the corresponding mutant viruses.
We previously reported the molecular role of the K65R substitution in the multiple resistance to dideoxynucleotides (3). Lysine at position 65 is located in the flexible ␤3-␤4 loop located in the finger domain of the p66 monomer of RT. The side chain of the Arg-65 interacts with the ␥-phosphate of the incoming nucleotide and alters the catalytic rate of creation of the phosphodiester bond for dideoxynucleotides specifically (3). The K65R mutation is found infrequently (less than 2%) on viral isolates from patients treated with ddI and ddC and gives a moderate 2-10-fold decrease in the sensitivity to these two drugs (4,5). K65R also has been described to give in vitro RT-mediated resistance to tenofovir (6). We recently have shown that the molecular effect of K65R mutation on tenofovir is identical to the one previously described for ddNTPs (7). L74V, unlike K65R, is the most frequently observed mutation in patients receiving ddI (8). It has been found in up to 48% of the patients under ddI monotherapy (9). The level of resistance to ddI given by L74V is similar to that given by K65R (8, 10 -14). However, little is known about the precise role of the amino acid at position 74, which does not make a direct contact with the incoming nucleotide. Leu-74 makes a van der Waals contact with the ϩ1 DNA template base (15) and also interacts with the side chain of Gln-151 and Arg-72 whose contribution in the catalytic reaction is better understood (16 -18).
Many questions remain concerning the emergence, selection, preference, and persistence of one drug resistance mutation relative to another. As a matter of fact, these two substitutions, K65R and L74V, are not found associated together in the clinic (9), suggesting that their simultaneous occurrence would not be advantageous for the virus. Knowledge of the molecular role of these amino acids would yield useful information. For example, these substitutions might confer unexpected ddI sensitivity to HIV when present together in RT. Alternatively they might confer a "ddN-super-resistant" phenotype to HIV-1, suggesting their simultaneous occurrence could potentially be selected under a higher selective pressure. As a third possibility, they might not be tolerated together at the RT active site, yielding a non-functional or impaired RT. In any case, knowledge of the role of these amino acids in the RT active site might have some relevance for future drug design. Although the interplay of drug resistance and viral fitness is increasingly apparent (19), the molecular mechanisms describing how an impaired RT influences viral fitness are still unclear at present.
To address these questions, we studied the properties of K65R/L74V double mutant both at viral and enzymatic levels. The mechanism of drug resistance by K65R/L74V was investigated using recombinant viruses and elucidated at the molecular level using purified RT and pre-steady state kinetics. We find that K65R and L74V do not combine into a ddN-superresistant virus and that the simultaneous presence of both substitutions in RT affects viral replication capacity through alteration of the interaction of natural nucleotides with the RT active site.

Recombinant HIV-1 Molecular Clone Constructions-The 1258-bp
ClaI-Eco47III fragment of the p66RTB vector containing mutations K65R, L74V, or K65R/L74V was introduced in the same restriction sites as described previously into the p66RTB-AD8 vector (20). The resulting plasmid was then digested with MscI restriction enzyme, and the 1.9-kb MscI fragment was subsequently cloned back in the same restriction sites in the pNL4.3 HIV-1 proviral molecular clone to obtain the pNL4.3 RTB plasmids mutated in the RT gene.
Cell Culture, Transfections, and Infections-293T cells were maintained in Dulbecco's modified Eagle's medium supplemented with antibiotics and 10% fetal bovine serum. C8166 and H9 T cell lines were cultivated in RPMI 1640 medium supplemented with antibiotics and 10% fetal calf serum. To obtain wild-type (WT) and K65R, L74V, or K65R/L74V viral stocks, 293T cells were transiently transfected by the use of FuGENE 6 transfectant reagent (Roche Applied Science) as recommended by the manufacturer. Two days post-transfection, cellfree supernatants containing viruses were titrated by assaying the fusogenic properties on C8166 cells. C8166 cells (1 ϫ 10 5 cells/well in triplicates) were seeded on 96-well microtiter culture plates (Costar) and infected with serial 5-fold dilutions of each viral stock. Cells were examined for cytopathic effects on day 5 of culture, and the 50% tissue culture infective dose (TCID 50 ) of viral stocks was determined. To analyze the replicative properties of each viral stock, H9 T cells (1 ϫ 10 6 ) were infected with virus calibrated to infect at a multiplicity of infection (m.o.i.) of 0.1, 0.01, and 0.001. Two-milliliter cultures were done in duplicate using 6-well tissue culture plates (Costar). Supernatants (0.5 ml) from each culture were collected twice a week, and cultures were supplemented with fresh medium. The p24 antigen was quantitated in cell-free culture supernatants and used as a marker of replication kinetics.
IC 50 Measurement-The susceptibility of RTs from recombinant viruses to ddI was determined by measuring IC 50 values using reverse transcriptase activity assay. C8166 cells (1 ϫ 10 6 cells/well in triplicates) seeded on 24-well tissue culture plates (Costar) were treated with increasing concentrations (2, 5, 10, 20, 40, or 60 M) of ddI and immediately infected during 24 h with each viral stock calibrated for a m.o.i. of 0.005. Cells were then washed twice in serum-free RPMI 1640 medium before being grown in RPMI 1640 medium supplemented with antibiotics and 10% fetal calf serum in the presence of increasing amounts of ddI. Day two postinfection, the cell-free supernatant was removed and monitored for the reverse transcriptase (21). Inhibition (IC 50 ) was expressed as the concentration of inhibitor producing a 50% inhibition of virus growth.
Determination of Viral DNA Copy Number by Real Time PCR-Viral stocks obtained from transfected 293T cells were filtered through 0.45-m-pore-size filters and treated with 20 g/ml DNase I (Promega) for 30 min at 37°C in the presence of 10 mM MgCl 2 . H9 cells (1 ϫ 10 6 ) were infected for 2 h by spinoculation (22) with equal amounts of each of the RT mutant viruses, and then cells were washed extensively with phosphate-buffered saline and cultured in fresh medium for 15 additional h. The total DNA was isolated from the infected cells by the QIAamp DNA minikit protocol (Qiagen) and eluted with 200 l of elution buffer. DNA samples were quantitated by optical density at 260 nm. Viral DNA products corresponding to the different steps of reverse transcription were assessed by quantitative real time PCR as described in the literature (23). Real time PCR was performed with an ABI PRISM 7000 apparatus (Applied Biosystems) using PCR primers and Taqman probes described previously (23). The mixture contained 15 l of Universal PCR Master Mix (Applied Biosystems), 500 ng of nucleic acids (10 l), 0.9 l of primers, and 0.3 l of probe. The PCR primers and Taqman probes were present at a final concentration of 200 nM. Unknown samples were compared with duplicate samples containing 10 7 -10 3 copies of plasmid DNA. The cycling conditions were 2 min at 50°C and 10 min at 95°C and then 40 cycles of 95°C for 15 s and 60°C for 1 min. The Taqman primer sets used to monitor initiation were as follows: Forward primer, 5Ј-AAC TAG GGA ACC CAC TGC TTA AG-3Ј; Reverse primer, 5Ј-CTG CTAG AGA TTT TCC ACA CTG AC-3Ј; Taqman probe, 5Ј-R-ACA CTA CTT GAA GCA CTC AAG GCA AGC TTT-Q-3Ј, and the Taqman primer sets used to monitor elongation were as follows: Forward primer, 5Ј-TGT GTG CCC GTC TGT TGT GT-3Ј; Reverse primer, 5Ј-GAG TCC TGC GTC GAG AGA TC-3Ј; Taqman probe, 5Ј-R-CAG TGG CGC CCG AAC AGG GA-Q-3Ј where R is the reporter dye 6-carboxyfluorescin and Q is the quencher 6-carboxytetramethylrhodamine.
HIV-RT Plasmid Constructions, Enzyme Preparations, and Reagents-The wild-type RT gene construct p66RTB was used to obtain K65R, L74V, and K65R/L74V RT as described previously (24). All constructs were verified by DNA sequencing. The recombinant RTs were co-expressed with HIV-1 protease in Escherichia coli to obtain p66/p51 heterodimers, which were later purified using affinity chromatography. All enzymes were quantitated by active site titration before biochemical studies. DNA oligonucleotides were obtained from Invitrogen. Oligonucleotides were 5Ј-32 P-labeled using T4 polynucleotide kinase (New England Biolabs). ␥-32 P-Labeled adenosine 5Ј-triphosphate was purchased from Amersham Biosciences.
Pre-steady State Kinetics of Single Nucleotide Incorporation into DNA-Pre-steady state kinetics were performed using dATP, dTTP, dCTP, dGTP, and ddATP in conjunction with wild-type, K65R, L74V, and K65R/L74V RT. Rapid quench experiments were performed with a Kintek Instruments Model RQF-3 using reaction times ranging from 10 ms to 30 s. All indicated concentrations are final. The primer DNA/DNA oligonucleotides used for the rapid reaction were a 5Ј-labeled 21-mer primer (5Ј-ATA CTT TAA CCA TAT GTA TCC-3Ј) annealed to a 31-mer template, 31T-RT (5Ј-TTT TTT TTT AGG ATA CAT ATG GTT AAA GTA T-3Ј), for the incorporation of dTTP, 31A-RT (5Ј-AAA AAA AAA TGG ATA CAT ATG GTT AAA GTA T-3Ј) for the incorporation of dATP/ddATP, 31C-RT primer (5Ј-TTT TTT TTT GGG ATA CAT ATG GTT AAA GTA T-3Ј) for the incorporation of dCTP, and 31G-RT primer (5Ј-TTT TTT TTT CGG ATA CAT ATG GTT AAA GTA T-3Ј) for the incorporation of dGTP. For natural nucleotides, the reaction was performed by mixing a solution containing 50 nM (active sites) HIV-1 RT bound to 100 nM primer/template in RT buffer (50 mM Tris-HCl, pH 8.0, 50 mM KCl, 0.05% Triton X-100), and a variable concentration of dNTP in 6 mM MgCl 2 . Reactions involving ddATP were conducted with excess concentrations of enzyme (200 nM) over primer/template duplex (100 nM). As described before (18), these conditions were chosen to eliminate the influence of the enzyme turnover rate (k ss ) that interferes in the measurements of low incorporation rates. The products of reactions were analyzed using sequencing gel electrophoresis (14% acrylamide, 8 M urea in Tris borate-EDTA buffer) and quantified using photostimulated plates and a FujiImager. The formation of product (P) over time was fitted with a burst equation, where A is the amplitude of the burst, k app is the apparent kinetic constant of formation of the phosphodiester bond, and k ss is the enzyme turnover rate, i.e. the kinetic constant of the steady state linear phase. The dependence of k app on dNTP concentration is described by the hyperbolic equation, where K d and k pol are the equilibrium constant and the catalytic rate constant of the dNTP for RT, respectively. K d and k pol were determined from curve fitting using Kaleidagraph (Synergy Software).
Assays of RT DNA Polymerization Rate and Processivity-The rate of polymerization was measured using a 5Ј-labeled oligo(dT) 21 primer annealed to poly(rA) or a 5Ј-labeled oligo(dA) 21 primer annealed to poly(rU). Extension products were analyzed using a gel assay. The primer/template (25 nM) was incubated with RT (50 nM) in RT buffer at 37°C. The reaction was initiated by the addition of 500 M dNTP (dTTP or dATP) in 6 mM MgCl 2 and quenched at various times by 0.3 M EDTA. The processivity assay was performed under the same experimental conditions with poly(rA)/oligo(dT) for 2 min with 2 g/l heparin and an excess of unlabeled primer/template added to the dTTP/MgCl 2 mixture. The control reaction was assayed by adding heparin together with the enzyme before starting the reaction.
Molecular Modeling of dATP and ddATP in the RT Active Site-Molecular modeling was performed using GenMol (25,26) with its all-atoms force field module from the x-ray crystal structure of wildtype RT⅐dNTP⅐DNA ternary complex (15). Modeling was performed with a 25-Å-radius sphere centered on the catalytic DNA polymerase domain. This model of HIV-RT included the residues of both p66 and p51 domain, DNA template/primer, two magnesium ions, and dTTP. Close attention was given to both magnesium ions as they have a key role in positioning the incoming dNTP or nucleoside reverse transcriptase inhibitor and in catalysis per se.
To prepare the closed RT⅐DNA⅐dTTP complex (15), the authors used a ddGMP primer terminus positioned at the so-called "Priming site" (P site (27)). We replaced ddGMP by dGMP, and interactions of 3Ј-OH with Mg 2ϩ were optimized. In the crystal structure, both magnesium ions were complexed with residues Asp-185, Asp-110, and Val-111. The complexation process of Mg 2ϩ with the incoming nucleotide via its triphosphate chain and residues Asp-110, Val-111, and Asp-185 was optimized using a conformational analysis and found to be the same as in the crystal structure of the ternary complex made of RT⅐DNA⅐dTTP. The complexation of the second Mg 2ϩ was also optimized. Positions of Asp-186 and Asp-110 side chains were adjusted and optimized. This second Mg 2ϩ was complexed with Asp-186 (O␦-1 carboxylate side chain), Asp-185 (O␦-2 carboxylate side chain), and the 3Ј-OH dGMP primer terminus.
All minimizations were conducted using Linux RedHat 7.3 workstations and a 2.0 GHz bi-Xeon Pentium® 4 processor. Hydrogen atoms were added to the enzyme, DNA duplex, and dTTP within the Biopolymer module of GenMol. Hydrogen positions were then optimized. The construction of the nucleotide triphosphate was based on x-ray conformation of dTTP present in the ternary complex RT⅐DNA⅐dTTP. All atom charges were computed with GenMol for RT-DNA, both Mg 2ϩ , dATP, and ddATP (25,26). The substituted RT structures were obtained by changing selected amino acids (K65R and L74V). The adenine at position 5 of the template was mutated to thymine, the counterpart of the incoming nucleotide analogue dATP and ddATP. The conformation of this fifth template nucleotide and the side chain of selected amino acids 65 and 74 were minimized within the Biopolymer module of GenMol using a conformational analysis method. Finally the incoming nucleotide was docked into the DNA-polymerase active site from the positioning of dTTP present in the x-ray structure ternary complex (nucleoside binding site, N site). The resulting ternary complexes were optimized using the GenMol all-atoms force field module. Fig. 2 was done and rendered using Swiss-PdbViewer and Povray, 2 respectively.

RESULTS
K65R or L74V substitutions in RT are known to be associated with ddN resistance in nucleoside reverse transcriptase inhibitor-treated patients (4,5,8). However, viruses carrying the K65R/L74V double substitution are almost never reported in the clinic. To understand this apparent mutual exclusion, we constructed and studied a K65R/L74V recombinant virus. The corresponding K65R/L74V RT was expressed independently, purified, and biochemically characterized. Viral replication capacity and drug resistance were measured at the virus and enzyme level comparatively.
Resistance of Wild-type, K65R, L74V, and K65R/L74V Recombinant Viruses to ddI-Resistance to ddI by wild-type, K65R, L74V, and K65R/L74V viruses was investigated by examining the replication of viruses in ddI-treated cells. C8166 cells were infected, in the presence of increasing amounts of ddI, with each viral stock calibrated to the same m.o.i. Two days postinfection, the reverse transcriptase activity in the cell-free supernatant was measured, and the resistance was expressed as the concentration of inhibitor inducing a 50% inhibition of virus growth. As shown in Table I, mutant viruses exhibited a significantly higher resistance to ddI compared with wild-type viruses with a 5-fold increase for L74V and a 9.5-fold increase for K65R and K65R/L74V virus. The sole K65R mutation conferred the resistance to ddI at levels similar to that of the K65R/L74V mutation (IC 50 ϭ 38 M), indicating that the additional L74V mutation did not increase ddI resistance. The L74V mutation alone displayed a lower IC 50 (20 M) than that of the double K65R/L74V mutations. It is tempting to speculate that resistance of the double mutant is due to the K65R mutation and is not increased further with the addition of the L74V mutation.
Resistance of K65R, L74V, and K65R/L47V RT to ddATP-Using pre-steady state kinetics with a single nucleotide incorporation, we tested the role of K65R, L74V, and K65R/L74V substitutions in RT-mediated ddI resistance. ddATP, the active metabolite of ddI, was used in conjunction with purified RT. The nucleotide affinity K d is calculated as the nucleotide concentration that gives half of the maximum incorporation rate k pol . A nucleotide analogue 5Ј-triphosphate is characterized by its efficiency of incorporation (k pol /K d ) into DNA as compared with that of its natural counterpart. For a given RT, comparing k pol /K d values between a natural dNTP and its corresponding analogue is a convenient manner to evaluate selectivity for (or discrimination of) the analogue. Comparing selectivity between RTs defines in vitro resistance (or susceptibility) at the enzymatic level.
Single nucleotide incorporation constants are presented in Table II. Wild-type RT discriminated ddATP relative to dATP. The difference of incorporation efficiency between the two substrates comes exclusively from an altered incorporation rate: 50 s Ϫ1 for dATP and 7.2 s Ϫ1 for ddATP, leading to a 7.4-fold discrimination. The affinity, however, remained the same for dATP and ddATP (K d ϭ 7.5-8 M). As described before, K65R RT exhibits a "k pol -dependent resistance" to ddATP (3). Note that the differences in kinetic constants for ddATP (compared with Selmi et al. (3,21)) originate in the variation of RT/DNA ratio used in this study as described under "Experimental Procedures." The efficiency of incorporation (k pol /K d ) of dATP by K65R RT was 1.7 s Ϫ1 ⅐M Ϫ1 , which was 60-fold better than in the case of ddATP (0.029 s Ϫ1 ⅐M Ϫ1 ). This high discrimination of the dideoxynucleotide over its natural counterpart was generated by a low incorporation rate of ddATP by K65R RT: k pol ϭ 0.33 s Ϫ1 , which was 22-fold lower than in the case of WT RT (illustrated in Fig. 1). It yielded an overall 8-fold resistance to ddATP by K65R RT. In the case of L74V RT, it is interesting to notice that the dATP incorporation properties resembled those of WT RT rather than K65R RT. It is especially true for the incorporation rate: k pol(dATP) ϭ 41 s Ϫ1 , resulting in a 2.9 s Ϫ1 ⅐M Ϫ1 incorporation efficiency. The incorporation rate of ddATP by L74V RT was also impaired (k pol ϭ 1.4 s Ϫ1 ) but to a 2 See www.povray.org. a The concentration of drug that produced a 50% inhibition of virus growth estimated by the reverse transcriptase activity in the cell-free supernatant was determined as described under "Experimental Procedures." b Values indicate drug resistance of mutants relative to wild-type HIV-1.
lesser extent than in the case of K65R RT ( Fig. 1 and Table I). Hence ddATP resistance by L74V RT was only 4.9-fold. In conclusion, we observed using pre-steady kinetics that the K65R substitution gave a 2-fold higher resistance to ddATP than that for L74V as it was first reported with recombinant viruses (Table I). When both substitutions combine on the enzyme, the single nucleotide incorporation experiments using K65R/L74V RT revealed minor changes in affinity for dATP and ddATP between K65R/L74V RT and K65R RT. However, the incorporation rate of ddATP using K65R/L74V RT was only 0.086 s Ϫ1 (4-fold lower than that of K65R RT). This promoted a 59-fold discrimination of ddATP and brought resistance up to 8-fold. Thus, there was no clear additive or synergistic effect of the two mutations for ddATP resistance. K65R/L74V RT had an overall resistance profile comparable to that of K65R RT, and not L74V RT, as judged by the 8-fold resistance to ddATP. This is consistent with viral resistance phenotypes observed in vitro (Table I). Our data show that a lysine to arginine substitution had a dominant and stronger effect than the L74V substitution on both dATP and ddATP incorporation.
Molecular Modeling of dATP and ddATP in the Nucleotide Binding Site-An attempt was made to understand these results in a structural context. Both dATP and ddATP were modeled in the active site of RT containing one or two mutations (Fig. 2). The presence of the guanidinium function of Arg-65 ( Fig. 2A) induces a strong electrostatic interaction with the oxygens bonded to ␤and ␥-phosphates of the incoming nucleotides. In the case of dATP, the constraints on these phosphates are partially compensated by the presence of the 3Ј-OH group, which interacts with a non-bridging oxygen of the ␤-phosphate (2.6 Å) through a strong intramolecular hydrogen bond and with the main chain amido group of Tyr-115 (3.2 Å). The absence of these contacts in the case of ddATP allows Arg-65 to destabilize the positioning and alignment of the triphosphate chain (11°angle rotation), giving a possible explanation for the effects observed on catalysis. In the case of the mutation of amino acid 74 to a valine (L74V, Fig. 2B), the stabilizing interaction between the nucleotide base of the incoming nucleotide and the side chain of Leu-74 ( Fig. 2A, blue  arrow) was lost (Fig. 2B, red arrow). This may induce a rotation of the base (7°for ddATP compared with dATP), which indi-  . (3). b K d and k pol were determined as described under "Experimental Procedures." Standard deviations were Ͻ20%. c Selectivity and resistance were determined as described under "Results." The selectivity is the ratio of (k pol /K d(nucleotide analogue) )/(k pol / K d(nucleotide) ). rectly affects the positioning of the phosphates. Consistent with this observation, we note in Table II a 2-fold increased K d that might be a reflection of this lost stabilizing interaction. Again in the case of dATP, rotation constraints are partially compensated by the presence of the 3Ј-OH group that stabilizes the nucleotide for catalysis and may explain why dATP is incorporated more efficiently than ddATP. When the two mutations K65R and L74V are introduced together, the effect of each single mutation might be cumulative, resulting in an increased misalignment of the phosphates of ddATP compared with dATP (13°angle rotation). This might translate into the overall decrease of incorporation efficiency of ddATP by K65R/L74V RT reported in Table II.
Replication Capacity of K65R, L74V, and K65R/L74V Viruses-We compared the replicative capacity of wild-type, K65R, L74V, and K65R/L74V viruses to determine whether a low replication capacity of K65R/L74V could explain why viruses carrying the double mutation are not stably selected in vitro or observed in the clinic. Replicative capacity can be assayed by a variety of methods of which the viral growth kinetic assay is the most widely used (19). Viral growth kinetics was followed by measuring the p24 antigen levels in the cell-free supernatant of H9 T cells infected with a standardized viral inoculum. To enhance kinetic differences and sensitivity, viral growth was monitored for three distinct multiplicities of infection (Fig.  3). Results showed that L74V and K65R viruses replicate with kinetics similar to wild-type viruses irrespective of the m.o.i. used to infect cells, although they both showed a 3-day delay at m.o.i. ϭ 0.01. The most striking difference was observed with the K65R/L74V virus for which replication was severely impaired for each of the multiplicities of infection used. Its replication was barely detectable when cells were infected with a m.o.i. of 0.001. The viral peak was found to occur slightly later than with wild-type, K65R, or L74V. These data demonstrate that the K65R/L74V double mutation confers a significant replicative disadvantage to the virus. The loss of apparent fitness of the virus carrying the K65R/L74V mutation is likely to be correlated with the inability to recover such doubly mutated viruses in infected patients.
We next asked whether the loss of replication of K65R/L74V double mutant upon multiple cycles of replication could also be observed during a unique cycle of replication. H9 cells were infected for 15 h with similar amounts of viral stocks (monitored by amounts of capsid p24 antigen). DNA copies of neosynthesized viral DNA were quantified by real time PCR as described under "Experimental Procedures." Fig. 4 presents quantification of initiated (R-U5) and elongated (R-Gag) viral cDNA. Viral DNA copy number estimated using either R-U5 or R-Gag primers was similar for wild-type, K65R, and L74V viruses. As expected, an ϳ10-fold decrease in initiated and elongated viral cDNA was observed for K65R/L74V viruses, indicating that these viruses were unable to ensure early steps of reverse transcription. The order of replication capacity was: HIV-1 WT ϭ HIV-1 L74V Ͼ HIV-1 K65R Ͼ HIV-1 K65R/L74V . In conclusion, both long term virus growth kinetics and the single round of infection showed that K65R/L74V virus displays a severe replication capacity defect when compared with the wild-type virus.
Processivity and Polymerization Rate of HIV-1 RT Mutants-In some nucleoside-resistant viruses such as L74V, M184V, K65R, and K65R/M184V, the processivity of the RT has been incriminated as a cause of fitness loss (28 -30). Processivity of DNA synthesis was examined for WT and K65R/ L74V RT. Fig. 5A shows a polymerization assay conducted with heparin, which traps the enzyme as soon as it dissociates from its template. Under non-limiting nucleotide concentrations, there was no difference in processivity between WT and K65R/ L74V RT. Hence, under the definition of processivity of DNA synthesis, the substitutions do not force the enzyme to come off its template, and the RT processivity cannot be considered as one of the parameters involved in the replication capacity defect of the K65R/L74V virus. An apparent decrease in processivity may or may not appear under limiting nucleotide concentrations, but in fact it would be diagnostic of an altered nucleotide traffic at the RT active site. This point is addressed in the following sections.
Indeed the substitutions mentioned above (L74V, M184V, K65R, and K65R/M184V) are located at the RT active site, and thus it is likely that they affect directly the nucleotide traffic there. We noticed that most of these substitutions affect the catalytic rate of incorporation of the nucleotide into  (Table II for dATP and see Refs. 3 and 18). Therefore, the apparent rate of DNA synthesis by K65R/L74V RT was visualized using a primer extension and gel assay. Fig. 5B shows the apparent velocity of the RT in its 1st s of multiple incorporation of a pyrimidine (dTMP) and a purine nucleotide (dAMP) into a homopolymeric product. Using a poly(rA)/oligo(dT) template, WT RT displayed an apparent rate of polymerization of roughly 15 nucleotides/s (k app ϭ 15 s Ϫ1 ). K65R/L74V RT was half as fast with a k app around 8 s Ϫ1 . Using a poly(rU)/oligo(dA) template, WT RT was also more efficient than K65R/L74V RT. We conclude that K65R/L74V RT has a slower apparent polymerization rate than WT RT. These findings prompted us to study in closer detail the kinetic and catalytic constants of natural nucleotides toward RT bearing K65R, L74V, and K65R/L74V substitutions.
Single Incorporation of Natural Nucleotides by WT, K65R, L74V, and K65R/L74V RT-We investigated the connection between the RT-mediated replicative defect and the decreased catalytic step of the reaction by measuring the incorporation kinetics of one single natural nucleotide into DNA. As it was shown for dATP on Table II, incorporation efficiency of a given nucleotide can be determined using pre-steady state kinetics. The incorporation efficiency of dATP by WT RT was 6.7 s Ϫ1 ⅐M Ϫ1 , which was over 2-fold higher than when using L74V RT (Table II). It was 6-fold lower in the case of K65R/L74V RT relative to WT RT, and this 6-fold decrease represents the lowest dATP incorporation efficiency among the four enzymes. Incorporation rates of the three other natural nucleotides were measured with WT, K65R (values reported from Ref. 7), L74V, and K65R/L74V RT. We observed that once again K65R/L74V RT displayed very low incorporation efficiencies compared with other enzymes (Table  III). The incorporation of dGTP by K65R/L74V RT was about 65% impaired compared with WT RT (Fig. 6) as a consequence of the low k pol(dGTP) and the increased K d(dGTP) of the double mutant (Table III). The incorporation efficiency of dCTP was only 0.24 s Ϫ1 ⅐M Ϫ1 for K65R/L74V RT, which was about 4-fold lower than that for WT RT (Fig. 6). This defect was caused by a 4-fold lower incorporation rate (from 1.7 to 7.3 s Ϫ1 for WT RT). The incorporation rate k pol of dTTP by K65R/L74V RT was also impaired (from 0.75 to 0.37 s Ϫ1 for WT RT) without any drastic change in the dissociation constants. Hence it appears that none of the four natural nucleotides is incorporated by K65R/L74V RT as efficiently as by WT RT (15% for dATP, 36% for dGTP, 50% for dTTP, and 25% for dCTP; Fig. 6). When these values were averaged, the global nucleotide incorporation efficiency of K65R/L74V RT was about 3-fold lower than that of WT RT. The order of nucleotide incorporation efficiency was RT WT Ͼ RT L74V Ͼ RT K65R Ͼ RT K65R/L74V . The low replication capacity of K65R/ L74V RT seems to come from a contribution of each of the single mutations. K65R mutation accounted for the low dNTP incorporation rates (41% of WT RT, Table III), whereas L74V substitution slightly affected both the binding affinity and the incorporation rates (Table III). We noticed an overall lower binding affinity of purines to K65R/L74V RT that was also found in K65R and L74V mutants to a lesser extent. Thus, there is a correlation between these enzymatic results and the in vitro virus infection data showing that K65R/L74V virus has a significant disadvantage in its replication capacity over the three other variants.
We conclude that K65R and L74V are not synergistic in terms of resistance, that the inability to select and maintain the K65R/L74V virus is due to moderate resistance to ddNs combined with a loss of viral replication capacity, and that this loss of replication capacity is correlated to a decreased ability of RT to use natural nucleotides. DISCUSSION Although HIV drug resistance is a major problem in current antiretroviral drug therapies, the stable selection of certain drug-resistant HIV mutants may actually be responsible for the prolonged efficacy of treatments. Indeed many drug-resistant variants have a reduced viral fitness compared with the wild-type virus. In many cases, drug resistance occurs at the expense of viral fitness. The negative impact of drug resistance on viral fitness is apparent upon treatment interruption: in the absence of nucleoside reverse transcriptase inhibitors, drugresistant viruses are quickly replaced by wild-type virus (31). This release of the selective pressure is being studied in response to therapeutic failure, and antiretroviral treatments can be interrupted to resensitize the virus to drugs (32)(33)(34). In this context, it is of importance to understand thoroughly the mechanisms underlying drug resistance and viral fitness as well as how they impact each other. K65R and L74V constitute a good example and test case for such a study as they promote resistance to the same drug and they are not found associated together, and the effect of their association in RT is not known either in terms of resistance or viral fitness.
In the present study, we have described the properties of the K65R/L74V substitutions in RT both at the virus and the enzyme levels. We evaluated the individual contribution of K65R and L74V to the resistance to ddI/ddATP and the impact of the resulting drug-resistant viruses on the replication capacity. Our study shows that contribution of the K65R mutation was stronger over that of L74V in terms of resistance levels. Both K65R and K65R/L74V viruses displayed a 10-fold resistance to the dideoxynucleoside in agreement with pre-steady state kinetics showing similar resistance levels (8-fold) between K65R and K65R/L74V RT. We did not see any cumulative effect of the individual mutations on the resistance level. It might not be advantageous to the virus to carry them together as resistance levels are not additive, which by itself is sufficient to explain why the two mutations do not combine in vivo.
Interestingly HIV-1 L74V and HIV-1 K65R variants were very comparable to HIV-1 WT in our assays in terms of replicative capacity both with the long term viral growth and the single cycle of replication method. It has previously been shown that the K65R recombinant virus displays about 50% of the replication capacity of the WT virus (7,29). Other studies have pointed out that L74V mutation is also associated with a reduced replication capacity of the HIV-1 virus isolated from patients (28,35). Although L74V provides less ddI resistance to HIV than K65R does, L74V is nevertheless selected preferentially over K65R in the clinic. Our virus replication assays failed in detecting any large difference between the two single mutants. It is very likely, however, that a more sensitive replication fitness assay (e.g. competition assay) would monitor a better replication capacity of HIV-1 L74V over HIV-1 K65R . Thus, viral fitness would be dominant over the selective pressure induced by ddI treatment. This is what we observed with recombinant enzymes as L74V RT showed a better nucleotide incorporation efficiency than K65R RT. Considering the fitness of the K65R/L74V virus, we observed an attenuated replication capacity compared with the other variants. It seems that the simultaneous presence of the two mutations had a strong negative impact on replication capacity, and this provides a second argument as to why the selection pressure of dideoxynucleosides does not select K65R and L74V together.
What is the molecular basis of this decreased replication capacity? Again K65R, L74V, or K65R/L74V substitutions provide a unique opportunity to understand one cause of replication capacity decrease and, in turn, fitness impairment. Indeed these substitutions are located at the nucleotide binding site in the RT catalytic center, and it is thus very likely that they do not have any other interacting partner than the DNA and/or nucleotide. Thus, interaction of RT with nucleotides is at the origin of the observed decreased replication capacity. Our results indicate that the loss of apparent fitness is due to the decreased ability of drug-resistant RT to use natural nucleotides. Furthermore two aspects of this finding are interesting.
First, these substitutions affected the catalytic rate of incorporation k pol rather than the equilibrium binding constant K d when K65R was present. This is what is observed in the vast majority of drug resistance substitutions involving ddNs where the analogue is not discriminated by a decreased binding affinity (or increased K d ) to RT relative to its natural counterpart but by a decreased incorporation rate k pol (3,18,36). The L74V substitution also promoted a decrease in the incorporation rates except with dATP for which only the affinity was altered. We note that our "enzymatic fitness" results do not highlight a defect in K65R/L74V double mutant as markedly as in the cell-based assays especially compared with K65R mutant alone. But both our enzymatic assays and molecular modeling show that the effect of each individual mutation combines in the double mutant. One can hypothesize that the dATP incorporation efficiency of K65R/L74V RT is too low to render a viable virus in vivo (Fig. 6).
Second, the lower fitness of K65R virus relative to the wildtype virus can be tentatively attributed to the poor purine incorporation efficiency of K65R. In other terms, there is an imbalanced nucleotide efficiency of incorporation between purines and pyrimidines in K65R. On the other hand, the loss of efficiency promoted by L74V is spread more between different nucleotides and might not impair HIV replication fitness as much as K65R does. If a 50% decrease in nucleotide incorporation efficiency has such an effect on viral fitness, the dNTP concentration must not be saturating in the cell. Intracellular   Table III. dNTP concentrations of 0.14 -5.6 M in resting lymphocytes (37-41), 2.4 -26 M in mitogen-stimulated lymphocytes (39,40), and 15-170 M in CEM lymphoblasts (37,38) have been reported. Therefore, we conclude that viral fitness must be drastically different according to cell type, and thus, viral fitness must be correlated to the existence of virus reservoir through the intracellular dNTP concentration.
Our results linking nucleotide efficiency of incorporation and viral replication capacity shed light on earlier works incriminating the processivity of DNA synthesis as the cause of decreased viral fitness (42)(43)(44). Processivity of DNA synthesis is defined as the number of nucleotides incorporated during a single cycle of polymerization prior to dissociation of the polymerase from the nascent DNA strand. Inherent to this definition, many altered biochemical properties of RT can account for a defect in processivity along the DNA polymerization pathway, such as primer/template binding (k on and/or k off ), nucleotide binding (K d ), nucleotide incorporation (k pol ), pyrophosphate release, or translocation to the next DNA template base. Unless a substitution is affecting the off-rate of RT from its primed DNA (k off ), and more exactly the ratio k pol /k off (45), the term processivity of DNA synthesis should not be used in a comparative analysis. Indeed, depending on the nucleotide concentration used, one could detect or not detect a change in processivity. This may explain why K65R has been reported to exhibit an increased (46), unchanged (5), or decreased processivity (29). We performed our processivity experiments under saturating nucleotide concentrations and found no difference between enzymes. In agreement with our analysis, White et al. (29) have performed a processivity assay under limiting nucleotide concentrations and found a difference between WT and K65R RT in the observed DNA product pattern. Our study points to the rate of nucleotide incorporation into DNA, which makes a difference between the enzymes studied here.
Finally our present data with the K65R and L74V mutations provide a second example of how combined resistance mutations affect viral replicative capacity. A prior example is given by M184V mutation whose presence in RT might confer a clinical benefit for antiretroviral therapies (47). The same observation has been described with the combination of 2Ј,3Јdideoxy-3Ј-thiacytidine and tenofovir selecting both M184V and K65R mutations (29). The doubly mutated virus, in a single cycle replication assay, could only retain 25% of the replication capacity of the WT virus. This defect is due to an impaired capacity of RT to incorporate natural nucleotides (7). The results presented here may help the selection of fitnessimpaired viruses that are potentially less pathogenic as well as aid the selection of drug combinations that have a prolonged efficacy through mutually exclusive resistance mechanism.