Enzymatic characterization of human immunodeficiency virus type 1 reverse transcriptase resistant to multiple 2',3'-dideoxynucleoside 5'-triphosphates.

A set of five mutations (A62V, V75I, F77L, F116Y, and Q151M) in the polymerase domain of reverse transcriptase (RT) of human immunodeficiency virus type 1 (HIV-1), which confers on the virus a reduced sensitivity to multiple therapeutic dideoxynucleosides (ddNs), has been identified. In this study, we defined the biochemical properties of RT with such mutations by using site-directed mutagenesis, overproduction of recombinant RTs, and steady-state kinetic analyses. A single mutation, Q151M, which developed first among the five mutations in patients receiving therapy, most profoundly reduced the sensitivity of RT to multiple ddN 5′-triphosphates (ddNTPs). Addition of other mutations to Q151M further reduced the sensitivity of RT to ddNTPs. RT with the five mutations proved to be resistant by 65-fold to 3′-azido-2′,3′-dideoxythymidine 5′-triphosphate (AZTTP), 12-fold to ddCTP, 8.8-fold to ddATP, and 3.3-fold to 2′,3′-dideoxyguanosine 5′-triphosphate (ddGTP), compared with wild-type RT (RTwt). Steady-state kinetic studies revealed comparable catalytic efficiency (kcat/Km) of RTs carrying combined mutations as compared with that of RTwt (<3-fold), although a marked difference was noted in inhibition constants (Ki) (e.g.Ki of a mutant RT carrying the five mutations was 62-fold higher for AZTTP than that of RTwt). Thus, we conclude that the alteration of RT's substrate recognition, caused by these mutations, accounts for the observed multi-ddN resistance of HIV-1. The features of multi-ddNTP-resistant RTs should provide insights into the molecular mechanism of RT discriminating ddNTPs from natural substrates.

The accumulating data suggest that the development of HIV-1 1 variants with reduced susceptibility to reverse transcriptase (RT) inhibitors is related to clinical deterioration in patients receiving RT inhibitors (1)(2)(3)(4)(5)(6). It has been shown that the high error rate of HIV-1 RT, approximately 1-10 misincor-porations/HIV-1 genome/round of replication (7)(8)(9), suggests that misincorporation by HIV-1 RT is responsible for the hypermutability of HIV-1, enabling HIV-1 to rapidly acquire drug resistance. This notable diversity of HIV-1 genome resulting from error-prone RT is likely to be associated with "natural" drug resistance seen in HIV-1 from patients receiving no prior antiretroviral therapy (10). In this regard, the combined use of multiple antiretroviral agents has been postulated to block or retard the emergence of HIV-1 less susceptible to therapeutic agents. However, several recent reports have demonstrated that HIV-1 can acquire resistance to multiple drugs in vitro and in vivo (11)(12)(13)(14)(15)(16). These data may suggest that even combination therapy may ultimately fail to suppress the replication of this hypermutable virus, although it is possible that combination chemotherapy continues to be efficacious if HIV-1 variants have a substantial replication disadvantage due to the altered enzymatic conformation/function as compared with wild type HIV-1; thus, the progression of the disease may be significantly delayed (17).
We and others have recently demonstrated that HIV-1 can develop a novel set of five mutations (A62V, V75I, F77L, F116Y, and Q151M) during combination chemotherapy with 3Ј-azido-2Ј,3Ј-dideoxythymidine (AZT; zidovudine) plus 2Ј,3Јdideoxyinosine (ddI; didanosine) or 2Ј,3Ј-dideoxycytidine (ddC; zalcitabine) and that these mutations confer resistance to various antiretroviral 2Ј,3Ј-dideoxynucleoside analogs (ddNs) on HIV-1 (11,13,18). In this study, we attempted to define the biochemical properties of RT carrying all or a subset of the five mutations. We constructed a cartridge mutagenesis system, which enabled us to introduce the desired mutations into relatively short fragments to generate both infectious HIV-1 clones (18) and recombinant RTs. We then examined the susceptibility of mutant RTs to multiple ddNTPs and determined steady-state kinetic constants. We also discuss the relationships of the observed HIV-1 resistance to multiple ddNs and altered RT functions caused by all or a subset of the five mutations.
Inc. All other reagents used in this study were of analytical grade. Oligonucleotide synthesis and purification using reverse phase chromatography were carried out by Lofstrand Labs (Gaithersburg, MD).
Plasmid Construction-Various RT expression vectors were generated using a BH10-derived RT overexpression vector, pKRT2 (20). First, an excised EcoRI-HindIII fragment containing the RT-coding region was subcloned to pTZ19R (Pharmacia), generating pTZRT2, which produced a single-stranded DNA using helper phage M13K07 (Life Technologies, Inc.). Following single-stranded DNA purification, site-directed mutagenesis was performed as directed in the manufacture's protocol (T7-Gen In Vitro Mutagenesis, U. S. Biochemical Corp.). Two restriction sites, XmaI and NheI, were introduced to pTZRT2 without deduced amino acid substitution (Fig. 1), and the first amino acid, proline, was substituted with alanine (an original BH10 amino acid), resulting in the loss of the NcoI site, giving rise to pTZRT06. The XmaI-NheI region of pTZRT06 was replaced with the XmNh oligonucleotide linker (5Ј-CCGGGATTTGATAGTCTAGAACAG-3Ј), generating pTZRT07, which was digested with EcoRI and HindIII and reinserted into pKRT2, generating pKRT07. An XmaI-NheI fragment excised from pTZRT06 was subcloned to pTZT, whose multicloning site of pTZ19R had been replaced with a linker (5Ј-AATTCGACGTCGTTAACGGGC-CCGCCGGCCCCGGGCTAGCTGGCCAT-3Ј) containing XmaI and NheI sites, generating pTZNX1. The entire RT-encoding regions of pTZRT07 and pTZNX1 were sequenced by the dideoxy sequencing method using Sequenase version 2 (U.S. Biochemical Corp.) as directed in the manufacturer's protocol, confirming the absence of any base substitutions other than the intended mutations. Mutations of interest (A62V, V75I, F77L, F116Y, and Q151M) were introduced by oligonucleotide-directed mutagenesis using the plus strand of pTZNX1 as a template. Mutagenesis was carried out by introducing one mutation in each mutagenesis reaction. Additional mutations were introduced by the same method as needed. Insertion of these intended mutations was confirmed by determination of nucleotide sequence. To construct a wild-type RT expression vector, pKRT08, the XmaI-NheI fragment of pTZNX1 replaced the XmNh linker region of pKRT07.
Enzyme Preparation-Escherichia coli JM109 (Promega, Madison, WI), transformed with a wild-type or mutant RT expression vector, were cultured overnight in ampicillin (100 g/ml)-containing 2 ϫ YT medium (0.32 g of tryptophane, 0.2 g of yeast extract, 0.1 g of NaCl, in 20 ml of distilled water), further propagated in the same medium (200 ml) with shaking for 2 h at 37°C and then exposed to 1 mM isopropyl-1-thio-␤-D-galactoside and cultured for an additional 7 h. Cells were harvested, frozen at Ϫ20°C overnight, thawed on ice, suspended in 1 ml of lysis buffer (50% sucrose, 50 mM Tris-Cl, pH 8.0, 5 mM EDTA, 50 g/ml lysozyme), and then disrupted by osmotic shock through rapid dilution with 10 ml of TE buffer (10 mM Tris-Cl, pH 8.0, 1 mM EDTA). Following a 15-min incubation on ice, the cell lysate was subjected to sonication and centrifugation, and the supernatant recovered was dialyzed against 50 mM Tris-Cl, pH 8.0, 200 mM NaCl overnight at 4°C. This RT-containing fraction was applied to a DE52-packed disposable column (5-ml bed, pre-equilibrated with the dialysis buffer), and the RT-containing flow-through fraction was collected. This pooled fraction was then applied to a Sepharose 4B column (1-ml bed) coupled to an RT-specific monoclonal antibody, M33, which reacts with both 51-(p51) and 66-(p66) kDa subunits of RT as described previously by Di Marzo Veronese et al. (19) and pre-equilibrated with 100 mM Tris-Cl, pH 8.0, 500 mM NaCl. Following thorough washing (15 ml) with the equilibration buffer, RT was eluted with 200 mM glycine-HCl, pH 2.8, and the RT-containing fraction was immediately neutralized with 1 ⁄10 volume of 1 M Tris. This RT preparation was concentrated using a Centricon-30 concentrator (Amicon, MA), dialyzed against 50 mM Tris-Cl, pH 8.0, 50 mM NaCl, 3 mM dithiothreitol overnight at 4°C, and stored at Ϫ70°C until use. The purity of thus prepared RT-derived polypeptides was Ͼ90% as assessed by a polyacrylamide gel analysis (see "Results").
Product Analysis-Products from experiments using 3 H-labeled nucleotides were analyzed using a DE81 filter (Whatman) binding assay. In order to remove unincorporated [ 3 H]nucleotides, filters were washed four times in 2 ϫ SSC (0.3 M NaCl, 0.03 M sodium citrate, pH 7.0), twice in ethanol, and once in acetone and dried in the air. The amount of [ 3 H]nucleotides incorporated into DNA and adsorbed on the filter was determined in liquid scintillation fluid using a scintillation counter.
Steady-state Kinetic Analysis-The buffer used for steady-state kinetic analysis for the homopolymeric template-primers was the same as buffer A or B with the exception of [ 3 H]dNTP concentrations and the volume of reaction mixture (100 l). The buffer composition for conducting a single nucleotide incorporation assay using 0.5 M MS2/22A (expressed as 3Ј-hydroxyl primer termini) was the same as that of buffer C except that only [ 3 H]dATP was present as a nucleotide substrate. Enzyme concentrations used were 1-3 nM for homopolymeric template-primer and 5-10 nM for MS2/22A. After the reaction mixtures were equilibrated at 37°C, the reaction was initiated by the addition of enzyme at 37°C. Portions of the reaction mixture (15 or 20 l) were removed periodically four or five times during the course of the assay and spotted on a DE81 filter, and the radioactivity in the product formed was counted as described above. The K m , K i , and V max values were determined from initial linear steady-state velocities with Lineweaver-Burk plot analyses, and the k cat values were calculated by dividing V max by active enzyme concentrations.

RESULTS
Overproduction and Purification of RT-Wild-type RT (designated RT wt ) and all mutant RTs were overproduced at about 1 mg of protein/200-ml culture and were purified using a Sepharose 4B column coupled to a monoclonal antibody, M33 (19). Thus obtained RT-derived polypeptides were Ͼ90% pure as assessed by an SDS-polyacrylamide gel analysis with Coomassie Brilliant Blue staining (Fig. 2). Although the mature HIV-1-associated RT is a heterodimer of p51 and p66 (19), all RT preparations produced were p66-dominant with a small amount of two truncated proteins of molecular masses of 62 and 55 kDa. Extended incubation of the p66-containing cytoplasm fraction (overnight at room temperature) did not change the p66/p51 ratio but decreased the total RT activity (data not shown).
In order to confirm that p66 homodimer RT preparations were suitable for the study of substrate recognitions between RT wt and mutant RT preparations, we subjected a p51/p66 heterodimer RT wt (21) to our assay system. With regard to dTMP incorporation into poly(rA)⅐(dT) [12][13][14][15][16][17][18] , the K m and k cat values of the p51/p66 heterodimer RT wt were 3.3 Ϯ 0.8 M and 2.0 Ϯ 0.1 s Ϫ1 , respectively (data not shown). These values were comparable with the K m and k cat values of our p66 homodimer RT wt , being 6.5 Ϯ 0.5 M and 1.8 Ϯ 0.3 s Ϫ1 , respectively (Table  II). Indeed, Beard et al. (24) have recently demonstrated that the two forms of RT wt , homodimer (p66/p66) and heterodimer (p51/p66), have comparable kinetic parameters of enzyme-template/primer interactions, concluding that the homodimeric form of RT can serve as a model for the interactions of heterodimeric form of RT with template-primers.
Sensitivity of RT with Various Mutations to Dideoxynucleotides-We first asked which of the five mutations (A62V, V75I, F77L, F116Y, and Q151M), seen in HIV-1 variants isolated from patients receiving combination chemotherapy (11,13,18,25), had the most profound effect on the sensitivity of RT against ddNTPs (Table I). The A62V mutant RT (designated RT 62 ) was slightly more sensitive to all ddNTPs examined, relative to RT wt . The V75I mutant RT (RT 75 ) was slightly less sensitive to ddNTPs except to ddATP, and both F77L and F116Y mutant RTs (RT 77 and RT 116 , respectively) had IC 50 values to ddNTPs tested comparable with that of RT wt . In contrast, the Q151M mutant RT (RT 151 ) proved to be substantially less sensitive to all ddNTPs tested (Table I).
In order to examine the effects of various combinations of mutations on the sensitivity of RT against ddNTPs, we introduced four mutations (A62V, V75I, F77L, F116Y) in addition to the Q151M mutation. In contrast to inconsiderable changes observed with single substitution of each of these four amino acids, combined mutations brought about marked increases in IC 50 values (Table I, Fig. 3). RT with four mutations (RT 75/77/116/151 ) and that with all five mutations (RT 62/75/77/116/ 151) had the most profound effect on the sensitivity of RT to ddNTPs tested (Table I).
We also examined the sensitivity of selected RTs to a nonnucleoside RT inhibitor, nevirapine (22), using [ 3 H]dGTP and poly(rC)⅐(dG) 12-18 as a nucleotide substrate and a templateprimer, respectively. The IC 50 values of nevirapine with RT wt and RT 62/75/77/116/151 were 0.50 and 0.59 M, respectively. These results are consistent with our previous observations that the infectious mutant HIV-1 carrying all five mutations was as sensitive to nevirapine as wild-type HIV-1 (18).
Determination of Concentrations of Active RT Molecules-Because of the nature of immunoaffinity chromatography, it was possible that improperly unfolded and/or degraded proteins were co-purified with active enzyme molecules. We therefore determined the concentration of active RT using the method previously described by Reardon and Miller (26). Since the dissociation of enzyme⅐template-primer complex (k off ) is slow relative to the rate of polymerization, the amplitude of the burst formation at a pre-steady state under the condition of single nucleotide incorporation is stoichiometric with regard to the number of active enzyme molecules (26 -28). In this assay, the incorporation of a single  Table I were analyzed, and virtually identical patterns were observed (not shown).  (Tables II and  V), and also agreed with previously reported values ranging 0.0065-0.06 s Ϫ1 for RT wt (26 -29). From the replot of the burst amplitudes (extrapolated y intercept) versus RT amounts added in the reactions, the concentration of active RT was determined (as shown in Fig. 4). These data indicated that approximately 40% of the total protein in the RT preparations represented active RT molecules. Substrate Analysis of RTs with Various Mutations-We also determined steady-state kinetic constants of RT wt and mutant RT preparations in processive (using poly(rA)⅐(dT) [12][13][14][15][16][17][18] or poly(rC)⅐(dG) 12-18 as a template-primer) and distributive (using poly(rI)⅐(dC) [12][13][14][15][16][17][18] or MS2/22A as a template-primer) modes. It was found that steady-state constants of RT wt to natural substrates were in the range previously reported by other groups (12, 26 -30) and that there were only up to 3-fold differences in K m , k cat , and k cat /K m values of RT wt and mutant RT preparations (Table II), suggesting that, like other ddN-associated RT mutations such as L74V, L74V/T215Y, T215Y (29), or K65R (30) the mutations examined in this study caused no significant detectable changes in the catalytic efficiency of RT. It should be noted, however, that a small difference in the enzymatic activity of RT may produce a considerable difference in the replication rate of HIV-1 in vivo (17).
We also determined steady-state kinetic constants using [ 3 H]ddATP as a nucleotide substrate. The K m(ddATP) value of RT wt and mutant RTs (Table IV) was virtually identical to its K i value with ddATP (Table III), consistent with a previous report demonstrating that the K i value of ddNTP against dNMP incorporation into a heteropolymeric template-primer was almost equal to K m(ddNTP) (27), indicating that ddNTP and dNTP behave as classical competitive substrates to each other under single nucleotide incorporation assay conditions (27).
RT Carrying Q151M and T215Y Mutations-Most of the HIV-1 variants carrying all or several of the five mutations, isolated from patients receiving combination chemotherapy with AZT plus ddC or AZT plus ddI, had the previously reported AZT, ddC, or ddI-associated pol gene mutations except for K219Q and K219E, and none had the most potent AZTassociated mutation, T215Y (11,13,18). In this regard, it is possible that one or more of the five mutations are incompatible with AZT, ddC, or ddI-associated mutations. In order to test this possibility, we produced a recombinant RT (RT 151/215 ) carrying the Q151M mutation and T215Y (32). We found, however, that RT 151/215 had an enzymatic activity comparable with that of RT wt and was less susceptible to all four ddNTPs tested, ddATP, ddCTP, ddGTP, and AZTTP (Table V). DISCUSSION The mutations responsible for resistance of HIV-1 against nucleoside RT inhibitors have been mapped to the RT-encoding region of the pol gene, and accumulated data suggest that altered substrate recognition by RT is associated with drug resistance (29,30,33). However, the changes in the sensitivity of RT to the triphosphate of a ddN appear to account partly for in vitro resistance of HIV-1 to the ddN (29,34). Indeed, several enzymatic studies have demonstrated that the sensitivities of RT to ddNTPs differ from the in vitro viral sensitivities to ddNs (31,34,35). For instance, a recombinant infectious HIV-1 carrying four AZT-associated amino acid substitutions (D67N, K70R, T215Y, and K219Q) is about 100-fold less sensitive to AZT than the wild-type HIV-1 strain in vitro (34); however, an RT with the same four amino acid substitutions is as sensitive to the inhibition of AZTTP as is the wild-type RT (RT wt ) (31).
Prasad et al. (35) have also demonstrated that HIV-1 carrying a mutant RT with an E89G substitution, which showed crossresistance to various ddNTPs including ddGTP, was sensitive  Table III. K i values determined from the x intercept of the replot are summarized in Table IV. to ddG in a cell culture system. These observations do not appear to be easily explained under one unifying theory at present. In the present study, we demonstrate that a unique set of five mutations (A62V, V75I, F77L, F116Y, and Q151M) alters RT's substrate recognition and confers resistance to various ddNTPs on RT, which may fully account for the observed in vitro resistance of HIV-1 to multiple ddNs (18).
It is evident that Gln 151 plays a critical role in the enzymatic function of RT and is linked to the development of multi-ddN resistance of HIV-1 (18). The amino acid Gln 151 consists of the highly conserved sequence, Leu-Pro-Gln-Gly, in the corresponding region within motif B of RT from various animal and human retroviruses (36); however, the substituted amino acid, methionine, is found in the corresponding region of hepatitis B viruses (HBVs), suggesting that Q151M is not overly detrimental to the function of RT. We found in this study that the Q151M substitution decreases the polymerase activity by 30% (as assessed with k cat /K m ; Table II) with several dNTPs and significantly reduces the sensitivity of RT to ddNTPs. In this regard, various groups have examined the effect of amino acid substitution of Gln 151 (37)(38)(39). A Q151N substitution has shown no significant alteration in the polymerase or RNase H activities of RT (38), A Q151H substitution, however, has been shown to reduce the polymerase activity by 30% and the sensitivity of RT to AZTTP by 4 -50-fold (39), and a Q151E substitution has been reported to reduce the polymerase activity by 30% (37) as compared with RT wt .
The amino acid Phe 116 , which is located close to the proposed dNTP-binding site of HIV-1 RT (40), is also part of a highly conserved region (motif A) in RTs from various retroviruses, but the corresponding amino acid in RT in hepatitis B viruses has been found to be tyrosine (36), which may explain why the RT with such a substitution at a highly conserved amino acid remains functional. In fact, Boyer et al. (38) have recently reported that F116Y alone does not affect the polymerase and RNase H activity of RT . In our study, RT 116 carrying F116Y was also functional and behaved similarly to RT wt against ddNTPs (Table I). It is not clear how F116Y affects the function of RT when other mutations are added. The addition of the V75I mutation to RT 77/116/151 , generating RT 75/77/116/151 , further reduced the sensitivity to ddNTPs, although V75I alone had little effect on the sensitivity (Table I). Codon 75 mutation with a different amino acid, V75T, has been identified in HIV-1 variants less susceptible to 2Ј,3Ј-didehydro-2Ј,3Ј-dideoxythymidine, ddC, and ddI in vitro (41). These data suggest that the amino acid at codon 75 of HIV-1 RT plays a crucial role in substrate recognition.
Detailed steady-state kinetic analyses revealed that the altered nucleotide substrate recognition of RT is evidently caused by all or a subset of the five mutations, suggesting that these amino acids interact with an incoming nucleotide substrate, although presently it is not possible to distinguish direct effects caused by amino acid substitutions from those mediated via template-primer interactions (40,42). It has been reported that Gln 151 is located close to the first single-stranded base of the template and that three amino acids (Val 75 , Phe 77 , and Gln 151 ) form part of the "template grip" (43). This may lead to repositioning or conformational change of the template-primer, causing a distortion of the geometry of the polymerase active site, enabling RT to discriminate natural substrates from multiple ddNTPs (40,43,44). In this regard, we attempted to define the difference between RT wt and mutant RTs with respect to the RT's interaction with the template-primers. We then found that all RT wt and mutant RTs examined had a virtually identical profile in the time course of dAMP incorporation into MS2/22A (Fig. 4). This steady-state rate, dominated by the dissociation of enzyme⅐template-primer complex (26 -28), was found comparable in all mutant RTs as compared with that of RT wt (within 2-fold), suggesting that none of the five mutations significantly affects the dissociation of the enzyme from the enzyme⅐template-primer complex. We also found that there was no significant difference in k off of ddAMP-forced terminated template-primer for all RTs tested (Table V; k cat ϭ k off ), suggesting that the resistant phenotypes of RTs do not involve the dissociation of the ddAMP-forced terminated templateprimer. Furthermore, among RTs examined we have found no significant difference (3-fold) in their K d values to poly(rA)⅐(dT) 10   RT 62/75/77/116/151 , and RT 151/215 , respectively. 2 Moreover, we have found no significant changes in the processivity of RTs as measured by dTMP incorporation into poly(rA)⅐d(T) 16 (data not shown). It is worth noting that in the present work, bacterially expressed p66 homodimer RTs were employed; thus, the observed changes caused by amino acid substitutions might differ from changes that may occur in the p51/p66 heterodimer RTs. In this regard, Becerra et al. have reported a difference in association constants upon dimerization of p66/p66 and p51/ p66 subunits of RT (46), which may produce changes in enzyme-substrate interactions differently between heterodimer and homodimer RTs when amino acid substitutions occur. Further experiments using proper heterodimer RTs and pre-steady kinetic analysis are needed to define the structures and functions of HIV-1 RT when these mutations develop. It is noteworthy that HIV-1 strains carrying Q151M often do not bear amino acid mutations that have been associated with viral resistance to AZT, although HIV-1 variants were isolated from patients receiving long-term therapy including AZT (11,13,18). It is possible that certain nucleotide or amino acid sequence(s) inherent to HIV-1 strains in certain patients predispose these viruses not to acquire any AZT-related mutations but to develop the Q151M substitution and ultimately the rest of the five mutations. Comparative nucleotide sequence analysis of pretherapy HIV-1 that later acquired Q151M and pretherapy strains that later developed the AZT-related mutations may define such predisposing nucleotide or amino acid sequence(s). It remains to be determined whether HIV-1 carrying Q151M can subsequently acquire the AZT-associated amino acid substitutions. It is also of interest to study the mechanisms of substrate recognition or discrimination of ddNTPs by three-dimensional structure analysis using mutant RTs with all or a subset of the five mutations. Such studies may provide us with insight to the understanding of molecular mechanisms of substrate recognition and catalytic function of HIV-1 RT.