HIV-1 Reverse Transcriptase (RT) Polymorphism 172K Suppresses the Effect of Clinically Relevant Drug Resistance Mutations to Both Nucleoside and Non-nucleoside RT Inhibitors*

Background: The effect of HIV polymorphisms in drug resistance is unknown. Results: RT polymorphism 172K suppresses resistance to nucleoside (NRTIs) and non-nucleoside RT inhibitors (NNRTIs) by decreasing DNA binding and restoring NNRTI binding. Conclusion: 172K is the first HIV polymorphism suppressing resistance to diverse inhibitors. Significance: Results provide new insights into interactions between the polymerase active site and NNRTI-binding sites. Polymorphisms have poorly understood effects on drug susceptibility and may affect the outcome of HIV treatment. We have discovered that an HIV-1 reverse transcriptase (RT) polymorphism (RT172K) is present in clinical samples and in widely used laboratory strains (BH10), and it profoundly affects HIV-1 susceptibility to both nucleoside (NRTIs) and non-nucleoside RT inhibitors (NNRTIs) when combined with certain mutations. Polymorphism 172K significantly suppressed zidovudine resistance caused by excision (e.g. thymidine-associated mutations) and not by discrimination mechanism mutations (e.g. Q151M complex). Moreover, it attenuated resistance to nevirapine or efavirenz imparted by NNRTI mutations. Although 172K favored RT-DNA binding at an excisable pre-translocation conformation, it decreased excision by thymidine-associated mutation-containing RT. 172K affected DNA handling and decreased RT processivity without significantly affecting the kcat/Km values for dNTP. Surface plasmon resonance experiments revealed that RT172K decreased DNA binding by increasing the dissociation rate. Hence, the increased zidovudine susceptibility of RT172K results from its increased dissociation from the chain-terminated DNA and reduced primer unblocking. We solved a high resolution (2.15 Å) crystal structure of RT mutated at 172 and compared crystal structures of RT172R and RT172K bound to NNRTIs or DNA/dNTP. Our structural analyses highlight differences in the interactions between α-helix E (where 172 resides) and the active site β9-strand that involve the YMDD loop and the NNRTI binding pocket. Such changes may increase dissociation of DNA, thus suppressing excision-based NRTI resistance and also offset the effect of NNRTI resistance mutations thereby restoring NNRTI binding.

There are two main mechanisms for resistance to NRTIs. In the first mechanism, RT preferentially decreases incorporation of NRTI-triphosphates (TPs), while retaining the ability to use natural nucleotide substrates. This type of resistance is typically imparted by mutations close to the nucleotide-binding site of RT. For instance, K65R, L74V, and Q151M decrease the incorporation rate of NRTI-TPs (7)(8)(9)(10), whereas M184V sterically hinders productive binding of lamivudine (3TC)-TPs at the dNTP-binding site (11). The second mechanism of NRTI resistance involves unblocking of the NRTI-terminated primers by an excision activity that uses ATP (12)(13)(14)(15). This excision activity is enhanced (12) in the presence of mutations such as M41L, D67N, K70R, L210W, T215Y/F, and K219Q/E (thymidine-associated mutations, TAMs), which are selected during zidovudine (AZT) or stavudine (d4T) therapy (16,17). It has been reported by several groups, including ours, that other RT mutations located far from the polymerase active site at the connection subdomain (E312Q, G335C/D, N348I, A360I/V, V365I, and A376S) confer resistance to NRTIs and/or NNRTIs (18 -25). It has been proposed that reduction of RNase H cleavage caused by connection subdomain mutations contributes to NRTI resistance by providing more time for RT to carry out excision and resume productive DNA synthesis (24 -28).
There are several examples of RT mutations that cause resistance to one drug and affect the emergence of resistance mutations to another drug. For example, treatment of patients with AZT and 3TC combinations often results in the emergence of the 3TC resistance mutation M184V, but it also delays acquisition of TAMs (29,30). In addition, appearance of the primary didanosine (ddI) resistance mutation L74V precludes AZT resistance conferred by TAMs (29,31). Conversely, appearance of the first TAM (T215Y) during AZT and ddI combination therapy suppresses emergence of L74V (32). Tenofovir (TDF) resistance mutation K65R has a strong negative association with TAMs but not with other NRTI mutations, including Q151M complex (Q151Mc) (33). The bidirectional phenotypic antagonism between K65R and TAMs suppresses not only AZT resistance conferred by TAMs but also abacavir (ABC), ddI, and TDF resistance conferred by K65R (33,34). Moreover, NNRTI resistance mutations L100I and Y181C are also antagonistic to AZT resistance by TAMs (32,35,36) because they reduce ATPmediated unblocking of AZT-terminated primers (34,35,(37)(38)(39). Such antagonistic interactions between resistance mutations impart significant clinical benefits. Hence, to optimize clinical decisions, it is very important to understand how mutations may affect the phenotype of known drug resistance mutations.
Although codon 172 of HIV-1 RT is usually an arginine (172R), a lysine (172K) polymorphism is also found in clinical samples (up to 1.0%, as reported at the HIV Drug Resistance Database) and in the BH10 laboratory strain, which is very commonly used in drug resistance studies. This study uses virological, biochemical, and structural tools to reveal the effect of 172K on NRTI NNRTI. We report that 172K significantly suppresses resistance to both NRTIs and NNRTIs, and we propose specific biochemical mechanisms for these effects.

EXPERIMENTAL PROCEDURES
Antiviral Agents-AZT, d4T, and ddI were purchased from Sigma. 3TC and TDF were purchased from Moravek Biochemicals, Inc. (Brea, CA). ABC, nevirapine (NVP), and efavirenz (EFV) were provided by the AIDS Research and Reference Reagent Program (National Institutes of Health).
RT mutations were introduced by site-directed mutagenesis as described previously (20,43). The pNL101 HIV-1 infectious clone was provided by Dr. K.-T. Jeang (National Institutes of Health, Bethesda) and used for generating recombinant HIV-1 clones. Wild-type HIV-1 (HIV-1 WT ) was constructed by replacing the pol-coding region of pNL101 (nucleotides position; nucleotide 2006 of ApaI site to 5785 of SalI site of pNL101) with HIV-1 BH10 strain. We introduced a silent mutation at nucleotide 4232 (TTTAGA to TCTAGA) of the pol-coding region to generate a unique XbaI site. The desired mutations were introduced into the XmaI-XbaI region (1643 bp), which encodes nucleotides 2590 -4232 of pNL101. This cassette was cloned into the respective sites of pBluescript vector (Stratagene) and introduced mutation(s) using an oligonucleotidebased site-directed mutagenesis method. After mutagenesis, the XmaI-XbaI cassettes were inserted back into pNL101 and confirmed by sequencing. Viral stocks were obtained by transfection of each molecular clone into COS-7 cells, harvested, and stored at Ϫ80°C until use.
Drug Susceptibility Assays-Single replication cycle drug susceptibility assays were performed in triplicates using MAG-IC-5 cells as described previously (42). Briefly, MAGIC-5 cells were infected with diluted virus stock at 100 blue forming units in the presence of increasing concentrations of RTIs, cultured for 48 h, fixed, and stained with 5-bromo-4-chloro-3-indolyl-Dgalactopyranoside (X-gal). The stained cells were counted under a light microscope. The susceptibility of RTIs was calculated as the concentration that reduces blue forming units (infection) by 50% (50% effective concentration [EC 50 ]).
Enzymes and Nucleic Acid Substrates-Mutant RTs were generated through site-directed mutagenesis and replaced into the pRT dual vector using restriction sites PpuMI and SacI for the p51 subunit and SacII and AvrII for the p66 subunit. Heterodimeric HIV-1 RTs (p66 and p51) were expressed in Escherichia coli, BL21, and purified as described previously (44,45).
Primer Extension Assays-Primer extension assays were performed in the presence of AZT-TP or NVP as we described previously (44). Reactions contained 50 nM T 100 /5Ј-Cy3-P 18long mixed with 120 nM RT (in experiments with AZT-TP) or 80 nM RT (in experiments with NVP) in a buffer containing 50 mM Tris-HCl, pH 7.8, and 50 mM NaCl, 1 M of each dNTP, 0.5 mM EDTA, and varying concentrations of AZT-TP or NVP. In NVP-containing reactions, RT was preincubated with template/primer (T/P) and inhibitor at 37°C for 5 min. DNA synthesis was initiated by the addition of 10 mM MgCl 2 . Reactions were terminated after 40 min (AZT-TP) or 30 min (NVP) by adding an equal volume of 100% formamide-containing traces of bromphenol blue. Extension products were loaded on a 7 M urea, 15% polyacrylamide gel. The gels were scanned using a FLA-5000 PhosphorImager (FujiFilm, Stamford, CT). The amounts of full-length extended and unextended products were quantified by densitometry using MultiGauge, and the results were plotted using GraphPad Prism 4 (GraphPad Software Inc.).
Site-specific Fe 2ϩ Footprinting Assays-Site-specific Fe 2ϩ footprinting assays were performed using 5Ј-Cy3-labeled DNA templates as described previously (46,47). Briefly, 100 nM 5Ј-Cy3-T 43 /P 30 was incubated at 37°C with HIV-1 RT (600 nM) for 30 min in a buffer containing 120 mM sodium cacodylate, pH 7.0, 20 mM NaCl, 6 mM MgCl 2 , and 10 M AZT-TP to allow quantitative chain termination. Complexes were preincubated for 7 min with increasing concentrations of the next nucleotide (dATP), and 1 mM ammonium iron sulfate was added. The reactions were quenched after 5 min with 30 l of formamide containing bromphenol blue. The products were resolved with gel electrophoresis in 7 M urea, 15% polyacrylamide gels.
Surface Plasmon Resonance Assay-We used surface plasmon resonance (SPR) to measure the binding affinity of RT 172K and RT 172R to double-stranded DNA. Experiments were performed on a Biacore T100 instrument (GE Healthcare). To prepare the sensor chip surface, we used the 5Ј-biotin-T d37 /P d25 (5Ј-biotin-TAG ATC AGT CAT GCT CCG CGC CCG AAC AGG GAC TGT G-3Ј, annealed to P d25 5Ј-CAC AGT CCC TGT TCG GGC GCG GAG C-3Ј). Approximately 100 resonance units of this DNA duplex were bound in a channel of a streptavidin-coated sensor chips (Series S Sensor Chip SA) by flowing 0.1 M DNA (flow rate 10 l/min) in 50 mM Tris, pH 7.8, 50 mM NaCl. The binding constants were determined as follows: RT binding was observed by flowing solutions contain-ing increasing concentrations of the enzyme (0.5, 1, 2, 4, 7.5, 10, 15, and 20 nM) in 50 mM Tris, pH 7.8, 60 mM KCl, 10 mM MgCl 2 in sample and background channels at 30 l/min. The background traces were subtracted from the traces of the corresponding samples to obtain the binding signal of RT. This signal was analyzed using the Biacore T100 Evaluation software to determine K D , k on , and k off values.
Enzyme Processivity Assays-Processivity assays were performed in the presence of a heparin trap to ensure that each synthesized DNA molecule resulted from a single processive cycle. Twenty nanomolar T 100 /5Ј-Cy3-P 18long was preincubated with 500 nM RT at 37°C in a buffer containing 50 mM Tris-HCl, pH 7.8, 50 mM NaCl, 50 M of each dNTP, and 0.1 mM EDTA for 5 min. DNA synthesis was initiated by the addition of 10 mM MgCl 2 and 2 mg/ml heparin. Reactions were terminated after 0, 15, and 90 min by adding an equal volume of 100% formamide containing traces of bromphenol blue. Extension products were loaded on a 7 M urea, 15% polyacrylamide gel, quantified, and analyzed as described above.
Steady State Kinetics-Steady state parameters K m and k cat for incorporation of a nucleotide were determined using a single nucleotide incorporation gel-based assay. Reactions with RT 172K and RT 172R were carried out in 50 mM Tris-HCl, pH 7.8, 50 mM NaCl, 6 mM MgCl 2 , 100 nM T/P, 10 nM RT, respectively, and varying concentrations of dNTP in a final volume of 10 l. The reactions for HIV-1 RT were carried out in Reaction Buffer with 100 nM T 31 /5Ј-Cy3-P 18 . Reactions were stopped after 1 min at 37°C, and the products were resolved and quantified as described above. K m and k cat values were determined graphically by using the Michaelis-Menten equation.
Crystallization of RT-RT with mutations K172A and K173A was prepared as described by Bauman et al. (48). These changes have been reported to strongly improve diffraction of RT. The enzyme was crystallized by the hanging drop vapor diffusion technique at 4°C. The well solution contained 50 mM BisTris, pH 6.8, 100 mM ammonium sulfate, 10% glycerol, and 12% PEG 8000. Two l of the well solution was mixed with 2 l of RT (30 mg/ml) containing 10 mM MgCl 2 , 10 mM Tris, pH 7.8, and 25 mM NaCl. The drop was equilibrated against the well solution by hanging a coverslip over the sealed well. The drops were streak-seeded with crushed RT crystals after 7 days of incubation. Crystals were obtained 7-14 days after the drops were seeded. Single crystals were soaked in cryoprotectant solution containing the well solution supplemented with 22.5% ethylene glycol for 15-60 s and frozen in liquid nitrogen.
Structure Solution-Diffraction datasets were collected at the Advanced Light Source Synchrotron beamline 4.2.2. The data were processed to 2.15 Å using MOSFLM (49) (supplemental Table 1). Molecular replacement phasing was performed using Phaser (50) and a previously solved RT structure as a model (Protein Data Bank (PDB) code 3KLI). The final model was obtained after cycles of iterative model building in COOT (51) and restrained refinement with CNS (52) and REFMAC (53). Final statistics for data processing and structure refinement are listed in supplemental Table 1. Coordinates and structure factors for the crystal structure were deposited to the PDB (PDB code 4DG1).
Structural Analysis-The program COOT was used to align crystal structure coordinates (typically using residues 107-112 and 151-215 of the p66 subunit) of the following complexes: RT 172R RT-EFV with RT 172K RT-EFV (PDB code numbers 1FK9 and 1IKW), RT 172R/K103N RT-EFV with RT 172K/K103N RT-EFV (PDB code numbers 1FKO and 1IKV), and RT 172R RT-NVP with RT 172K RT-U05 (a NVP analog) (PDB code numbers 1RTH and 3HVT). We also compared RT-DNA-dNTP ternary complexes (1RTD and 3JYT, as examples of RT 172R and RT 172K RTs). Figures were generated using PyMOL Molecular Graphics Program.

RESULTS
Effect of 172K on NRTI Resistance-To determine the effect of 172R or 172K on NRTI resistance, we generated RT 172R and RT 172K mutants carrying various NRTI-resistant mutations ( Fig. 1 and supplemental Table 2) and compared the EC 50 value of HIV-1 172R with that of HIV-1 172K in the background of NRTI-resistant mutations shown as a "fold reduction" in the Fig. 1. Although 172K alone had no effect on AZT or ddI susceptibility, 172K significantly reduced the AZT resistance of the following mutants that are associated with the excision mechanism: N348I, 69ins complex N348I/TAM-1 and TAM-2. Similarly, 172K appeared to suppress ddI resistance of a mutant that is associated with the excision mechanism TAM-2. In contrast, 172K had no statistically significant impact on AZT or ddI resistance of viruses with mutations that cause drug resistance by decreasing incorporation of NRTIs (K65R, L74V, M184V, and Q151Mc) (Fig. 1).
When we examined the effect of 172K on viruses that had combinations of mutations that cause NRTI resistance through the excision mechanism and through the decreased incorporation mechanism, we observed that 172K significantly suppressed resistance of L74V/TAM-1, L74V/N348I/TAM-1, and M184V/N348I, to AZT and to ddI (Fig. 1). These data suggested that 172K augments the suppressive effects of L74V and M184V on AZT and ddI resistance.
To further examine the effect of 172K on resistance to other Food and Drug Administration-approved NRTIs, we focused on the susceptibility of the 69ins complex to d4T, ABC, TDF, and 3TC (Table 1). Our data show that the 172K polymorphism significantly suppressed resistance of the 69ins complex to d4T (3-fold reduction; Table 1) and to AZT (11-fold reduction; Fig.  1). However, the resistance of 69ins complex to 3TC (54), ABC, and TDF was not significantly affected by 172K (less than 1.2fold reduction in resistance). These results indicate that the suppressive effect of 172K can be NRTI-specific.
Effect of 172K on NNRTI Resistance-The antagonistic effects of 172K on the NNRTI resistance of RTs with mutations at K103N, V106M, V108I, Y181C, Y188L, and N348I are shown in Fig. 2 (and also supplemental Table 3). Fold reduction was FIGURE 1. Effect of the 172K polymorphism on HIV-1 susceptibility to AZT and ddI. Antiviral activities of HIV-1 carrying NRTI resistance mutations with 172R (black bars) or 172K (gray bars) were determined by MAGIC-5 cell-based assay and shown as the concentration required for 50% inhibition of virus infection (EC 50 ). "TAM-1 and -2" have the "M41L and T215Y" and "D67N, K70R, T215F, and K219Q" mutations, respectively. 69ins complex carries 69 insertion and TAM-1. Q151Mc complex carries Q151M, A62V, V75I, F77L, and F116V. Error bars represent standard deviations from at least three independent experiments. Asterisks mark statistically significant differences (p Ͻ 0.05 by t test) in EC 50 values comparing 172K with 172R in the background of NRTI-resistant mutations. Susceptibility of the Q151Mc complex with 172R or 172K to ddI was not assessed because the EC 50 values were over the detection limit for this assay (Ͼ100 M). calculated as "EC 50 of 172R/EC 50 of 172K." 172K significantly suppressed NVP resistance conferred by K103N, V106M, V108I, and N348I. The impact of 172K on NVP resistance in the context of Y181C or Y188L could not be assessed, because both exhibited very high resistance (Ͼ10 M) under these conditions. 172K also reduced EFV resistance of V106M, V108I, and Y181C. Hence, the impact of 172K on EFV resistance was considerably lower than that on NVP resistance. Collectively, our data demonstrate that RT resistance to NNRTIs (especially to NVP) can also be affected by 172K. A similar effect has been observed in assays using purified RTs containing either 172R or 172K. A lysine at position 172 conferred a higher susceptibility to NVP, while displaying only marginal differences for EFV, etravirine, and rilpivirine when compared with RT containing an arginine at that position (data not shown).
Effect of 172K on AZT Resistance of HIV-1 RT-TAMs cause AZT resistance by enhancing the ability of RT to remove the terminal AZT-monophosphate (AZT-MP) from the 3Ј-primer terminus using a pyrophosphate donor such as ATP. To deter-mine whether 172K suppresses AZT resistance through a reduction in the efficiency of ATP-mediated excision, we purified RT 172R and RT 172K with TAM-2. We measured the susceptibility of RTs to inhibition by AZT in the presence or absence of ATP in a primer extension assay that uses long T/P (T 100 /5Ј-Cy3-P 18long ) ( Table 2 and also supplemental Fig. 1). Any changes in AZT susceptibility only in the presence of ATP would suggest that the effect is excision-dependent. RT 172R/TAM-2 showed ϳ4-fold higher IC 50 in the presence of ATP (995 versus 250 nM), although RT 172K/TAM-2 exhibited only a 1.8-fold excision enhancement by ATP (517 versus 290 nM). Hence, the ATP-based (or excision-based) increase in AZT resistance in TAM-2 was at least twice as much in the presence of 172R than 172K (4-fold versus 1.8-fold excision enhancement). However, in the absence of TAMs the effect of ATP was comparable for RT 172R and RT 172K (1.3-fold versus 1.2-fold excision enhancement) and smaller than in the case of the TAM enzymes. To further confirm these results, we also performed ATP-rescue assay using RT 172R/TAM-2   and RT 172K/TAM-2 . The rescue assays involve unblocking of AZT-terminated primers through ATP-based excision and DNA synthesis. RT 172K/TAM-2 also decreased ATP-excision reactions (data not shown). Hence, the suppressive effect of 172K in AZT resistance is ATP excision-dependent.

Effect of 172K on the Translocation Site of HIV-1 RT Measured by Fe 2ϩ
Footprinting-A single cycle of DNA polymerization involves dNTP binding, incorporation, and translocation of the elongated DNA primer from the dNTP-binding site (pretranslocation site, N-site) to the priming site (post-translocation site, P-site). For efficient ATP-based excision, the AZT-MP-terminated primer needs to be positioned at the pre-translocation site (14,47). To determine whether 172K influences the positioning of AZT-MP-terminated primers, we performed a site-specific Fe 2ϩ footprinting assay (Fig. 3). We mixed various RTs with AZT-terminated DNA and different concentrations of the incoming dATP substrate to form dead end complexes that position DNA in a post-translocation site. Treatment with hydroxyl radicals prepared in the presence of Fe 2ϩ allowed cleavage at positions that correspond to pre-or post-translocation sites, and we monitored translocation as a function of dATP required to translocate 50% of RT. In this assay, lower numbers indicate more efficient translocation. Our data indicated that RT 172R/TAM-2 (or RT 172K/TAM-2 ) required higher dATP concentrations than RT 172R (or RT 172K ) to translocate by 50% (327 versus 128 M or 940 versus 388 M). These data are consistent with previous reports that AZT-resistant RTs with excision mutations primarily bind in a pre-translocation site (47) and is thus more accessible for ATP-based excision. Interestingly, although RT 172K/TAM-2 has suppressed AZT resistance compared with RT 172R/TAM-2 , it still bound the AZT-MP-terminated primer more efficiently in an excision-competent pre-translocation site (940/327 ϭ 2.9-fold). Taken together, our data suggest that the decreased excision by 172K is not due to repositioning of the AZT-MP-terminated primer at the nonexcisable post-translocation site.

DNA Binding by HIV-1 RT Measured by Surface Plasmon
Resonance-We hypothesized that the decreased excisionbased resistance of 172K is due to a more efficient dissociation of the nucleic acid substrate, such that it decreases the opportunities to unblock chain terminated primers, thus resulting in suppression of AZT resistance. Hence, we used SPR to measure DNA binding and compare the DNA binding affinities of RT 172K and RT 172R . We chose SPR because measurements of the dissociation constant, K D. DNA , using gel-mobility shift assays, do not offer insights regarding the kinetics of binding (k on ) and release (k off ) of nucleic acid from the enzyme. We attached biotinylated DNA on a streptavidin sensor chip and flowed various concentrations of either enzyme over the chip to measure the association (k on ) and dissociation (k off ) rates of the enzymes in real time (Fig. 4). The k off value for RT 172K was markedly increased (31-fold) with a slightly changed k on value (2.9-fold) compared with those for RT 172R . The K D. DNA (ϭk off / k on ) value for RT 172K was 11-fold higher than that for RT 172R (1.8 and 0.16 nM, respectively). Our results demonstrate that RT 172K had lower DNA binding affinity than RT 172R due to a significant difference in the dissociation rate of RT from the DNA.
Processivity of HIV-1 RT-Considering that low DNA binding affinity by 172K may contribute to decreased processivity, we carried out processivity assays using RT 172K or RT 172R in the absence or presence of the NRTI resistance background. Assays were carried out using a long DNA template in the presence of heparin as a competitive trap (Fig. 5). Full-length product formation was less observed in RT 172K compared with RT 172R , indicating that 172K attenuates processivity. When introduced into RT TAM-2 , 172K lowered processivity even more. In contrast, Q151Mc may enhance RT processivity, as Q151Mc was more processive than WT, especially in the presence of RT 172K . These data suggest that 172K could decrease RT processivity in the background of only excision and not discrimination resistance mutations. Steady State Kinetics of Nucleotide Incorporation-Initial polymerase activity comparisons of RT 172R and RT 172K showed that 172K slowed the polymerase activity of RT. This observation led us to investigate the steady state nucleotide incorporation properties of RT 172R and RT 172K using single nucleotide incorporation assays ( Table 3). The estimated values for k cat and K m.dNTP show that RT 172K and RT 172R had comparable efficacies (k cat /K m.dNTP of 0.05 versus 0.03 min Ϫ1 nM Ϫ1 ) under steady state conditions. These data suggested that 172K decreased RT processivity without affecting catalytic efficiency for nucleotide incorporation.
Susceptibility of HIV-1 RT to NVP-To examine the suppressive effect of 172K on NNRTI resistance at the HIV-1 RT enzyme level, we purified HIV-1 RT with V106M in the background of Arg or Lys at RT codon 172 and performed primer extension assays in the presence of NVP (Table 4). When introduced into RT V106M , 172R results in an enzyme (RT 172R/V106M ) with 57-fold resistance to NVP. In contrast, RT 172K/V106M showed decreased resistance to NVP (6-fold). Hence, the extent of the suppressive effect of 172K on NVP resistance at the enzyme level was comparable with the effect observed at the virus level.
Crystal Structure of HIV-1 RT with Mutation at Codon 172-The structure of RT with the K172A mutation was determined at 2.15 Å resolution (supplemental Table 1). This is one of the  A and 172K in B). Increasing concentrations (0.5, 1, 2, 4, 7.5, 10, 15, and 20 nM) of each RT were flowed over a streptavidin chip with biotinylated double-stranded DNA (5Ј-biotin-T d37 /P d25 ) immobilized on its surface as described under "Experimental Procedures." The experimental trace (red) shown is the result of a subtraction of the data obtained from the channel containing the immobilized DNA minus the signal obtained from a control/background channel. DNA binding constants for RT 172R or RT 172K are shown in C.

FIGURE 5. Effect of changes at codon 172 on the processivity of HIV-1 RT.
A, processive DNA synthesis was measured at 0, 15, and 90 min after initiation of the reaction in the presence of heparin trap for each RT enzyme (WT, TAM-2, and Q151Mc) with 172R or 172K. Conditions were selected so that in the absence of heparin trap (1st lane of every set, labeled as ϪHeparin), the sum of processive and distributive DNA synthesis was the same (comparable full-length product in ϪHeparin lanes). All experiments were repeated three times and a representative gel is shown here. B, amount of full-length extended and un-extended products at 90 min after initiation of the reaction were quantified by densitometry using MultiGauge. Percentages of fulllength DNA synthesis are shown in B.
The data are means Ϯ S.D. from at least three independent experiments. b Fold increase was compared with WT 172R. c These data were obtained from primer extension assay in the presence of NVP. d These data were obtained from cell-based assay using MAGIC-5.
highest resolution structures of HIV-1 RT. In addition to this mutation, RT had also the K173A mutation to improve the crystallographic properties of the enzyme (48). However, we confirmed K173A did not affect susceptibility of RT to NVP and AZT, and it also did not alter the effect of the 172R or 172K change to susceptibility to RTIs at the virus or at the enzyme level (data not shown). The overall conformation of the mutant RT is similar to that of WT unliganded RT in which the p66 thumb subdomain is folded down into the DNA binding cleft (55). However, there are notable changes proximal to mutated residue 172 of the ␣-helix E, which affect interactions with residues of the ␤9-strand that leads to the YMDD loop of the polymerase active site. The Lys-172-Ile-180 interaction is lost in the 172A structure. This change is accompanied by a repositioning of the amide side chain of Gln-182, which in the 172K structure stabilizes the Met-184 main chain amide of the YMDD loop, whereas in the 172A structure it makes hydrogen bonds with the Thr-165 hydroxyl group and the Gln-161 side chain amide. Hence, the changes in the interactions between 172 and 180 propagate toward the polymerase active site (Fig. 6).
Effect of 172K on the Structure of HIV-1 RT-To further examine how 172K can simultaneously decrease to both NRTI and NNRTI resistances, we performed structural comparisons by aligning the crystal structures of RT 172R and RT 172K RTs in complex with NNRTIs. Interestingly, the structural changes observed among RT complexes that have different amino acids at position 172 follow the same pattern, albeit with small variations, as we observed in the comparisons among the unliganded structures (see above). In all cases, 172K and 172R of the ␣-helix E interact differently with residue 180 of the ␤9-strand, and this change affects in turn interactions between other residues of these structural elements as follows: 161 and 165 of the ␣-helix E, with 182 and 184 of the ␤9-strand (Fig. 7). Specifically, the RT 172R structure in complex with NVP shows that 172R and Gln-161 in ␣-helix E interact with Ile-180 and Gln-182 and indirectly with Met-184 in the ␤9-strand (Fig. 7A). The RT 172R structure in complex with EFV also shows that the 172R side chain contacts Ile-180 and has an additional interaction with Gln-182 through a water molecule (Fig. 7B). The RT 172R complex with EFV also shows that ␣-helix E helps stabilize ␤9-strand by a hydrogen bond between Thr-165 and Gln-182. In contrast, RT 172K in complex with NVP analog U05 has no interactions with Ile-180, Met-184, and Gln-161, following repositioning of surrounding residues (Fig. 7A). Similarly, RT 172K in complex with EFV shows loss of interactions of Ile-180 and Thr-165, but additional interactions are gained through Gln-161 with both Met-184 and Gln-182 (Fig. 7B). This loss of interaction between 172K and Ile-180 and Gln-182 is also observed in the presence of nucleic acid substrate (Fig. 7C). Furthermore, in the reported crystal structure of RT 172R/K103N , the aromatic side chain of Tyr-181 seems to flip almost 90°in the opposite direction and is likely to affect NNRTI susceptibility (Fig. 7D). Taken together, this information suggests that the residues in ␣-helix E may be involved in stabilizing the ␤9-strand, which is a part of the polymerase active site where NRTIs bind and of the NNRTI binding pocket. Changes in the interactions between the residues of ␣-helix E and ␤9-strand may affect the positioning of the YMDD loop, thus affecting not only NRTI and NNRTI binding but also substrate or DNA binding. This is consistent with previous reports on reduction of RT processivity due to changes in the YMDD loop (56,57).

DISCUSSION
Our virological and biochemical data demonstrate that 172K suppresses resistance to both NRTIs and NNRTIs. Moreover, we established that the suppression of NRTI resistance by 172K involves a decrease in ATP-mediated excision. Previous studies have demonstrated that some NRTI and NNRTI resistance mutations can also affect excision-based NRTI resistance. Hence, NRTI resistance mutations K65R, K70E, L74V, and M184V and NNRTI resistance mutations L100I and Y181C block ATP-mediated excision and suppress AZT resistance (32-35, 37, 38). In contrast, other NRTI resistance mutations (V118I and T215Y) impart NNRTI hyper-susceptibility to HIV-1 (58). To our knowledge, 172K is the first polymorphism that suppresses resistance to two independent types of inhibitors, NRTIs and NNRTIs.
Pathak and co-workers (24 -27) proposed that connection subdomain mutations enhance NRTI and NNRTI resistance by reducing RNase H activity, thereby providing additional time for RT to bind in an NRTI excision-competent mode or allow NNRTIs to dissociate from RT (RNase H-dependent resistance mechanism). It is likely that connection subdomain mutations can also cause RNase H-independent AZT resistance. For example, mutation G333D does not reduce RNase H function, but it increases ATP-mediated excision, likely the result of long range interactions (27,59). In addition, N348I confers NRTI and NNRTI resistance and increases ATP-mediated excision of AZT by both RNase H-dependent and independent mechanisms (60). The antagonistic effects of Y181C or M184V on phenotypic AZT resistance cannot be counteracted by N348I (19,61). Interestingly, 172K reduces the resistance (to NRTIs or NNRTIs) that N348I imparts to HIV when added to WT, L74V, L74V/TAM-1, M184V, or M184V/TAM-1 backgrounds (Fig.  1). We showed that 172K can either slightly increase (RT 172K versus RT 172R ) or decrease (RT 172K/TAM-2 versus RT 172R/TAM-2 ) the RNase H activity (supplemental Fig. 2). Hence, 172K does not have a consistent effect on RNase H function and is thus unlikely to reduce NRTI or NNRTI resistance by restoring the RNase H defect introduced by N348I.
How then is 172K reducing NRTI resistance? There are several possible mechanisms by which a residue could enhance susceptibility to NRTIs. First, it could improve NRTI incorporation into DNA. However, this is not the case with 172K because the IC 50 values for AZT inhibition of RT 172R/TAM-2 or RT 172K/TAM-2 are not significantly different under conditions where only incorporation and not excision are examined (in the absence of ATP) ( Table 2 and also supplemental Fig. 1). Second, a residue could increase sensitivity to NRTIs by decreasing excision. This could be accomplished by several ways including the following: (a) by promoting translocation of the NRTI-ter- In addition to changes in the interactions between side chains of ␤-9 -10 sheet and ␣-helix E also observed in A and B, the side chain of Tyr-181 "flips" in the 172K versus the 172R structure.

172K RT Polymorphism Suppresses Resistance to RTIs
minated primer to the post-translocation site (47), which is not accessible to ATP, thus preventing NRTI excision (47). This is also not the case with 172K, as RT 172K favors the pre-translocated mode of binding (Fig. 3), which allows unblocking of chain-terminated primers; (b) by decreasing the excision efficiency through impaired use of the excision substrate ATP. However, our data (not shown) do not support this possibility, as we do not find any significant difference in ATP binding between RT 172R/TAM-2 and RT 172K/TAM-2 or RT 172R and RT 172K (the apparent K m for ATP in a rescue-type assay for all enzymes is ϳ3 mM); (c) by increasing the dissociation of chain-terminated T/P, thereby limiting unblocking. Our data are consistent with this apparently novel mechanism of suppression of NRTI resistance. Specifically, SPR measurements indicate that RT 172K had decreased affinity for nucleic acid due to an increased dissociation rate (Fig. 4) (k off was ϳ31-fold higher in the case of RT 172K ). Hence, our data suggest that RT 172K has decreased AZT resistance primarily because the AZT-terminated template/primer falls off of the enzyme and cannot be unblocked. This is also consistent with the observed decrease in processivity of RT 172K and RT 172K/TAM-2 (Fig. 5). Interestingly, the processivity of RT 172K/Q151Mc was not reduced (Fig. 5). Hence, it is possible that the relatively higher prevalence of RT 172K/Q151Mc (9.9%, supplemental Table 4) compared with RT 172K/TAM-2 is likely due to the higher processivity and presumably higher fitness of RT 172K/Q151Mc .
The structural basis of decreased NNRTI resistance and AZT excision by RT 172K is likely related to how changes at 172 affect the positioning and mobility of the active site YMDD loop during the course of polymerization. The structural analysis of the crystal structures solved here and previously (Figs. 6 and 7) predict some changes in the interactions of the ␤9-strand with the ␣-helix E. The changes are propagated toward the polymerase active site and may affect the relative motions of the YMDD loop during the course of polymerization. Our previously published work has established that the mobility of YMDD is important for NRTI excision, processivity (57), and NNRTI susceptibility (56). Hence, it is possible that 172K affects NRTI and NNRTI resistance by changing the local environment of YMDD.
The R172K change emerges during serial passages of HIV-1 in MT-2 cells in increasing concentrations of Reverset (RVT, D-2Ј,3Ј-didehydro-2Ј,3Ј-dideoxy-5-fluorocytidine), a nucleoside analog with potent activity against HBV and AZT-resistant HIV-1, especially in the presence of TAMs (62,63). Phenotypic analysis revealed that K70N/V90I/R172K had a 3.9-fold increase in RVT resistance (64). Our findings indicate that appearance of 172K during RVT therapy may also affect resistance to other drugs and should be taken into consideration in designing RVT-based therapeutic strategies.
In conclusion, we have identified the first RT polymorphism that suppresses resistance to both NRTIs and NNRTIs. The effect of polymorphisms in suppressing NRTI and NNRTI resistance provides new information into the interactions between the polymerase active site and the NNRTI-binding site of RT. These findings provide valuable insights for the design of antiviral regimens and new RT inhibitors.