Novel mechanism of inhibition of HIV-1 reverse transcriptase by a new non-nucleoside analog, KM-1.

2-Naphthalenesulfonic acid (4-hydroxy-7-[[[[5-hydroxy-6-[(4 cinnamylphenyl)azo]-7-sulfo-2-naphthalenyl]amino]-carbonyl]amino]-3-[(4-cinnamylphenyl)]azo (KM-1)) is a novel non-nucleoside reverse transcriptase inhibitor (NNRTI) that was designed to bind at an unconventional site on human immunodeficiency virus type 1 reverse transcriptase (RT) (Skillman, A. G., Maurer, K. W., Roe, D. C., Stauber, M. J., Eargle, D., Ewing, T. J., Muscate, A., Davioud-Charvet, E., Medaglia, M. V., Fisher, R. J., Arnold, E., Gao, H. Q., Buckheit, R., Boyer, P. L., Hughes, S. H., Kuntz, I. D., and Kenyon, G. L. (2002) Bioorg. Chem. 30, 443-458). We have investigated the mechanism by which KM-1 inhibits wild-type human immunodeficiency virus type 1 RT by using pre-steady state kinetic methods to examine the effect of KM-1 on the parameters governing the single nucleotide incorporation catalyzed by RT. Analysis of the pre-steady-state burst phase of dATP incorporation showed that KM-1 decreased the amplitude of the reaction as previously shown for other NNRTIs, because of the slow equilibration of the inhibitor with RT. In the ternary enzyme-DNA-KM-1 complex (E-DNA-I), incorporation of the next nucleotide onto the primer is blocked. However, unlike conventional NNRTIs, the inhibitory effect was caused primarily by weakening the DNA binding affinity and displacing DNA from the enzyme. Wild-type RT binds a 25/45-mer DNA duplex with an apparent K(d) of 3 nm, which was increased to 400 nm upon saturation with KM-1. Likewise, the apparent K(d) for KM-1 binding to RT increased at higher DNA concentrations. We therefore conclude that KM-1 represents a new class of inhibitor distinct from nevirapine and related NNRTIs. KM-1 can bind to RT in both the absence and presence of DNA but weakens the affinity for DNA 140-fold so that it favors DNA dissociation. The data suggest that KM-1 distorts RT conformation and misaligns DNA at the active site.

Since the introduction of antiretroviral therapy, the life span of patients infected with HIV 1 or suffering from AIDS has been dramatically extended (1). However, the rate of infection is still on the rise (2), the current antiviral drugs do not eliminate HIV infection, and the high rate of mutation of the virus leads to rapid emergence of new variants of HIV resistant to each class of drugs (3)(4)(5)(6)(7)(8). Thus, current treatments call for the frequent screening to look for increases in viral load indicative of new drug-resistant forms of the virus, which must be countered by a new combination of drugs. The current strategy for drug design is to address the changes in HIV by developing new drugs with a different resistance pattern. The purpose of this study is to investigate the properties of a new series of compounds to quantitatively evaluate their potency and to examine their mechanism of inhibition of HIV RT.
RT is responsible for replication of single-stranded viral RNA into double-stranded DNA, which is subsequently integrated into the host cellular DNA during the course of viral infection (9). Two types of anti-HIV medications target RT: the nucleoside analog RT inhibitors and non-nucleoside RT inhibitors (NNRTIs). Nucleoside analog RT inhibitors, such as 3Ј-azido-3Ј-deoxythymidine and 2Ј,3Ј-dideoxycytidine, are nucleosides that become phosphorylated by cellular enzymes to their triphosphate form and are then incorporated by RT, acting as chain terminators to prevent further polymerization. To varying degrees, they also block cellular DNA polymerases, in particular the mitochondrial DNA polymerase (Pol ␥) (10), and their kinetics of incorporation are correlated with the toxic side effects of these drugs (reviewed in Ref. 11). NNRTIs are structurally diverse hydrophobic compounds that show fewer toxic side effects, but they are predominantly effective toward HIV-1 RT and not HIV-2 RT (12). NNRTIs bind in a hydrophobic pocket ϳ10 Å away from the RT polymerase active site in the palm domain of the p66 subunit and distort the key residues that comprise the aspartic triad (D110, D185, and D186) so that the carboxyl groups of the side chains are out of alignment leading to slower rates of catalysis (13)(14)(15). Spence et al. (15) showed that the first generation NNRTIs, nevirapine and tetrahydroimidazo [4,5,1-jk] [1,4]-benzodiazepin-2(1H)-thione, are allosteric, slow-tight binding inhibitors that reduce the rate of the chemical reaction and increase the affinity of nucleotide binding to the inhibitor and DNA-bound RT (E-D-I) complex. As a result of a slower chemical reaction, the two steps of nucleotide binding entailing the initial ground-state binding and the subsequent conformational change reach equilibrium leading to tighter, albeit non-productive nucleotide binding.
The biggest barrier in the battle against AIDS is mutation, which confers resistance toward RT inhibitors. Drug-resistant RT mutants have been proposed to obstruct the binding of the drug (16,17), reposition the template/primer (18 -20), or facil-itate the removal of chain terminators (21). Because of a high frequency of mutation by RT, HIV is incessantly generating new drug-resistant strains (22,23). In patient plasma, drugresistant mutations of HIV-1 RT have been detected in the presence of all RT inhibitors (24), including the most recent NNRTI, efavirenz (25). Because of cross-resistance, the combination drug regimen exacerbated the problematic mutations in HIV RT (26,27). Therefore, new distinct drugs are required to counter the rapid RT evolution.
Herein, we report the kinetic analysis of a new class of NNRTIs selected and designed initially by computational structure-based design ( Fig. 1) (28). Compared with two other derivatives with modified terminal moieties, KM-1 was predicted to be the most potent and the least toxic based upon results from cell culture screening and preliminary activity assays (28). Further investigation into the mechanism of inhibition by KM-1 is essential for the assessment of its potential as a new class of inhibitors of HIV RT. Our data revealed that, unlike other NNRTIs, KM-1 interfered with template-primer binding. We propose that KM-1 binds to RT with or without DNA and precludes proper alignment of DNA at the polymerase active site, depleting the active DNA-bound RT (E-D) complex required for nucleotide incorporation. KM-1 is similar to other NNRTIs in that each seems to slow the rate of catalysis in the E-D-I complex by binding to a site adjacent to the catalytic site. However, it also weakens the binding of DNA, whereas conventional NNRTIs are thought to strengthen DNA binding by slowing the DNA dissociation rate from the closed complex formed after the binding of the next correct nucleotide (15).

EXPERIMENTAL PROCEDURES
Materials-All chemicals were purchased from Sigma Aldrich (St. Louis, MO) unless otherwise noted. Bacterial expression cell lines, dATP, T4 polynucleotide kinase, and kinase buffer were purchased from Invitrogen Corp. [␥-32 P]ATP (3000 Ci/mmol) was purchased from PerkinElmer Life and Analytical Sciences. The non-nucleoside inhibitors were synthesized as described previously (28). The P-11 resin for the phosphocellulose column and filter paper and discs were purchased from Whatman (Chilton, NJ). The Sepharose resin was purchased from Amersham Biosciences. The non-nucleoside inhibitors were dissolved and stored in Me 2 SO at Ϫ20°C and diluted by at least 700-fold to a stock solution with water. Controlled experiments verified that the final concentrations of Me 2 SO did not affect the measured kinetic rates.
Buffers-All experiments were carried out at 37°C. RT was incubated with DNA in a reaction buffer containing 50 mM Tris acetate pH 7.5, 100 mM potassium acetate, 0.1 mM EDTA, and 0.1-1 mg/ml BSA. Nucleotides (such as dATP) were diluted into the same buffer with 4 mM magnesium acetate.
Protein Expression and Purification-Transformed DH5␣ cells producing the p66 and p51 subunits were grown separately in Lennox L broth supplemented with ampicillin (0.1 mg/ml) at 37°C. When the optical density reached 0.6, protein expression was induced by adding 0.5 mM isopropyl ␤-D-thiogalactopyranoside. Cells were then grown for 16 h at 37°C and harvested by centrifugation. The crude cell lysates were analyzed by SDS-PAGE stained with Coomassie Blue. Based upon the gel-staining intensities of the bands corresponding to the p66 and the p51subunits, cell pellets of the large and small subunit were combined in a ratio of 3.9:1 to achieve a ratio of 1.25:1 at the end of the purification process. The procedures for purification of wild-type RT were described previously by Kati et al. (29,30). The concentration of RT was determined by absorbance at 280 nm with an extinction coefficient of 260,450 M Ϫ1 cm Ϫ1 . 〈ctive enzyme concentration was measured by active site titration as described below.
Synthetic Oligonucleotides-All experiments were performed with the same DNA duplex 25/45-bp as described by Kati et al. (29) in which the next base coded for incorporation was dATP. The sequence of the 25-mer primer was GCCTCGCAGCCGTCCAACCAACTCA. The sequence of the 45-mer template was GGACGGCATTGGATCGAGGTT-GAGTTGGTT GGACGGCTGCGAGGC. Oligonucleotides were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA) and purified by denaturing polyacrylamide gel electrophoresis. Concentrations of the oligonucleotides were determined by absorbance at 260 nm. Equimolar amounts of 25-mer and 45-mer were annealed by placing the solution in a heating block at 95°C and then allowing the heating block to gradually cool to room temperature.
5Ј-32 P Labeling of 25-bp Primer-Before annealing, the primer was 5Ј-radiolabeled with [␥-32 P]ATP by T4 polynucleotide kinase. The kinase was denatured after 1 h by placing the reaction at 95°C for 5 min. Un-reacted and contaminating nucleotides were removed from the labeled primer using a Micro Bio-Spin-30 column (Bio-Rad Laboratories, Inc.). The final concentration of labeled 25-mer was determined by thin layer chromatography of the samples taken before and after the Micro-Spin column.
Pre-steady-state Kinetics of Single Nucleotide Incorporation-Experiments were carried out at 37°C in a KinTek RQF-3 rapid quench flow (www.kintek-corp.com). The experiments were initiated by mixing a pre-equilibrated complex of enzyme-DNA, in the absence or presence of inhibitor, with an excess of dATP (15 l each), and quenched with 80 l of 0.5 M EDTA, pH 8.0. The reaction sample was collected into a microcentrifuge tube containing 30 l of polyacrylamide gel loading buffer. We will refer to the concentrations of all reactants after 1-to-1 mixing in the quench-flow instrument.
Product Analysis-We quantified the extension of 5Ј-labeled 25-mer to 26-mer. Products were resolved on a denaturing gel (15% polyacrylamide, 0.8% bis-acrylamide, 7 M urea). The dried gel was exposed to a phosphor-screen then scanned with a Storm 860 PhosphorImager (Amersham Biosciences) and quantified using ImageQuant 5.0 software.
Data Analysis-Data were fitted by nonlinear regression using the program GraFit 5.0.1 (Erithacus Software). Data points from the presteady-state burst experiments were fitted to the burst equation y ϭ A ϫ (1 Ϫ e Ϫkt ) ϩ m ϫ t, where y represents the concentration of the 26-mer product, A is the burst amplitude, k is the observed burst rate, m is the slope of the linear steady-state phase, and t the reaction time. The steady state rate was calculated by dividing the slope by the concentration of active enzyme. The observed burst rates were plotted against nucleotide (dATP) concentration and fitted to a hyperbolic equation, y ϭ (k max ϫ S) / (K d ϩ S), where k max represents the maximum incorporation rate, K d the apparent equilibrium dissociation constant for the nucleotide, and S the concentration of nucleotide. For single nucleotide incorporation experiments done in the presence of increasing concentrations of the inhibitor, the plot of the amplitude of each time course as a function of inhibitor concentration was fitted to a hyperbolic equation, y ϭ E Ϫ [(E ϫ I)/(K d ϩ I)] ϩ B, where y represents the amplitude, E is the total enzyme concentration, I is the total inhibitor concentration, B is the background signal, and K d is the apparent equilibrium dissociation constant for the non-nucleoside inhibitor.
Titration of Active Enzyme with DNA-The concentration of active enzyme and the K d for DNA in the presence of different concentrations of the inhibitor were determined by an active site titration experiment as described previously (29). RT was added to increasing concentrations of 5Ј-32 P-labeled 25/45-mer DNA, in the presence or absence of inhibitor and incubated for 5 min. The sample was then mixed with an equal volume of dATP (100 M) for 0.5 s and then quenched with 0.5 M EDTA, pH 8. The amplitudes were obtained by fitting each time course to the burst equation. The amplitudes were then plotted as a function of DNA concentration. The results from experiments done in the absence of the inhibitor were fitted with a quadratic equation where A is the amplitude, E is the total enzyme concentration, D is the total DNA concentration, K d is the equilibrium dissociation constant for DNA, and F is the binding factor. The binding factor is the fraction of E-D complexes that are productively bound. It corrects for the actual concentration of DNA in productive E-D complexes because the data from DNA titration experiments showed that both inactive and active enzymes were bound to DNA. The maximum amplitude obtained was ϳ50% of the protein concentration determined by the UV absorbance measurements at 280 nm. All enzyme concentrations reported here are calibrated via active site titration. The results from DNA titration experiments done in the presence of the inhibitor were fitted with a hyperbolic equation, y ϭ [(E ϫ I)/(K d ϩ I)] ϩ B, where y represents the amplitude, D is the total DNA concentration, and K d is the equilibrium dissociation constant for the inhibitor.
TK Solver 4.0 (Universal Technical Systems, Inc.) was also used to estimate the equilibrium dissociation constant for KM-1 binding to E-D complex and free enzyme. Numerical estimates were entered to solve for seven equations simultaneously. The K d values calculated by TK Solver were used to simulate data, which were overlaid with the experimental data for validation.
When measuring the apparent binding rate of KM-1, normalized product formation as a function of time was fitted to an exponential decay equation, where [P] normalized ϭ A ϫ e Ϫkobs ϫ t ϩ C. A represents the amplitude, k obs is the apparent binding rate constant, t is the time, and C is the background signal. Association and dissociation rates for formation of the E-D-I complex were obtained by analysis of the KM-1 concentration dependence of the observed rate of binding as in the function, k obs ϭ k on ϫ [I] ϩ k off . An independent estimate of k off was not possible because of the large errors in the extrapolated intercept; therefore, the binding rate was calculated by fitting the data to the equation, k obs ϭ k on ϫ ([I] ϩ K d ), such that k off was constrained to agree with both the K d value and the observed binding rate.
Static Light Scattering-Inhibitor stock solutions at 14 mM in Me 2 SO were diluted with 50 mM potassium phosphate buffer. The buffer was filtered with 0.02 m Anodisc 47. All samples were analyzed with mini-DAWN triple-angle light scattering detector for ambient high performance liquid chromatography from Wyatt Technology Corporation (Santa Barbara, CA). A neon-helium ion laser at 632 nm was used. The laser power and integration times were comparable for all experiments. The detector angle was 90°. Each sample was measured three or more times at room temperature.
Fluorescence Spectroscopy and Fluorescence Polarization Anisotropy-Stock solutions of DNA and KM-1 were diluted in reaction buffer (50 mM Tris acetate, pH 7.5, 100 mM potassium acetate, and 0.1 mM EDTA). KM-1 is naturally fluorescent. The fluorescence intensities were measured with QuantaMaster (QM-1) spectrofluorometer from Photon Technology International (Lawrenceville, NJ). Each sample volume was 100 l. The emission wavelength scan was obtained by exciting at 352 nm and emitting from 365-650 nm. The excitation scan was obtained by examining emission at 440 nm, exciting from 250 -425 nm. The polarization angles for both the excitation and emission linear polarizing filters were set at 0°a nd 90°to obtain four sets of fluorescence intensity data. The signals from horizontal and vertical polarization were respectively denoted as I vv (0°e xcitation; 0°emission), I vh (0°excitation; 90°emission), I hh (90°excitation; 90°emission), and I hv (90°excitation; 0°emission). The anisotropy (r) value was calculated as r ϭ (I vv ϫ I hh Ϫ I hv ϫ I vh )/(I vv ϫ I hh ϩ 2 ϫ I hv ϫ I vh ). The anisotropy value was plotted against DNA concentration to check for signs of interaction (31).
Transmission Electron Microscopy-Solutions were prepared at final concentrations of 20, 50, 250, and 500 M KM-1 in double-distilled water. At room temperature, 5 l of each solution was placed onto a carbon-coated grid from Electron Microscopy Sciences (Fort Washington, PA) for 30 s. The grid was then blotted on filter paper to remove excess solution, washed three times with water, and negatively stained with 2% aqueous uranyl acetate for 10 s, blotted, and air dried. Images were obtained with a transmission electron microscope (Philips TEM 208) at 80 kV. Micrographs were recorded at 44,000ϫ magnification.

Dose-dependent Inhibition of Single Nucleotide
Incorporation by KM-1-We first examined the effect of KM-1 on the presteady state burst of single nucleotide incorporation catalyzed by HIV RT with a defined primer/template. In this experiment, the burst rate provides a direct measurement of the rate of polymerization at the active site and the amplitude defines the concentration of reactive enzyme-DNA (E-D) complexes, whereas steady state turnover is limited by the release of product DNA from enzyme (29,32). We incubated the E-D complex with 0 -20 M inhibitor before mixing with Mg 2ϩ ⅐dATP to start the reaction. The results in Fig. 2A show an initial burst of polymerization followed by a slow steady-state turnover. In the absence of KM-1, the burst amplitude for dATP incorporation catalyzed by 50 nM RT (with 100 nM DNA) was 43 Ϯ 2 nM, the burst rate was 7.4 Ϯ 0.9 s Ϫ1 , and the steady state rate was 0.20 Ϯ 0.03 s Ϫ1 . Increasing the inhibitor concentration reduced the amplitude of the burst phase as well as the rate of the linear steady state phase to almost zero (Figs. 2A and 3A) but did not affect the burst rates. Reduction in the amplitude is presumably caused by the binding of KM-1 to E-D and/or E in a reaction that equilibrates slowly relative to the rate of the burst (15). The reduction in burst amplitude but not rate is consistent with the slow equilibration of the inhibitor so that the observed amplitude reflects only the reaction catalyzed by uninhibited enzyme. Therefore, plotting the amplitude of the burst as function of inhibitor concentration provides a direct titration to quantify inhibitor binding as shown in  Similar effects on the burst amplitude and steady state rates observed with EDC 11 and EDC 12 suggest identical inhibition mechanisms as KM-1. The modified moieties in EDC 11 and 12 ( Fig. 1) affect the binding affinities to HIV-1 RT only slightly. The hyperbolic function of amplitude versus inhibitor concentration defined the apparent equilibrium dissociation constants for EDC-11 and EDC-12 of 1.9 Ϯ 0.4 and 1.5 Ϯ 0.6 M, respectively. Two modes of action were possible for this series of inhibitors. The reduction of burst amplitude with increasing concentrations of the inhibitor suggests that they can act like conventional NNRTIs and inhibit the chemistry step of polymerization pathway by binding to the E-D complex. On the other hand, they could bind to the free enzyme and prevent productive binding of DNA to HIV-1 RT. Their structural resemblance to DNA base pairs suggests possible competition between DNA and the inhibitor for binding to HIV-1 RT.
To examine whether KM-1 acted by competing with DNA binding, we repeated the titration of the reaction amplitude tion dependence of the burst amplitude as described previously (29) at various concentrations of KM-1 (0, 2, 5, or 10 M) (Fig.  4). In each sample, DNA and KM-1 were combined before adding 200 nM RT and allowed to equilibrate for 15-30 min at 37°C before initiating the polymerization reaction. All reactions were initiated by adding Mg 2ϩ ⅐dATP to pre-incubated mixtures of RT and DNA with or without KM-1. The burst amplitude was plotted as a function of DNA concentration at each KM-1 concentration. In the absence of KM-1, DNA binds HIV-1 RT tightly; thus, the titration curve was fitted to a quadratic equation to obtain the K d for DNA binding to uninhibited RT. In the presence of KM-1, the apparent DNA binding affinity was sufficiently weak that the titration curves could be fitted to a hyperbolic equation to obtain the apparent K d for DNA binding at each concentration of KM-1 (Fig. 4A).
To a first approximation, the KM-1 concentration dependence of the apparent K d for DNA binding can be interpreted in terms of a simple model involving competition between the two ligands for binding to the primer/template binding site, KM1 ). The dependence of the apparent K d,DNA on inhibitor concentration (Fig. 4B) was fitted by linear regression to yield the slope of the line defining the ratio of K d,DNA /K d,KM-1 ϭ 0.08 Ϯ 0.02 implying that K d,KM-1 ϭ 35 Ϯ 16 nM for KM-1 binding free enzyme based upon a true K d,DNA value of 3 Ϯ 1 nM.
If KM-1 was directly competitive with DNA such that KM-1 could not bind in the presence of DNA, then high concentrations of DNA should overcome the inhibition. However, as seen in Fig. 4A and displayed in Fig. 4C, the maximum amplitudes of the burst reaction obtained by extrapolation to infinite DNA concentration were reduced in the presence of increasing concentrations of KM-1. These data indicate that KM-1 is still capable of binding to the enzyme, as the DNA concentration is extrapolated toward infinity and suggest that the ternary E-D-I complex is non-reactive. These results suggest that there is no absolute competition between DNA and KM-1 binding to RT in that higher concentrations of DNA cannot overcome inhibition by KM-1. Therefore, we propose that KM-1 and DNA do not share the same binding site in its entirety, although KM-1 and DNA interfere with each other in binding to HIV-1 RT such that the sites may overlap. This model is illustrated in Scheme I.
We can estimate the K d for KM-1 binding to the E-D complex by fitting the KM-1 concentration dependence of the reduction in maximum reaction amplitude at saturating DNA concentration as shown in Fig. 4C. Fitting these data to a hyperbolic function defines the K d ϭ 4.7 Ϯ 1.8 M for KM-1 binding to E-DNA as summarized in Scheme I (rounded off to 5 Ϯ 2 M). Thermodynamic box arguments allow computation of the final equilibrium constant, K d,4 ϭ 0.43 Ϯ 0.32 M, for the binding of DNA to the enzyme-inhibitor complex. Thus, the inhibitor weakens the affinity of DNA for the enzyme by a factor of ϳ140-fold.
Direct competition between KM-1 and DNA would represent one extreme of the model shown in Scheme I, where the ternary E-DNA-I complex was extremely weak or could not form at all. In the present case, all four species in the thermodynamic box contribute to the observed inhibition, and the only equilibrium constant that can be measured directly is that for forming the E-D complex in the absence of inhibitor. All other measurements in the presence of both DNA and inhibitor are complicated by the presence of an equilibrium mixture of E-D, E-I, and E-D-I leading to the observed apparent K d for either KM-1 or DNA.
The KM-1 inhibition pathway is summarized in Scheme I. The software TK solver was used to simultaneously solve the K d of KM-1 dissociation from E-D-I complex, the K d of KM-1 dissociation from E-I complex, and the K d of DNA dissociation from E-D-I complex. According to Scheme I, steps 1-4 in the inhibited reaction pathway were described by four definitions of equilibrium constants and three mass balance equations: and I ϭ I o Ϫ EI Ϫ EDI. E represents the equilibrium enzyme concentration, E o is the initial enzyme concentration, D is the equilibrium DNA concentration, D o is the initial DNA concentration, ED is the concentration of E-D complexes, EDI is the concentration of E-D-I complexes, and EI is the concentration of E-I complexes. K d,1 was determined in this experiment (see above and Fig. 4). E o , D o , and I o were known from the reaction conditions. Estimates were inserted into the equations along with the known values. The solved K d values were used to simulate the data for the DNA titration experiments. The simulations were superimposed with real data to check the accuracy of the calculations.
dATP Concentration Dependence of the Nucleotide Incorporation Rate in the Presence of Different Amounts of KM-1-Next, we examined the kinetics of single nucleotide incorporation (15,29,33). In this experiment, we measured the maximum incorporation rate (k pol ) and the dissociation constant of dATP (K d,dATP ) in the presence of 0 -5 M of KM-1. During the quench flow operation, equal volumes of the preincubated mixture (consisting of RT, DNA, and KM-1) and dATP⅐Mg 2ϩ were rapidly mixed and quenched after various times. Product concentration versus time (Fig. 5A) was fitted to a burst equation to obtain the rate and amplitude of the fast product formation. The burst rate was plotted against dATP concentration (Fig. 5B) and fitted to a hyperbola to obtain k pol (Fig. 5C) and K d,dATP (Fig. 5D). The k pol was approximately constant and the K d,dATP remained in the low micromolar range. These observations are consistent with the conclusion that the observed product formation was catalyzed only by uninhibited E-D complex. SCHEME 1. Pathway of inhibition by KM-1. The inhibitor binds to enzyme in the presence and absence of DNA as shown but weakens DNA by 140-fold.
KM-1 Binding Rates-Our data indicated that the equilibration of inhibitor with the E-D-I complex is slow relative to the rate of polymerization. We measured the rates of KM-1 binding by a rapid quench-flow method consisting of two sequential mixing steps. First, KM-1 was mixed with a solution of E-D complex and allowed to react for 0.1-30 s, then Mg 2ϩ ⅐dATP was added to initiate the reaction. Polymerization was then allowed to take place for 0.2 s, which is sufficient for approximately a single turnover. The reactions were terminated by mixing with EDTA. The product concentration was normalized to the maximum amount of turnover detected in 0.2 s without the inhibitor. The fraction of product formation was plotted as a function of time allowed for the binding of KM-1 to E-D (Fig. 6A). Four KM-1 concentrations were tested: 5, 10, 15, and 20 M. Each curve of the product formation was fitted to a single exponential decay function to calculate the binding rates, k obs , which were in turn linearly dependent on the KM-1 concentration to define the values for the association and dissociation rate constants (Fig. 6B). For reasons that are unclear to us, the observed binding rates at 15 and 20 M KM-1 were ϳ2-fold higher than expected from the linear relationship established at the lower KM-1 concentrations. Because the latter two concentrations are beyond the inhibition concentration range observed in An independent estimate of k off was not possible because of the large errors in the extrapolated intercept; therefore, the binding rate was calculated by linearly fitting the first two data points at KM-1 concentrations of 5 and 10 M and fixing the intercept to zero, such that k off was constrained to agree with both the K d value (ϭ k off /k on ) and the observed binding rate. The linear fit rendered a slope of 0.044 Ϯ 0.009 (dashed line). The estimated on rate was (4.4 Ϯ 0.9) ϫ 10 4 M Ϫ1 s Ϫ1 and the off rate was 0.06 Ϯ 0.01 s Ϫ1 .

Novel Mechanism of Inhibition of HIV-1 Reverse Transcriptase
prior experiments, and because it is possible that KM-1 aggregated at the higher concentrations, we used only the first two concentrations to estimate the binding rate.
Compared with nevirapine, with a k on of 9.9 ϫ 10 4 M Ϫ1 s Ϫ1 and a k off of 0.0019 s Ϫ1 (15), KM-1 displayed a similar k on of (4.4 Ϯ 0.9) ϫ 10 4 M Ϫ1 s Ϫ1 and a 30-fold higher k off of 0.06 Ϯ 0.01 s Ϫ1 . The differences in the on and off rate are not surprising because the apparent K d, KM-1 (1.3 M) is about 100-fold greater than K d,nevirapine (0.019 M). KM-1 displayed binding rates in the same order of magnitude as nevirapine and tetrahydroimidazo[4,5,1-jk][1,4]-benzodiazepin-2(1H)-thione within the range of inhibition concentrations. However, this simple analysis may not be entirely valid for KM-1, because the inhibition seems to be a function of binding to both E and E-D and leads to the dissociation of the E-D complex. Because these experiments were performed by mixing the preformed E-D complex with inhibitor, the time dependence of inhibition reflects the rate of binding to form the E-D-I complex, and it is possible that the DNA dissociates more rapidly after inhibitor binding. Given the complexities in performing this experiment, it was not possible to evaluate whether the time dependence was biphasic, reflecting a two-step process. Nonetheless, these data demonstrate that the binding and equilibration of KM-1 is slow relative to the rate of the burst reaction, thereby justifying the assumption inherent in the analysis of the burst amplitude as a measurement of the concentration of un-inhibited E-D complex, poised for rapid catalysis.
KM-1 Specificity-Recent reports on nonspecific inhibition by drugs via aggregation (34 -37) raised our concerns about the specificity of KM-1. Shortly after Skillman et al. (28) discovered KM-1 founded in part on structure-based design, Shoichet et al. (34 -37) reported promiscuous inhibition by aggregation of small molecules designed by high throughput screening or molecular docking. These ligands typically display three properties: 1) they innately bind to many targets; 2) they react chemically with proteins; and 3) they aggregate in solution at low micromolar concentrations. One of the first-generation NNR-TIs, delavirdine, displayed indiscriminate and aggregationbased inhibition of non-targeted enzymes (37). Because KM-1 was designed based on similar molecules (28) and consists of many aromatic rings, which are stereotypical for aggregating inhibitors, we used three approaches to examine the aggregation of KM-1. First, we examined the structures of KM-1 by electron microscopy at concentrations of 5, 50, 250, and 500 M. In comparison with the electron microscopy images published by McGovern et al. (35), no large aggregated particles were detected in the samples of KM-1 at any concentration. Small particles (ϳ30 nm) were observed only at the highest concentration (data not shown). Second, we examined aggregation by classic light scattering, showing no significant changes in the particle size as a function of concentration. Finally, we tested the sensitivity of the inhibition by KM-1 to nonionic detergent (34) by repeating the DNA-directed single nucleotide incorporation experiments with the addition of 0.1% Triton-X. Prior work had shown that Triton-X reversed nonspecific inhibition by disrupting aggregates so that they could not interact with arbitrary surface residues on an enzyme (34). In our experiments, the presence of Triton-X did not change the inhibition mechanism of KM-1. The burst amplitude and the steady state rate decreased, whereas the burst rate remained the same in response to the increase in KM-1 concentration (Fig. 7). The apparent K d,KM-1 was raised ϳ4-fold, from 1.3 Ϯ 0.2 to 5.2 Ϯ 1.4 M. A similar effect was observed when we repeated the experiment with an RNA template of the same sequence: the apparent K d,KM-1 was increased from 3.1 Ϯ 1.3 to 13 Ϯ 3 M (data not shown). Triton-X weakened the binding of KM-1 to RT al-though it did not eliminate the inhibitory effect of KM-1. These data argue against a model in which the aggregation of KM-1 was responsible for the observed inhibition. Moreover, the observed competition by DNA implies that KM-1 is interacting with the primer/template binding site.
Aside from the likelihood of aggregation, the aromatic structure of KM-1 also allows for possible intercalation with DNA. KM-1 resembles chemotherapeutic drugs such as daunomycin and nogalamycin, which inhibit cell replication by inserting themselves between two planar or stacked aromatic rings, such as adjacent bases in a genetic sequence. The interaction results in a distortion of DNA helix structure. We scanned the emission spectra of three mixtures of 10 M KM-1 with 100, 150, and 200 nM DNA, respectively, by fluorescence spectroscopy, but we observed no change in the fluorescence intensity. We also examined changes in fluorescence anisotropy of 2 M KM-1 when titrated with a high concentration of linearized DNA plasmid pUC18 (data not shown). The data fit an apparent K d of ϳ400 nM plasmid. Given a theoretical binding site size of 20 bp for the 2686-bp plasmid, this translates to a K d ϳ 50 M per site. Although the site size is not known, these data indicate the binding of KM-1 to DNA is very weak and do not support a model in which KM-1 intercalates with DNA to inhibit polymerization by HIV-1 RT.
KM-1 Also Inhibits the Mitochondrial DNA Polymerase-Further testing of KM-1 selectivity was carried out by inspecting its effect on the human mitochondrial DNA polymerase, Pol ␥. Pol ␥ (200 nM large subunit was combined with 0.5 M small subunit), 5Ј-32 P-labeled 25/45-mer DNA (300 nM), and KM-1 (0 -30 M) were incubated before reaction with Mg 2ϩ ⅐dATP. Individual curves of primer extension at each concentration of KM-1 were fitted to the burst equation to determine the amplitude (Fig. 8A). The hyperbolic function of amplitude versus KM-1 concentration provided the apparent dissociation constant of 2.2 Ϯ 0.7 M (Fig. 8B). As in the polymerization catalyzed by RT, the amplitude of the burst of polymerization catalyzed by Pol ␥ was reduced by KM-1. Although KM-1 did not inhibit prokaryotic DNA polymerases in the study reported by Skillman et al. (28), the data presented here demonstrates that KM-1 effectively inhibits Pol ␥ with an apparent K d similar to that seen for RT. DISCUSSION We have characterized a novel non-nucleoside inhibitor, KM-1, by examining its effect on the pre-steady state kinetics of single nucleotide polymerization catalyzed by HIV RT. The measurements afforded estimates of the equilibrium dissociation constants for KM-1 binding to RT in the presence and absence of DNA. We observed mutual dose dependence of dissociation constants but not an absolute competition between KM-1 and DNA in binding to RT. Unlike inhibition by nevirapine, there was no evidence of a slow but measurable rate of reaction in the ternary E-D-I complex (15). The data reveal that the major source of inhibition is displacement of the DNA from the enzyme. The ternary E-D-I complex can form but is unable to catalyze incorporation. Taken together, these results suggest that the KM-1 binding site overlaps with the primer/template binding site and KM-1 displaces DNA so that it cannot access the polymerase active site. The ternary complex forms only weakly. The binding of DNA to KM-1-bound RT is 140-fold weaker than to the free enzyme, as summarized in Scheme I.
The amplitude of the burst phase represents the amount of uninhibited E-D complex available for fast nucleotide incorporation in a single turnover. Therefore, the reduction in the burst amplitude in the presence of the inhibitor is caused by a reduction in the concentration of the reactive E-D complex as a result of formation of E-D-I and E-I. The reduced burst amplitude and unchanged burst rate suggest that the rate of equilibration of KM-1 with RT (dissociation and rebinding) is slower than the rate of polymerization catalyzed by uninhibited E-D complex in a single turnover. The unchanged rate of polymerization and constant K d for dATP indicate that the observed incorporation is catalyzed by the uninhibited E-D complex.
Previously characterized non-nucleoside inhibitors form a ternary E-D-I complex that is still capable of nucleotide incorporation, but at a reduced rate in that there is a slow but significant continuing reaction even at saturating concentrations of tetrahydroimidazo[4,5,1-jk][1,4]-benzodiazepin-2(1H)thione and nevirapine (15). KM-1, however, inhibits the polymerization catalyzed by RT to a greater extent because it abolishes the reaction seen during the slower phase. This suggests that the E-D-I complex is not reactive or is at least extremely slow relative to the rate seen in the presence of nevirapine. Because the only observed reaction was caused by the activity of the uninhibited E-D complex, we cannot ascertain whether KM-1 blocks the nucleotide incorporation at the binding step, conformational change, or chemistry. In the case of previous analysis of inhibition by nevirapine, analysis of the incorporation kinetics of the fully inhibited enzyme allowed us to establish that NNRTIs block chemistry without interfering with nucleotide binding or the conformational change step (15). A similar analysis is not possible with KM-1 because no reaction is seen that is not attributable to the uninhibited fraction of the enzyme.
In contrast to nevirapine and tetrahydroimidazo[4,5,1jk] [1,4]-benzodiazepin-2(1H)-thione, KM-1 leads to dissociation of DNA from the E-D-I to favor the E-I complex. Although the E-D-I complex may form, as suggested by the observation that saturating DNA cannot completely overcome the inhibition, the E-D-I complex seems to be non-reactive. It is reasonable to suppose that the binding sites for DNA and KM-1 overlap substantially but that the binding site for DNA is much larger. At high concentrations, KM-1 and DNA both may bind, but the DNA may be dislodged from the catalytic site in such a way as to prevent nucleotide incorporation. Increasing DNA concentration cannot overcome the inhibition because the DNA site is already occupied. We conclude that KM-1 binds to RT at a different site than that for the incoming nucleotide or conventional NNRTIs.
A clear picture of KM-1 selectivity awaits further experiments. KM-1 seemed to inhibit human mitochondrial DNA polymerase as effectively as HIV-1 RT. However, broader cellular toxicity assays are necessary because it is not known whether KM-1 will effectively inhibit the replication of DNA in cells. Our studies test only whether KM-1 can inhibit the mitochondrial polymerase in vitro and do not address questions regarding cellular and mitochondrial uptake. Moreover, it is possible that structure/activity studies based upon KM-1 as a lead compound could succeed in altering the relative affinities for HIV RT versus other cellular polymerases to overcome undesirable side reactions. This study and our prior work (15) outline a series of experiments that can be used to define the mechanism of inhibition by non-nucleoside inhibitors and quantify the affinity.
Like nevirapine, KM-1 binds slowly to the E-D complex, with an apparent second-order rate constant of 4 ϫ 10 4 M Ϫ1 s Ϫ1 and a dissociation rate of 0.04 s Ϫ1 . The rate of dissociation of the DNA from the E-D-I complex was not resolved in these studies. Parniak has described another NNRTI that he claims is unique in that it is a slow-tight binding inhibitor, unlike other NNRTIs (38). However, his claim of uniqueness is without merit, in that all NNRTIs characterized thus far are slow-binding inhibitors with binding rates on the order of 10 4 M Ϫ1 s Ϫ1 , and his rough estimates, using slow-onset inhibition kinetics produce similar rates. All data support the conclusion that Parniak's inhibitor is simply another NNRTI binding to the well established and exploited conventional NNRTI binding site.
It is likely that KM-1 interferes with the proper alignment of DNA at the polymerase active site with its bulky shape. The characterization of this non-nucleoside inhibitor may help in the design of more effective drugs that are potent toward wild type and drug-resistant strains of RT. Given the high affinity of KM-1 for the free enzyme, it is possible that further structure/ activity studies on compounds related to KM-1 could lead to a more potent and selective inhibitor of a new class.