|Home | About | Journals | Submit | Contact Us | Français|
S-adenosylmethionine (AdoMet) lies at an intersection of nucleotide and amino acid metabolism, and performs a multitude of metabolic functions. AdoMet formation is catalyzed by S-adenosylmethionine synthetase (ATP : L-methionine S-adenosyltransferase (MAT)) which is a target for development of anti-cancer and antimicrobial agents. High affinity MAT inhibitors have been found through computational docking of more than 200,000 compounds for predicted binding to the crystallographically-defined nucleotide binding region of the enzyme's active site. Two of the top scoring candidate compounds had IC50 values less than 10 nM,, more than 10,000-fold lower than the substrates' KM values. The compounds are structurally unrelated to the natural ligands of the enzyme. The enzyme is protected from inhibition by ATP, but not by methionine, consistent with binding at the adenosyl region of the active site. These results validate in silico screening as a robust approach to the discovery of inhibitors of this chemotherapeutically relevant enzyme.
S-adenosylmethionine (AdoMet) has a central role in the life of all cells, joining the pivotal processes of nucleotide and amino acid metabolism.1-4 AdoMet acts as methyl donor to DNA, RNA, proteins and small molecules, as a propylamine donor in the synthesis of the polyamines, as a 5′-deoxy-adenosyl free radical progenitor in a diverse family of “Radical SAM” enzymatic reactions, and has other specialized roles such as participation in the syntheses of hypermodified nucleotides in tRNA and the unusual amino acid hypusine. The roles of DNA and protein methylation in the epigenetic regulation of mammalian cell growth and differentiation have closely tied AdoMet to carcinogenesis.5, 6
S-adenosylmethionine biosynthesis involves the condensation of ATP and L-methionine which is catalyzed by S-adenosylmethionine synthetase (ATP: L-methionine S-adenosyltransferase, MAT).7 Alterations in the abundance of the two MAT isozymes in hepatic diseases have led to an enduring search for specific inhibitors of potential use in cell biological studies and as potential chemotherapeutic agents.8-15. More recently MAT has been proposed as a target for the development of anti-parasitic agents.16 However MAT's substrates, ATP and methionine, partake in numerous metabolic processes, thus generating specific, bioavailable, inhibitors as compounds that are structurally related to the substrates remains challenging. Furthermore MATs from eukarya and bacteria are generally highly conserved in sequence and structure, particularly in residues comprising the active site, further suggesting a difficult inhibitor design problem.17 Nevertheless numerous analogs of the substrate methionine have been synthesized as potential MAT inhibitors.8-14, 18-24 Some of the methionine analogs, particularly L-cis-AMB (L-2-amino-4-methoxy-cis-but-3-enoic acid10) and the commercially available cycloleucine (1-amino, 1-carboxy-cyclopentane), have had biological applications because of their ability to enter cells, despite their at best micromolar affinity for MAT and their unclear effects on other amino acid utilizing enzymes.8, 12, 13 Bi-substrate analogs that link portions of methionine to the 5′ position of the nucleotide substrate have been prepared in complex syntheses;14, 18-25 these compounds had Ki values in the low micromolar range and displayed some selectivity between rat MAT isozymes, however these scarce inhibitors have had limited application. This laboratory has described the potent, mechanism-based, inhibition of MAT by diimidotriphosphate (O3P-NH-PO2-NH-PO3; PNPNP) a non-hydrolyzable analog of the tripolyphosphate reaction intermediate with a 2 nM dissociation constant for the E. coli MAT.15 However PNPNP had no effects on the growth of Escherichia coli, Saccharomyces cerevisiae or the NCI 60-cell line panel, likely due to its highly charged nature, formally −5 at neutral pH (unpublished results).
The present work describes the identification of the first MAT inhibitors that are structurally unrelated to the reactants, intermediate or products of the enzyme. In a new approach for this enzyme, we have taken advantage of the availability of the crystal structure of the MAT from E. coli,26, 27 and have used computational docking to screen libraries of commercially available compounds for predicted binding to the enzyme's active site. Experimental tests of a limited selection of candidate compounds verified two potent inhibitors of the enzyme with IC50 values near 10−9 M, more than 10,000-fold lower than the KM for either substrate.28
Virtual screening of ca. 200,000 compounds as potential ligands to the active site of MAT from E. coli, followed by visual inspection of the docked poses of ~200 of the best scoring inhibitors, has led to the identification of two potent inhibitors (1 and 2, each illustrated schematically and in their predicted binding conformation in Figure 1). Eight compounds were selected for experimental testing based on both their predicted affinity (i.e. their Glide Scores) and the predicted structures of their enzyme-bound complexes (Figure 2), however only five of them were available. The method of selection relied upon the prediction of hydrogen bonding to key active site ligands (see below) and interaction with the sole aromatic active site residue, phenylalanine-230. Two compounds provided complete inhibition of MAT activity at a concentration of 1 mM in initial tests, and were then characterized in detail (see Methods). The other compounds did not cause detectable inhibition at 1 mM concentrations and were not further investigated. The binding poses predicted by Glide for 1 and 2, which we propose to be denoted AntiMAT-1 and AntiMAT-2 respectively, are illustrated in Figure 2. The poses predict hydrogen bonding to Histidine-14, Lysine-265, and Aspartate-271 which are key residues in binding and catalysis, and both compounds have a nitrophenyl moiety which is predicted to stack with Phenylalanine-230 upon which the adenine ring of the substrate normally lies.29
Measurements of enzyme activity at various inhibitor concentrations revealed that at the micromolar enzyme concentrations used for activity measurements there was complete inhibition by approximately equimolar concentrations of either 1 or 2 and the protein (Figure 3). The dependence of residual enzyme activity on ligand concentration was fit to the equation for tight-binding inhibition (i.e. one where the Ki value is comparable to the protein and inhibitor concentrations used in the experiments, see Methods). These analyses (Figures 3A and 3B) reported Ki values of 1.9 ± 0.8 nM and 6.9 ± 2.1 nM for 1 and 2, respectively. Figures 4A and 4B show that when 10 mM ATP was included during an incubation of enzyme and ligand, and aliquots were removed for activity determination by addition of methionine, no inhibition was observed with either compound. In contrast when 1 mM methionine (the highest concentration compatible with the activity assay) was included in the incubation of enzyme and ligand (before addition of ATP to start the assay) the same degree of inhibition was observed as in the absence of methionine.
The inhibition by both 1 and 2 has a slow onset, consistent with the high affinities and slow dissociation rates (see below). At 200 μM 1 and 10 μM enzyme, half-maximal maximal inhibition was achieved in ca. 7 min., corresponding to a rate of 1.6 × 10−3 s−1, see Figure 4C. A similar experiment with 2 revealed a half time of 20 min. A limit on the dissociation rate of the Enzyme•AntiMAT complexes was estimated from the lack of recovery of detectable activity after a 24 hr dialysis of a mixture of enzyme and either inhibitor against 1250 volumes of buffer (enzyme and inhibitor were initially present at 0.25 μM and 0.3 μM, respectively; if equilibrated, the final concentration of inhibitor would have been 0.25 nM). An estimate of the dissociation rate was obtained from the limit of detection of 10% recovery of the enzyme activity, from which the off rate was estimated as < 3 × 10−6 s−1. (at 4° C where the dialysis was conducted). For 1, a dissociation rate of ~10−6 s−1 can be estimated from the 1.6 × 10−3 s−1 rate of inhibition onset and the 1.9 nM Ki value. An analogous analysis indicates a similar dissociation rate for 2.
1 and 2 were tested for potential inhibition of the MAT homolog from the archaeon M. jannaschii, which has less than 20% sequence identity to the E. coli MAT and for which protein circular dichroism spectra indicate a structure that is likely to be different (a crystal structure is not available for this protein or a close relative).30 At concentrations up to 1 mM neither compound inhibited the archaeal M. jannaschii enzyme, even after a 20 min preincubation with the protein before activity measurement (see Methods).
We examined whether either of 1 or 2 inhibited the growth of either Saccharomyces cerevisiae or E. coli on solid media by placing a filter disk that was soaked with either 1 mM candidate compound dissolved in DMSO, or a DMSO control, on a freshly spread lawn of cells. No inhibition was observed in any case. Since labeled compounds were not available, we could not determine whether they were taken up by the cells and thus could become available to the cytosolic MAT. In addition both 1 and 2 were evaluated by the Developmental Therapeutics Program of the National Cancer Institute for potential inhibition of the growth of their panel of 60 mammalian tumor cell lines (http://dtp.nci.nih.gov/branches/btb/ivclsp.html);31, 32 concentrations up to 100 μM were tested. Neither compound caused detectable growth inhibition of any of the cell lines; as in the case of the trials with E. coli and yeast, it was not possible to determine whether the compounds were transported into the cells which would be required for MAT inhibition.
The present report describes success in finding novel types of MAT inhibitors by an initial virtual screen, docking several hundred thousand compounds to the active site of cMAT, followed by experimental evaluation of a restricted number (5) of candidates which were selected by visual inspection of the poses of the best scoring compounds. The two new inhibitors are structurally distinct from the natural ligands and may provide useful leads to novel bioavailable inhibitors. The success rate of this screen was atypically good; for example we recently reported the discovery of one compound that stabilizes a minor hexameric assembly of the predominantly octameric porphobilinogen synthase upon experimental testing of 76 candidates identified by using Glide to screen ca. 60,000 compounds from a different library of candidate ligands for potential binding to a postulated inter-subunit interfacial binding site;33 in that case the success rate may be abnormally low because the site chosen was an adventitious location formed upon hexamer assembly rather than an evolved ligand binding site. The reason for our current high success rate may also be related to the small and highly polar nature of the enzyme's active site, which is lined with 4 cationic residues and 3 anionic residues and to which 2 Mg2+ ions and a K+ ion bind as well as the substrates.26, 27 These ionic protein residues may select for the binding of compounds that can have several directional interactions (e.g. hydrogen bonds) that are complimented by stacking with Phenylalanine-230, the sole aromatic residue at the active site. It should be noted that an experimental search of a library of this magnitude for MAT inhibitors could not have been easily performed using the standard assay methods, not only due to the cost of assembling the library of candidate ligands, but also the precautions in handling the radioactive materials, the cost of 14C labeled substrates, and the disposal of radioactive waste.
The two MAT inhibitors described herein did not inhibit cell growth in the systems tested. Nevertheless these inhibitors may provide scaffolds upon which to develop compounds with more desirable properties. Most importantly their discovery validates docking as a satisfactory method for the discovery of MAT ligands with unexpected structures.
The libraries of structures of compounds that were to be screened for candidate inhibitors were obtained in SDF or Mol2 format from the Sigma-Aldrich Rare Chemical Library, Fluorous Technologies, Lancaster Chemicals, Maybridge Chemicals, and Peakdale Chemicals. These docking libraries were largely prepared before the announcement of the ZINC compendium;34 while ZINC and our collection overlap, the differences in the processing of the structures to make them “dockable” will make the runtime libraries vary in the diverse structures that are present, even when the same 2-dimensional representations were initially used.
Modeling software from Schrödinger L.L.C. (New York, NY) was used throughout. The structures were obtained as two-dimensional representations typically without hydrogens and without designations of stereochemical configuration at chiral centers. Perl scripts were used to add appropriate hydrogen atoms to each compound, and the resultant structures were converted to a geometry-optimized, three-dimensional representation using the program MacroModel with the Merck Molecular Force field (MMFF) and the GB/SA continuum solvation model for water.35, 36 Compounds that did not successfully energy minimize, due for example to atoms that were not represented in MMFF (e.g. metal ions), were of necessity discarded. This procedure generated a single representation of each starting structure. The structures were then expanded to include all ionization states predicted to be present at neutral pH using the Ionizer utility from Schrödinger, followed by repetition of the geometry optimization. The resulting combined library had ~200,000 structures.
The MAT protein structure was taken from the coordinates of the enzyme•ADP•Pi complex deposited in the Protein Data Bank (PDB code 1mxb, 2.80 Å resolution). 26 Since six amino acid residues that form an active site lid are disordered in this structure, the low energy conformer from our previous loop modeling work was used;37 these lid residues do not directly contact the substrates (or potential inhibitors) and do not intrude into the active site, unlike their positions in the crystal structure of the apo enzyme.26 All ligands and metal ions were removed from the protein structure before further manipulation. The protein structure had been prepared to have side chains in their appropriate ionization states at neutral pH and had been energy minimized using the AMBER force field in the course of loop modeling.37 In order to enhance the conformity of the protein with the OPLS-AA force field used for docking by Glide,38 the protein structure was subjected to a restrained energy minimization using the OPLS-AA(2001) force field39 the crystallographically defined heavy atoms were allowed to deviate from up to 0.3 from their experimental positions without energy penalty, beyond which a 100 kJ/mol-Å2 penalty was applied, as in our previous work.37
The docking protocol used the program Glide (version 2.0) from Schrödinger, L.L.C, (New York, NY, 2005).38, 40 A 16 Å3 box centered on the ADP ligand was chosen as the region which the ligand center must occupy, and a larger 30 Å3 box was used to define the region in which the entire docked ligand must lie. Glide allows the ligand to have fully flexible torsional geometry, but retains the bond lengths and angles of the input structure; hence it was important to have optimized the geometries of the ligand structures before docking. A ‘soft’ protein surface in which the van der Waals radii of the polar protein atoms were reduced by a factor of 0.8 was used to simulate protein malleability. The van der Waals radii of the ligands were also reduced by a factor of 0.8 for a similar reason. Several crystal structures of E. coli MAT have shown that the ATP binding site accepts differing nucleotide conformations while enzymatic activity retains substantial specificity for the adenosine moiety; for example, a comparison of the positions of the nucleoside moiety in the complex with ADP and Pi (originating from hydrolysis of ATP during the crystallization, pdb code 1m×b)26 with that observed in a complex with both substrate sites occupied (by the ATP analog AMPPNP and methionine, pdb code 1p9l)29 shows that while the phosphoryl groups occupy analogous locations the adenosine moiety swaps position between alternate sites located near either end of the phosphate chain, depending in part on the presence of the methionine co-substrate).29
The output of Glide docking is a list of compounds ranked by “Glide Score” which is a modified version of Chemscore,41 with units of kcal/mol. In order to estimate the Glide Scores that might be expected for potent inhibitors, and to evaluate the reliability of ligand positioning, we first docked the product AdoMet which forms a complex for which the crystal structure has been determined (pdb code 1rg9).26 The 3 μM dissociation constant of AdoMet is the lowest of any compound known to bind without complex kinetics.7 Glide positioned AdoMet with a heavy atom rmsd of 1.1 Å from the crystallographically determined location,29 indicating reliable docking. We choose a cutoff for further evaluation of the candidate ligands as those which docked with poses (poses being defined as the combination of ligand conformation and position) having Glide Scores at least 2 kcal/mol lower (predicted higher affinity) than the Glide Score of AdoMet (−6.5 kcal/mol). Only 10% of the library docked with Glide Scores better than AdoMet, and only 0.08% (170 structures) docked with Glide Scores less than −8.5 kcal/mol. The docked poses of the compounds passing this criterion were then visually examined, and a small selection of compounds which were predicted to cover a significant portion of the active site were chosen for experimental evaluation.
Reagents were acquired from Sigma-Aldrich except where noted. MAT from E. coli (cMAT) was isolated from an overproducing strain as described previously.15 cMAT concentrations were based on the UV absorbance at 280 nm using an extinction coefficient of 1.3 (mg/ml-cm)−1 and a subunit molecular mass of 41941 Daltons. Purified recombinant MAT from the archaeon Methanococcus jannaschii was obtained from Dr. Z. Lu of this laboratory.30 The concentration of MJ-MAT was determined from UV absorbance at 280 nm using an extinction coefficient of 1.0 (mg/ml-cm)−1 and a subunit molecular mass of 45252 Daltons.30
A group of compounds that were identified from the docking screen as possible inhibitors were purchased from the Sigma-Aldrich Rare Chemical Library and their identifying catalog numbers are given along with their chemical names and Chemical Abstracte Service (CAS) Registry Number. These were: 1, S903566, (1-(3-(2-ethoxyphenyl)ureidoacetyl)-4-(2-methyl-5-nitrophenyl)semicarbazide, CAS Registry Number 198704-90-4); 2, S867349, (1-(4-chloro-2-nitrophenyl)-3-(4-sulfamoylphenyl)urea), CAS Registry Number 197160-37-5; 3, S702633, 1,1′-((2-chlorobenzyl)-2-methylphenylene)bis(3-antipyrinylurea), CAS Registry Number 883838-93-5; 4, S720844, 1-(4-methyl-2-nitrophenyl)-3-(4-sulfamoylphenyl)urea, CAS Registry Number 200347-89-3; and 5, S890901, 1,1′-(4-Methyl-1,3-Phenylene)Bis-(3-(3-(Trifluoromethyl)phenyl)urea), CAS Registry Number 200416-79-1.
MAT activity determinations measured incorporation of L-[methyl-14C]-methionine (Moravek Biochemicals) into AdoMet, which was absorbed onto Whatman P-81 cation exchange filter paper and quantified by scintillation counting of the filters after extensive washing with water.15 Routine activity measurements utilized 50 mM Tris•Cl, 50 mM KCl, 10 mM MgCl2, 10 mM ATP, 0.5 mM L-[methyl-14C]-methionine, pH 8.0. Typically reactions were allowed to proceed for 5 min at 25° C before quenching by the addition of 3 volumes of a 25 mM EDTA (pH 8.0) stop solution.
Candidate inhibitors were dissolved in 100% dimethyl sulfoxide (the final amount of DMSO in the assays was maintained at 15% (by volume) unless otherwise noted). Control experiments showed that 15% DMSO did not affect the enzyme activity, either when included with the enzyme for up to an hour before activity measurement, or when it was present during the assay period.
Initial inhibitor screening assays contained reduced substrate concentrations of 0.1 mM ATP and 0.1 mM L-[methyl-14C]-methionine; these substrate concentrations approximate their KM values (0.11 mM for ATP and 0.08 mM for L-methionine).42 Assays were conducted at 25° C, typically for a 5 - 10 min incubation period after initiation by addition of enzyme; assays were terminated by addition of an excess of 25 mM EDTA, pH 8.0. Product formation was linear with time for all assay protocols used. The M. jannaschii MAT was assayed using the same method but reactions were conducted at 37° C due to the low activity of this thermophilic enzyme at 25° C; the KM values of M. jannaschii MAT were 0.29 mM for ATP and 0.24 mM for L-methionine.30
Since we had previously observed a slow onset of MAT inhibition by intermediate and product analogs,15 possible inhibition was evaluated after a 20 min preincubation of the candidate inhibitor at either 0.2 mM or 1 mM with enzyme (0.25 μM) before addition of 1.5 volumes of concentrated assay reagents (final concentrations as above). Experiments were also performed in which the reactions were initiated by addition of enzyme to a solution containing substrates and candidate inhibitor to yield the same final concentrations.
In typical inhibitor screening, cMAT was preincubated at room temperature for 20 min. in 50 mM Tris•Cl pH 8.0, 50 mM KCl, 20 mM MgCl2 with the indicated amount of potential inhibitor or a 15% DMSO control. The protein plus inhibitor solution was then mixed with 1.5 volumes of the assay reagents (final concentrations 50 mM Tris•Cl, pH 8.0, 50 mM KCl, 10 mM MgCl2, 0.1 mM ATP, 0.1 mM [Methyl-14C]-L-methionine) and the reaction was allowed to proceed for 5 min. at 25° C before quenching with 25 mM EDTA, pH 8.0, and followed by determination of the amount of AdoMet formed.
Kinetic data were fit to the equation for tight binding inhibition
where vobserved is the fractional activity observed at a given concentration of enzyme and inhibitor, and [E] and [I] are the total concentrations of enzyme and inhibitor, respectively. Ki is the inhibition constant which was determined by non-linear regression using SigmaPlot v. 10. (SPSS Inc.).
Both 1 and 2 were tested as inhibitors of MJ-MAT (0.3 μM) at co ncentrations from 0.1 to 1.0 mM; enzyme activity was measured after a 20 min preincubation of enzyme plus inhibitor, as described above. No alteration in the enzyme activity was observed.
The reversibility of the inhibition by 1 and 2 was investigated by determining the enzyme activity at various times after dialysis of an enzyme plus inhibitor mixture. A solution containing inhibitor (0.3 μM) and enzyme (0.25 μM) in 50 mM Tris•Cl, pH 8.0, 10 mM MgCl2, 10 mM KCl was initially incubated at room temperature for 20 min. A portion of this inhibited mixture was then dialyzed against 1250 volumes of 50 mM Tris•Cl pH 8.0, 10mM MgCl2, 10 mM KCl at 4° C. Aliquots were removed after 5, 10, 20, 30, 40, 50 60, 70, and 80 min. and 24 hours, added to assay concoction, and the enzyme activity quantified in the usual fashion. A control in the absence of inhibitors showed no loss in enzyme activity through the time period of the experiment.
The structures and purity of compounds identified as MAT inhibitors were verified by LC-MS. Analyses were conducted on a Waters LC-MS system configured with a DeltaPak C-18 column (3.9 × 300 mm; 15 micron; 100 ) using a Model 2545 binary gradient pump with the following series of detectors: a Model 3100 mass spectrometer (in positive electrospray ionization mode), a Model 2424 ELS (Evaporative Light Scattering) detector, and a Model 2487 UV detector (254 nm). Elution used a 20 ml gradient from a mixture of 95% water, 5% methanol, 0.05% TFA to a final solution of 95% methanol, 5% water, 0.05% TFA at a flow rate of 0.8 ml per minute.
Reversed phase HPLC chromatograms of 1, S9035666,,showed the same lone peak eluting at 16.28 min in each detection mode. The purity was estimated as ≥ 97% based on the integrated peak areas. The mass was reported as 430.2 (predicted 430.4; C19H22N6O6). UV/visible λmax = 342 nm in DMSO. 1H NMR (600 MHz, D6-DMSO ): 10.04 (s, 1H); 8.61 (s, 1H); 8.31 (s, 1H); 8.06 (s, 1H); 8.03 (m, 1H); 7.84 (m, 1H); 7.46 (d, J = 9.7 Hz, 1H); 7.42 (m, 1H); 6.93 (d, J=7.4 Hz, 1H); 6.84 (td, 7.4 Hz, 1.7 Hz, 1H); 6.78 (m, 1H); 4.07 (q, J=7.1 Hz, 2H); 3.84 (d, J = 5.4 Hz, 2H); 2.33 (s, 3H); 1.39 (t, 7.1 Hz, 3H).
Reversed phase HPLC chromatograms of 2, S867349, ((1-(4-chloro-2-nitrophenyl)-3-(4-sulfamoylphenyl)urea), showed the same peak eluting at 14.15 min in each detection mode. The purity was estimated as ≥ 98% based on the integrated peak areas. The peak had a mass of 370.1 (predicted 370.7; C13H11ClN4O5S). UV/visible λmax = 380 nm in DMSO. 1H NMR (600 MHz, D6-DMSO): 10.25 (s, 1H); 9.68 (s, 1H); 8.28 (d, J= 8.8 Hz, 1H); 8.15 (d, J = 2.0 Hz, 1H); 7.80 (dd, J = 2.0 Hz, 8.8 Hz, 1H); 7.75 (d, J = 8.8 Hz, 2H); 7.63 (d, J = 8.8 Hz, 2H); 7.25 (s, 2H).
1 and 2 were tested for their ability to inhibit the growth of Escherichia coli strain RP523 (thr-1, leuB6(Am), fhuA21, lacY1, hemB220, glnV44(AS), e14−, rfbC1, glpR200(glpc), thi-1)43 which was obtained from Dr. Eileen Jaffe of the Fox Chase Cancer Center. RP523 is hyperpermeable to exogenous compounds, for example allowing entry of heme and conferring enhanced sensitivity to growth inhibition by antibiotics. Ca. 30 μl of an overnight cell culture were spread on a LB agar plate, and 5 mm diameter filter paper disks soaked with 30 μl of a 1 mM solution of MAT inhibitor or a DMSO control were placed on top. The size of the zone of growth inhibition was measured after an overnight incubation at 37° C.
Potential growth inhibition of Saccharomyces cerevisiae strain RDY84 (MATa, pdr1ΔKAN, pdr3ΔHIS5+, ade2, trp1, his3, leu2, ura3, can1) was investigated in an analogous fashion, using a growth temperature of 30° C on YDP agar plates. RDY84 was obtained from Dr. Warren Kruger of the Fox Chase Cancer Center.
The potential effects of 1 and 2 on the growth of the NCI-60 cell line panel were evaluated at the Division of Chemical Therapeutics of the National Cancer Institute.32 Concentrations up to 100 μM were tested in their standard protocol.
This work was supported by National Institutes of Health Grants GM31186, CA06927, and also by an appropriation from the Commonwealth of Pennsylvania. The work was utilized the Fox Chase High Performance Computing Facility, the Organic Synthesis Facility and Research Secretarial Services. Helpful discussions regarding Glide with technical support persons at Schrodinger L.L.C. are gratefully acknowledged. We appreciate the help of Dr. Warren Kruger who contributed the yeast strain and provided advice on yeast growth, Dr. Trevor Selwood who aided in analysis of the tight binding kinetic data, and Dr. Eileen Jaffe for invaluable advice on many aspects of this study. We thank Dr. Hong Cheng for obtaining the NMR spectra and Cindy Myers for the HPLC and mass spectral analyses. We appreciate the help of Dr. Maxim Pimkin in preparing Figure 2. We thank the NCI Division of Chemical Therapeutics program for testing 1 and 2 for inhibition of the growth of the NCI-60 cell line panel.