No job name

J. Am. Chem. Soc. 2000, 122, 12898-12900
Validation of a Model for the Complex of HIV-1
Reverse Transcriptase with Sustiva through
Computation of Resistance Profiles
Robert C. Rizzo, De-Ping Wang, Julian Tirado-Rives, andWilliam L. Jorgensen* Department of Chemistry, Yale UniVersity New HaVen, Connecticut 06520-8107 ReVised Manuscript ReceiVed October 24, 2000 All retroviruses depend on a virally encoded reverse tran- scriptase enzyme (RT) to convert viral RNA into DNA forsubsequent incorporation into the host cell genome.1 Drug-designefforts to arrest reverse transcription in HIV have led to the FDA Figure 1. Orientation of the four NNRTIs in the HIVRT binding site.
approval of three non-nucleoside reverse transcriptase inhibitors (A) Best docked structure of Sustiva. (B) Nevirapine from pdb entry 1vrt.
(NNRTIs), nevirapine, delavaridine, and efavirenz (Sustiva).
(C) MKC-442 from pdb entry 1rt1. (D) 9-Cl TIBO from pdb entry 1rev.
Additional compounds, including MKC-442, are in clinical trials(Table 1). Because of the low fidelity of HIVRT, the mutation the X-ray and docked structures was 0.43-0.60 Å. These low rate in the encoded proteins including HIVRT is great.2,3 As a rmsd values and the limited flexibility of Sustiva are favorable result, all HIVRT inhibitors incur resistance problems that for the accuracy of the docking calculations.
adversely affect their clinical value.4,5 A measure of a drug’s The best docked structure of Sustiva reveals that it makes effectiveness against a mutation is given by the fold resistance, interactions that are consistent with those for other NNRTIs and which is the ratio of mutant to wild-type activities. Sustiva has that it overlays well with the “butterfly” shape of nevirapine.
been shown to remain notably active against several common Unlike nevirapine, hydrogen bonds are present between Sustiva HIVRT point mutations including Val f Ala at position 106 and the protein backbone at position Lys101 that are similar to (V106A) and Tyr f Cys at position 181 (Y181C) (Table 1).
those observed in the crystal structures with 9-Cl TIBO and MKC- No HIVRT structure with Sustiva has been reported that may 442 (Figure 1). Nevirapine makes no formal ligand-protein help explain its improved resistance profile. Herein, we have (a) hydrogen bonds, but it does form a π-type hydrogen bond between computed a structure for the Sustiva/HIVRT complex, (b) the secondary amide hydrogen and Tyr18812 and water-mediated validated the structure through computations of the effects of the hydrogen bonds.10,12 The cyclopropyl ethynyl group of Sustiva V106A and Y181C mutations on binding affinities for four drugs, is positioned toward aromatic residues Tyr181 and Tyr188 in the and (c) obtained structural insights on the improved effectiveness same fashion as the methylpyridine fragment of nevirapine, the of Sustiva. A binding site model6 was constructed from the 2.55-Å benzyl ring of MKC-442, and the dimethylallyl group of 9-Cl crystal structure of the MKC-442/HIVRT complex (pdb 1rt1)7 TIBO (Figure 1). Presumably, these aryl-π interactions all with MKC-442 removed including only those residues within ∼15 contribute favorably to binding.4,5 Superposition based on the Å of MKC-442. The MATADOR program8 was then used to dock HIVRT CR atoms shows that these π fragments of the inhibitors Sustiva into the NNRTI site.9 The other complexes were prepared coincide spatially in the binding site and that Sustiva’s π fragment analogously starting with coordinates from the X-ray structures of nevirapine (pdb 1vrt),10 HEPT (pdb 1rti),7 and 9-Cl TIBO (pdb1rev)11 bound to HIVRT. The docking calculations placed Sustiva An alternative binding mode suggested by Maga et al. was in a reasonable position and orientation in the binding site in based on an alignment of nevirapine and Sustiva in which the comparison with the crystal structures for the complexes with (9) MATADOR uses a Monte Carlo-based Tabu search algorithm. To keep MKC-442, nevirapine, and 9-Cl TIBO (Figure 1). As controls, the Tabu search focused on the known NNRTI binding site during the docking MKC-442, nevirapine, 9-Cl TIBO, and HEPT were also docked runs, a 50 kcal/mol Å2 half-harmonic restraining force was applied if the back into their binding sites to verify that the docking protocol distance between the ligand and the binding site center was greater than 5 Å.
The defined binding site was roughly centered on the alkyne group of Sustiva.
could reproduce experimental structures. The lowest-energy The Tabu list was set to be 25 and constructed from unique structures structure generated during the docking runs was taken as the considering energetic as well as geometric criteria. In total, 100 Tabu cycles “best” structure and was found in all cases to reproduce closely were performed with each Tabu search generating 100 randomly placed ligandpositions around the binding site. The decision to accept a new structure onto the position and orientation observed in the crystal structures; the Tabu lists is made after an intermolecular energy minimization and is the rmsd for the non-hydrogen atoms of the four ligands between based on both energetic and geometric criteria. The protein and ligand wererigid during the docking. The CM1P augmented OPLS-AA force field [Duffy, (1) Katz, R. A.; Skalka, A. M. Annu. ReV. Biochem. 1994, 63, 133-173.
E. M.; Jorgensen, W. L. J. Am. Chem. Soc. 2000, 122, 2878-2888] provided
(2) Preston, B. D.; Poiesz, B. J.; Loeb, L. A. Science 1988, 242, 1168-
the initial structure of Sustiva; it was also used to determine the nonbonded energies, which were stored on a spherical grid for efficiency. The total (3) Roberts, J. D.; Bebenek, K.; Kunkel, T. A. Science 1988, 242, 1171-
intermolecular interactions between the ligand and protein amount to a measure of both steric and electrostatic complimentarity; the lowest energy structure (4) Tantillo, C.; Ding, J. P.; Jacobomolina, A.; Nanni, R. G.; Boyer, P. L.; found during the simulations was taken as the “best” docked system. A Hughes, S. H.; Pauwels, R.; Andries, K.; Janssen, P. A. J.; Arnold, E. J. Mol. distance-dependent dielectric constant of 4 ( ) 4r) was used for all docking Biol. 1994, 243, 369-387.
(5) De Clercq, E. AntiViral Res. 1998, 38, 153-179.
(10) Ren, J.; Esnouf, R.; Garman, E.; Somers, D.; Ross, C.; Kirby, I.; (6) Protein residues included in the binding site model were 91-110A, Keeling, J.; Darby, G.; Jones, Y.; Stuart, D.; et al. Nat. Struct. Biol. 1995, 2,
161-205A, 222-242A, 316-321A, 343-349A, 381-383A, and 134-140B.
(7) Hopkins, A. L.; Ren, J. S.; Esnouf, R. M.; Willcox, B. E.; Jones, E.
(11) Ren, J.; Esnouf, R.; Hopkins, A.; Ross, C.; Jones, Y.; Stammers, D.; Y.; Ross, C.; Miyasaka, T.; Walker, R. T.; Tanaka, H.; Stammers, D. K.; Stuart, D. Structure 1995, 3, 915-26.
Stuart, D. I. J. Med. Chem. 1996, 39, 1589-1600.
(12) Rizzo, R. C.; Tirado-Rives, J.; Jorgensen, W. L. J. Med. Chem. 2000,
(8) Plount Price, M. L. Ph.D. Thesis, Yale University, 2000.
J. Am. Chem. Soc., Vol. 122, No. 51, 2000 12899 Relative Free Energies of Binding (∆GFR) Estimated from Fold Resistance (FR) Valuesg a Young, S. D.; Britcher, S. F.; Tran, L. O.; Payne, L. S.; Lumma, W. C.; Lyle, T. A.; Huff, J. R.; Anderson, P. S.; Olsen, D. B.; Carroll, S. S.; Pettibone, D. J.; Obrien, J. A.; Ball, R. G.; Balani, S. K.; Lin, J. H.; Chen, I. W.; Schleif, W. A.; Sardana, V. V.; Long, W. J.; Byrnes, V. W.; Emini,
E. A. Antimicrob. Agents Chemother. 1995, 39, 2602-2605. b Levin, J http://www.natap.org/reports/NR5-nnrti_update2.resis.htm. c Byrnes, V. W.;
Sardana, V. V.; Schleif, W. A.; Condra, J. H.; Waterbury, J. A.; Wolfgang, J. A.; Long, W. J.; Schneider, C. L.; Schlabach, A. J.; Wolanski, B. S.;
Graham, D. J.; Gotlib, L.; Rhodes, A.; Titus, D. L.; Roth, E.; Blahy, O. M.; Quintero, J. C.; Staszewski, S.; Emini, E. A. Antimicrob. Agents
Chemother. 1993, 37, 1576-1579. d Balzarini, J.; Baba, M.; Declercq, E. Antimicrob. Agents Chemother. 1995, 39, 998-1002. e Baba, M.; Shigeta,
S.; Yuasa, S.; Takashima, H.; Sekiya, K.; Ubasawa, M.; Tanaka, H.; Miyasaka, T.; Walker, R. T.; Declercq, E. Antimicrob. Agents Chemother.
1994, 38, 688-692. f Balzarini, J.; Karlsson, A.; Meichsner, C.; Paessens, A.; Riess, G.; Declercq, E.; Kleim, J. P. J. Virol. 1994, 68, 7986-7992.
g FR ) mutant/wild-type activities, ∆G
RT ln FR in kcal/mol. The columns show the structure, compound name, the assay type and reference for the FR values, and ∆GFR for several common HIVRT mutations.
respectively, while ∆GA and ∆GB are the changes in free energyof binding for A and B with the mutant vs the wild-type protein.
The quantities are related as ∆G ∆∆G, where ∆∆G is the experimentally observable differencein the fold resistance values, RT ln FR - activity ratios from IC or EC values are expected to parallelbinding constant ratios for similar inhibitors.14 Computationally,one could mutate either the drug or the protein. However, we Figure 2. Overlays of the binding-site positions of nevirapine, MKC-
have chosen to perform the structurally simpler mutations of the 442, and 9-Cl TIBO with Sustiva in CPK colors.
protein; specifically, Val106 was mutated to Ala, and Tyr181 wasmutated to Cys in the presence of the four NNRTIs. Our resultsshould yield the observed experimental trends, if the proposedSustiva/HIVRT structure is correct.
Molecular dynamics (MD)15 equilibration and Monte Carlo16 free energy perturbation (MC/FEP)17 simulations were performedon the docked Sustiva structure and the equivalent nevirapine,MKC-442, and 9-Cl TIBO binding-site models, including 850water molecules in the FEP calculations. Aside from normalthermal oscillations, the positioning of Sustiva in the binding sitefrom the docking calculations was maintained in the MD andMC simulations. There is significant variability in the reportedfold resistance data, presumably due to the use of different assayconditions (Table 1). Sustiva, however, consistently emerges as Figure 3. Thermodynamic cycle used to compute relatiVe fold resistance
more tolerant toward the Y181C and V106A mutations than the values. The wild-type side-chain (magenta) is perturbed to the mutant (14) Cheng, Y.; Prusoff, W. H. Biochem. Pharmacol. 1973, 22, 3099-
side chain in the presence of Drug A (solid red) and Drug B (checkered red) while bound to a protein (cyan). Relative fold resistance ) ∆G - (15) The docked Sustiva/HIVRT structure was subjected to a molecular dynamics (MD) simulation to allow the HIVRT backbone and side chains torelax. The CM1P augmented OPLS-AA force field was used for all MDsimulations with the IMPACT program [Levy, R. M. et al., IMPACT Version amide moiety of both drugs was superimposed.13 The present c1.00; Schro¨dinger, Inc.: Jersey City, NJ, 1999]. Ten cycles of gradient-based docking calculations did not find this orientation. Furthermore, energy minimization were performed prior to the MD simulations, and the forced placement of Sustiva in this alternative geometry yielded complex was then restrained in the following manner. Protein residues wereallowed to move freely within ∼10 Å of the binding site (95-107A, 172A, steric and electrostatic protein-ligand interaction energies ∼5 and 177-182A, 188-192A, 198A, 227A, 229A, 234-236A, 318-319A, 321A, 15 kcal/mol, respectively, less favorable than for our docked and 135-139B). Movement was restrained for those residues in a 10-12 Å structure. The alternative orientation is unlikely since the hydrogen shell about the binding site, i.e., for residues 94A, 108A, 175-176A, 183A,187A, 225A, 237-239A, 317A, 320A, 349A, 382-383A, 134B, 140B with bonds to the backbone of Lys101 would be sacrificed.
harmonic potentials. All other residues were restrained to their positions after A computational experiment was then pursued to validate the conjugate-gradient minimization. The Verlet algorithm was used to integrate Sustiva model by predicting relative fold resistance values. The Newton’s equations of motion using a time step of 0.001 ps and constanttemperature was maintained through coupling to a Berendsen temperature bath methodology, which is a general computational approach to using a relaxation parameter of 0.2 ps for the velocity scaling. Bond lengths determining the impact of protein mutations on drug candidates, were fixed by the SHAKE algorithm and a distance-dependent dielectric hinges on the thermodynamic cycle in Figure 3. For two inhibitors, ) 4r) was used. First, 3 ps of initial equilibration was performed at 100 K followed by 50 ps of equilibration at 300 K. Quenching A and B, ∆GWT and ∆GMUT are the differences in free energy of of the structure was performed by reducing the simulation temperature over binding for B vs A with the wild-type and mutant proteins, 6 blocks of 4 ps each starting at 300 K and ending at 50 K. The same MDequilibration was also performed on the nevirapine, MKC-442, and the 9-Cl (13) Maga, G.; Ubiali, D.; Salvetti, R.; Pregnolato, M.; Spadari, S.
TIBO structures, and the resultant complexes were then used in the MC Antimicrob. Agents Chemother. 2000, 44, 1186-1194.
12900 J. Am. Chem. Soc., Vol. 122, No. 51, 2000 Relative Fold Resistance Energies (∆∆G) in kcal/mol for be partly compensated by better alignment of the NH-O hydrogen bond with Lys101 when the buttressing effect of the valine sidechain is reduced by conversion to alanine. In the MC simulations, the hydrogen bond between the oxazinone NH of Sustiva and the carbonyl oxygen of Lys101 is on average 0.1 Å shorter (1.77 vs 1.85 Å) when residue 106 is Ala rather than Val. The 3.88 ( 0.3 2.20, 2.71, 2.90 3.33 ( 0.4 2.34, 2.27, 2.95 interaction of the valine’s isopropyl group with the weakly polarizable trifluoromethyl group is also likely less attractive than 9-Cl TIBO 3.01 ( 0.3 1.05, 0.41, 2.27 1.32 ( 0.5 0.66, 1.22, 1.48 the corresponding interactions with the cyclopropyl group of a Values derived from Table 1.
nevirapine and the isopropyl and ethoxymethyl groups of MKC-442 (Figure 1). Thus, it is reasonable to propose, on the basis of other drugs, especially nevirapine and MKC-442. Indeed, the the present structure, that Sustiva’s improved resistance profile present FEP results do predict Sustiva to be less affected by both benefits from a combination of less favorable initial interactions mutations than the other three inhibitors (Table 2). The agreement with Tyr181 and Val106 and more favorable hydrogen bonding of the computed free energies with the experimental results with Lys101 in the V106A mutant. Consistently, the L100I strongly supports the correctness of our docked model.
mutation is more damaging (Table 1) because Leu100 forms a The structural model suggests some factors that render Sustiva snug lid over the ring systems for all four inhibitors (Figure 1).
less affected by the Y181C and V106A mutations in comparison Without adjustment, the branching at C rather than Cγ would with the other compounds. It is well-known that the NNRTI direct the methyl group of Ile100 directly into the rings. An binding site is capable of accommodating structurally diverse alternative strategy for improved resistance profiles is to enhance inhibitors and that different inhibitors give rise to strikingly interactions with immutable residues such as Trp229.18 different patterns of resistance mutations among ∼15 residues In this work, we have presented a molecular model for the that line the binding site.4,5 In general, this variability implies important anti-HIV drug Sustiva bound to HIVRT. The resultant that the effect of mutations on drug binding needs assessment on structure reveals that Sustiva overlays well with the butterfly shape a case-by-case basis. However, the Y181C mutant arises early of nevirapine and makes similar contacts with HIVRT as do other and confers resistance for many NNRTIs. This can be attributed reported NNRTIs including hydrogen bonds with the backbone to the loss of favorable aryl/π interactions, for example, between of Lys101 (Figure 1). FEP methodology for the assessment of the tyrosine and the methylpyridyl and benzyl rings of nevirapine relative resistance profiles for drug candidates has been defined.
and MKC-442, and the dimethylallyl group of 9-Cl TIBO (Figure Results from its application to four NNRTIs (Table 2) are in good 1). Loss of the interaction between Tyr181 and the smaller, less agreement with the experimental activity trends and provide polarizable cyclopropyl ethynyl group of Sustiva is expected to evidence that the proposed binding mode for Sustiva is correct.
Sustiva’s relative insensitivity to the Y181C and V106A mutants In the case of V106A, the residue is tucked under the benzene appears to arise from a mix of relatively weaker interactions with ring of Sustiva and is in van der Waals’ contact with the Tyr181 and Val106 and improvement of hydrogen bonding for trifluoromethyl group. Reduction of these interactions appears to Ala106. These findings highlight the power of molecular modelingfor structure and binding affinity predictions and its potential for (16) Each protein-inhibitor complex was briefly energy-minimized prior to the Monte MC simulations using a distance-dependent dielectric constantof 4 ( ) 4r). The CM1P augmented OPLS-AA force field was used with the Acknowledgment. This work was supported by the National Institute
MCPRO program [Jorgensen, W. L. MCPRO Version 1.65; Yale University: of Allergy and Infectious Diseases (AI44616). We thank Melissa L. Plount New Haven, CT, 2000]. For the MC simulations, a water cap with a 22-Åradius was used containing ∼850 TIP4P water molecules and the system was Price for computational assistance with the docking calculations, and Dr.
partitioned into rigid residues (91-94A, 109-110A, 116-178A, 184-185A, Marilyn B. Kroeger Smith, Professor Richard H. Smith, and Mark A.
192-197A, 199-205A, 222-224A, 230-232A, 240-242A, 316-317A, 320-321A, 343-349A, 381-383A, 134-135B, 137B, 140B) and flexibleresidues (95-108A, 179-183A, 186-191A, 198A, 225-229A, 233-239A, Note Added in Proof. A crystal structure has now appeared
318-319A, 136B, 138B). All HIVRT side chains within ∼10 Å from the center of the water cap were sampled, the protein backbone was fixed, and for a Sustiva/HIVRT complex [Ren, J.; Milton, J.; Weaver, K.
each inhibitor was fully flexible. Each solvated complex was subjected to 1 L; Short, S. A.; Stuart, D. I.; Stammers, D. K. Structure 2000, 8,
million configurations of solvent-only equilibration, 10 million of equilibration, 1089-1094]. It fully confirms the correctness of the structure and 10 million configurations of averaging per window during the FEPsimulations. For additional MC simulation parameters and protocols, see (17) For a recent discussion of FEP methodology, see: Jorgensen, W. L.
Free Energy Changes in Solution. In Encyclopedia of ComputationalChemistry; Schleyer, P. v. R., Ed.; Wiley: New York, 1998; Vol. 2, pp 1061- (18) Hopkins, A. L.; Ren, J.; Tanaka, H.; Baba, M.; Okamato, M.; Stuart, D. I.; Stammers, D. K. J. Med. Chem. 1999, 42, 4500-4505.

Source: http://rizzo.ams.sunysb.edu/~rizzo/StonyBrook/publications/rizzo003.pdf