Calculated and experimental low-energy conformations of cyclic urea HIV protease inhibitors

Article abstract of DOI:10.1021/ja972357h

One important factor influencing the affinity of a flexible ligand for a receptor is the internal strain energy required to attain the bound conformation. Calculation of fully equilibrated ensembles of bound and free ligand and receptor conformations are computationally not possible for most systems of biological interest; therefore, the qualitative evaluation of a novel structure as a potential high- affinity ligand for a given receptor can benefit from taking into account both the bound and unbound (usually aqueous) low-energy geometries of the ligand and the difference in their internal energies. Although many techniques for computationally generating and evaluating the conformational preferences of small molecules are available, there are a limited number of studies of complex organics that compare calculated and experimentally observed conformations. To assess our ability to predict a priori favored conformations of cyclic HIV protease (HIV-1 PR) inhibitors, conformational minima for nine 4,7- bis(phenylmethyl)-2H-1,3-diazepin-2-ones I (cyclic ureas) were calculated using a high temperature quenched dynamics (QD) protocol. Single crystal X-ray and aqueous NMR structures of free cyclic ureas were obtained, and the calculated low-energy conformations compared with the experimentally observed structures. In each case the ring conformation observed experimentally is also found in the lowest energy structure of the QD analysis, although significantly different ring conformations are observed at only slightly higher energy. The 4- and 7-benzyl groups retain similar orientations in calculated and experimental structures, but torsion angles of substituents on the urea nitrogens differ in several cases. The data on experimental and calculated cyclic urea conformations and their binding affinities to HIV-1 PR are proposed as a useful dataset for assessing affinity prediction methods.

Full text of DOI:10.1021/ja972357h

4570
J. Am. Chem. Soc. 1998, 120, 4570-4581
Calculated and Experimental Low-Energy Conformations of Cyclic
Urea HIV Protease Inhibitors
,
‡
‡
‡
‡
C. Nicholas Hodge,* Patrick Y. S. Lam, Charles J. Eyermann, Prabhakar K. Jadhav,
‡
‡
‡
‡
Y. Ru, Christina H. Fernandez, George V. De Lucca, Chong-Hwan Chang,
Robert F. Kaltenbach III, Edward R. Holler, Francis Woerner, Wayne F. Daneker,
George Emmett, Joseph C. Calabrese, and Paul E. Aldrich
‡
‡
‡
‡
‡
†
‡
Contribution from the Chemical and Physical Sciences Department, Research DiVision, DuPont Merck
Pharmaceutical Company, P.O. Box 80500, Wilmington, Delaware 19880-0500, and Central Research
Department, DuPont Company, Wilmington, Delaware 19880-0328
ReceiVed July 14, 1997
Abstract: One important factor influencing the affinity of a flexible ligand for a receptor is the internal strain
energy required to attain the bound conformation. Calculation of fully equilibrated ensembles of bound and
free ligand and receptor conformations are computationally not possible for most systems of biological interest;
therefore, the qualitative evaluation of a novel structure as a potential high-affinity ligand for a given receptor
can benefit from taking into account both the bound and unbound (usually aqueous) low-energy geometries of
the ligand and the difference in their internal energies. Although many techniques for computationally generating
and evaluating the conformational preferences of small molecules are available, there are a limited number of
studies of complex organics that compare calculated and experimentally observed conformations. To assess
our ability to predict a priori favored conformations of cyclic HIV protease (HIV-1 PR) inhibitors, conformational
minima for nine 4,7-bis(phenylmethyl)-2H-1,3-diazepin-2-ones I (cyclic ureas) were calculated using a high
temperature quenched dynamics (QD) protocol. Single crystal X-ray and aqueous NMR structures of free
cyclic ureas were obtained, and the calculated low-energy conformations compared with the experimentally
observed structures. In each case the ring conformation observed experimentally is also found in the lowest
energy structure of the QD analysis, although significantly different ring conformations are observed at only
slightly higher energy. The 4- and 7-benzyl groups retain similar orientations in calculated and experimental
structures, but torsion angles of substituents on the urea nitrogens differ in several cases. The data on
experimental and calculated cyclic urea conformations and their binding affinities to HIV-1 PR are proposed
as a useful dataset for assessing affinity prediction methods.
Introduction
proteases.5,6 As with other HIV protease inhibitors, even those
7
now in clinical use, virus has been identified with decreased
The search for inhibitors of HIV protease with clinically
useful antiviral properties has led to the discovery of a variety
of highly potent compounds. We recently described a novel
series of HIV protease inhibitors that were designed for specific
binding to retroviral, as opposed to mammalian, aspartyl
proteases by displacing a structural water molecule found only
in retroviral proteases, and to provide a compact, rigid structure
that optimized tight binding interactions with the enzyme while
minimizing unnecessary molecular weight and conformational
freedom. X-ray crystallographic and NMR studies of the
complexes have demonstrated that these compounds do in fact
bind to HIV protease in the expected manner. Since our initial
report we have identified cyclic ureas that are highly potent,
selective, orally available inhibitors of a broad set of retroviral
8
sensitivity to first-generation cyclic ureas, suggesting that the
compounds may be of limited benefit over the long term. Our
recent focus has been to apply structure-based design and
medicinal chemistry to identify cyclic urea analogues with high
1
2
(4) (a) Yamazaki, T.; Nicholson, L. K.; Wingfield, P.; Stahl, S. J.;
3
Kaufman, J. D.; Eyermann, C. J.; Hodge, C. N.; Lam, P. Y. S.; Torchia, D.
A.; et. al. J. Am. Chem. Soc. 1994, 116, 10791-10792. (b) Grzesiek, S.;
Bax, A.; Nicholson, L. K.; Yamazaki, T.; Wingfield, P.; Stahl, S. J.;
Eyermann, C. J.; Torchia, D. A.; Hodge, C. N.; Lam, P. Y.; Jadhav, P. K.;
Chang, C. H. J. Am. Chem. Soc. 1994, 116, 1581-2. (c) Nicholson, L. K.;
Yamazaki, T.; Torchia, D. A.; Grzesiek, S.; Bax, A.; Stahl, S. J.; Kaufman,
J. D.; Wingfield, P. T.; Lam, P. Y. S.; Jadhav, P. K. Hodge, C. N.; Domaille,
P. J.; Chang, C. H. Nat. Struct. Biol. 1995, 2, 274-280.
2
4
(5) (a) Hodge, C. N.; Aldrich, P. E.; Bacheler, L. T.; Chang, C.-H.;
Eyermann, C. J.; Garber, S.; Grubb, M.; Jackson, D. A.; Jadhav, P. K.;
Korant, B.; Lam, P. Y. S.; Maurin, M. B.; Meek, J. L.; Otto, M. J.; Rayner,
M. M.; Reid, C.; Sharpe, T. R.; Shum, L.; Winslow, D. L.; Erickson-
Viitanen, S. Chem. Biol. 1996, 3, 301-315. (b) Lam, P.; Ru, Y.; Jadhav,
P. K.; Aldrich, P. E.; DeLucca, G. V.; Eyermann, C. J.; Chang, C. H.;
Emmett, G.; Holler, E. R.; Daneker, W. F.; Li, L.; Confalone, P. N.;
McHugh, R. J.; Han, Q.; Li, R.; Markwalder, J. A.; Seitz, S. P.; Sharpe, T.
R.; Bacheler, L. T.; Rayner, M. R.; Klabe, R. M.; Shum, L.; Winslow, D.
L.; Kornhauser, D. M.; Jackson, D. A.; Erickson-Viitanen, S.; Hodge, C.
N. J. Med. Chem. 1996, 39, 3514-3525. (c) Nugiel, D. A.; Jacobs, K.;
Worley, T.; Patel, M.; Kaltenbach, R. F. I.; Meyer, D. T.; Jadhav, P. K.;
De Lucca, G. V.; Smyser, T. E.; Klabe, R. M.; Bacheler, L. T.; Rayner, M.
M.; Seitz, S. P. J. Med. Chem. 1996, 39, 2156-2169.
†
Central Research Department, DuPont Company.
Research Division, DuPont Merck Pharmaceutical Company.
‡
(
4.
1) West, M. L.; Fairlie, D. P. Trends Pharmacol. Sci. 1995, 16, 67-
7
(2) Lam, P. Y. S.; Jadhav, P. K.; Eyermann, C. J.; Hodge, C. N.; Ru,
Y.; Bacheler, L. T.; Meek, O. M. J.; Rayner, M. M.; Wong, Y. N.; Chang,
C. H.; Weber, P. C.; Jackson, D. A.; Sharpe, T. R.; Erickson-Viitanen, E.
V. Science 1994, 263, 380-384.
(3) Miller, M.; Schneider, J.; Sathyanarayana, B. K.; Toth, M. V.;
Marshall, G. R.; Clawson, L.; Selk, L.; Kent, S. B. H.; Wlodawer, A. Science
989, 246, 1149-1152.
1
S0002-7863(97)02357-3 CCC: $15.00 © 1998 American Chemical Society
Published on Web 05/02/1998

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4571
potency against resistant variants that retain the favorable
physicochemical properties of the initial series.9
which the active site of the receptor is used as a negative mold
for the unbiased construction of novel ligands and fragments
are fitted into the active site with fast scoring algorithms and
in some cases connected to form potent inhibitors, has also been
applied to HIV protease. Developing quantitative structure-
activity data from both ligand and macromolecular structure has
In this project, as in most drug discovery efforts, once a new
series of biologically active ligands with adequate in vitro
properties is identified it is necessary to optimize the physical
properties of the lead compound to obtain desired in vivo
properties for clinical use; these properties include oral bio-
availability, safety, stability, and ease of synthesis and formula-
tion. In the process of modulating physical properties the target
activity of the lead series is frequently lost or diminished,
resulting in a time-consuming cycle of analogue synthesis and
testing. Traditionally this process has been successfully ad-
dressed by applying medicinal chemistry and quantitative
structure activity relationship (QSAR) analyses, but these
methods do not adequately take into account the effect of
modifications of the ligand on its three-dimensional shape, either
free or bound to the enzyme. Computational methods that
provide even a crude estimate of the binding affinity of
suggested novel scaffolds and functional groups for HIV-1 PR
can assist medicinal chemists in prioritizing the large number
of possible candidate ligands conceived in their effort to balance
potency against the target receptor with the required physical
properties.10
1
5
yielded interesting results in HIV protease and phospholipase
1
6
A2.
A less sophisticated but more rapid approach is to consider
a potent, selective ligand with a defined manner of binding to
a known receptor as a positive mold and to screen novel
proposed ligands by their ability to attain the shape and
electrostatic character of this known reference compound:17
essentially a molecular field analysis in which the “field” is
defined by the shape of a single bound ligand of known high
affinity. If a proposed new compound cannot project the
important recognition elements into the correct regions of space
in an energetically favorable fashion, i.e., with internal strain
energy at or near its global minimum, then it is rejected as a
synthetic target. If the required shape is found among a large
number of other low-energy conformations that do not provide
a good match to the reference compound, then the target is
intermediate in quality; and if the proposed structure displays a
strong energetic bias toward the required shape, then it is,
relatively speaking, an attractive candidate for experimental
Considerable progress toward this endsthe qualitative as-
sessment of the affinity of a hypothetical ligand for a receptorshas
been made by the development of 3D QSAR and Molecular
Field Analysis models,11 wherein the activities of known
analogues are used to infer a set of binding determinants or a
receptor field, and the activity of a new analogue is predicted
based on interactions with that field. Docking,12 with and
without flexibility of the ligand and the protein, has been used
to evaluate modifications of known inhibitors and to screen
three-dimensional structure databases for novel ligands. Free
energy perturbation methods have been used to rationalize
observed binding affinity changes in stereoisomeric ligands or
isosteric replacements of small groups.13 De novo design, in
1
8
validation.
In practice, many flexible compounds that are proposed as
mimetics of a ligand whose binding conformation is known do
(12) (a) Meng, E. C.; Gschwend, D. A.; Blaney, J. M.; Kuntz, I. D.
Proteins: Struct., Funct., Genet. 1993, 17, 266-278. (b) Stoddard, B. L.;
Koshland, D. E., Jr. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 1146-1153.
(
c) Yamada, M.; Itai, A. Chem. Pharm. Bull. 1993, 41, 1203-5. (d) Cherfils,
J.; Janin, J. Curr. Opin. Struct. Biol. 1993, 3, 265-269. (e) Di Nola, A.;
Roccatano, D.; Berendsen, H. J. C. Proteins: Struct., Funct., Genet. 1994,
1
9, 174-82. (f) Goodsell, D. S.; Olson, A. J. Proteins: Struct., Funct.,
Genet. 1990, 8, 195-202. (f) Goodsell, D. S.; Lauble, H.; Stout, C. D.;
Olson, A. J. Proteins: Struct., Funct., Genet. 1993, 17, 1-10. (g) Kearsley,
S. K.; Underwood, D. J.; Sheridan, R. J. Comput.-Aided Mol. Des. 1994,
8, 565-582. (h) Luty, B. A.; Wasserman, Z. R.; Stouten, P. F. W.; Hodge,
C. N.; Zacharias, M.; McCammon, J. A. J. Comput. Chem. 1995, 16, 454-
464. (i) Guarnieri, F.; Wilson, S. R. J. Comput. Chem. 1995, 16, 648-653.
(j) Wasserman, Z. R.; Hodge, C. N. Proteins: Struct., Funct., Genet.1996,
24, 227-237.
14
(6) (a) Erickson-Viitanan, S.; Klabe, R. M.; Cawood, P. G.; O’Neal, P.
L.; Meek, J. L. Antimicrob. Agents Chemother. 1994, 38, 1628-1634. (b)
Grubb, M. F.; Wong, Y. N.; Burcham, D. L.; Saxton, P. L.; Quon, C. Y.;
Huang, S. M. Drug Metab. Dispos. 1994, 22, 709-712. (c) Rayner, M.
M.; Cordova, B. C.; Meade, R. P.; Aldrich, P. E.; Jadhav, P. K.; Ru, Y.;
Lam, P. Y. S. Antimicrob. Agents Chemother. 1994, 38, 1635-1640. (d)
Wong, Y. N.; Burcham, D. L.; Saxton, P. L.; Erickson-Viitanen, S.; Grubb,
M. F.; Quon, C. Y.; Huang, S.-M. Biopharm. Drug Dispos. 1994, 15, 535-
(13) (a) Cieplak, P.; Kollman, P. A. J. Comput.-Aided Mol. Des. 1993,
7(3), 291-304. (b) DeBolt, S. E.; Pearlman, D. A.; Kollman, P. A. J.
Comput. Chem. 1994, 15, 351-373. (c) Ferguson, D. M.; Radmer, R. J.;
Kollman, P. A. J. Med. Chem. 1991, 34, 2654-2659. (d) Hansson, T.;
Aaqvist, J. Protein Eng. 1995, 8, 1137-1144. (e) Rao, B. G.; Murcko, M.
A. J. Comput. Chem. 1994, 15, 1241-1253. (f) Reddy, M. R.; Viswanadhan,
V. N.; Weinstein, J. N. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 10287-
10291. (g) Reddy, M. R.; Varney, M. D.; Kalish, V.; Viswanadhan, V. N.;
Appelt, K. J. Med. Chem. 1994, 37, 1145-1152. (h) Varney, M. D.; Appelt,
K.; Kalish, V.; Reddy, M. R.; Tatlock, J.; Palmer, C. L.; Romines, W. H.;
Wu, B.-W.; Musick, L. J. Med. Chem. 1994, 37, 2274-2284.
(14) (a) Clark, T. D.; Ghadiri, M. R. J. Am. Chem. Soc. 1995, 117,
12364-12365. (b) Pearlman, D. A.; Murcko, M. A. J. Comput. Chem. 1993,
14, 1184-1193. (c) Baca, M.; Alewood, P. F.; Kent, S. B. H. Protein Sci.
1993, 2, 1085-1091. (d) ibid. idem., 623-32. (e) Boehm, H.-J. J. Comput.-
Aided Mol. Des. 1994, 8, 243-256. (f) Rotstein, S. H.; Murcko, M. A. J.
Med. Chem. 1993, 36, 1700-1710. (g) Bures, M. G.; Hutchins, C. W.;
Maus, M.; Kohlbrenner, W.; Kadam, S.; Erickson, J. W. Tetrahedron
Comput. Methodol. 1990, 3, 673-680. (h) Roe, D. C.; Kuntz, I. D. J.
Comput.-Aided Mol. Des. 1995, 9, 269-282. (i) Waszkowycz, B.; Clark,
D. E.; Frenkel, D.; Li, J.; Murray, C. W.; Robson, B.; Westhead, D. R. J.
Med. Chem. 1994, 37, 3994-4002.
(15) (a) Verkhivker, G. M.; Rejto, P. A. Proc. Natl. Acad. Sci. U.S.A.
1996, 93, 60-64. (b) Verkhivker, G.; Appelt, K.; Freer, S. T.; Villafranca,
J. E. Protein Eng. 1995, 8, 677-691.
(16) Ortiz, A. R.; Pisabarro, M. T.; Gago, F.; Wade, R. C. J. Med. Chem.
1995, 38, 2681-2691.
(17) Hodge, C. N.; Straatsma, T. P.; McCammon, J. A.; Wlodawer, A.
Rational Design of HIV Protease Inhibitors. In Structural Biology of Viruses;
Chiu, W., Burnett, R. M., Garcea, R., Eds.; Oxford University Press: New
York, 1997; pp 451-473.
4
4. (e) Ottto, M. J.; Reid, C. D.; Garber, S.; Lam, P. Y. S.; Scarnati, H.;
Bacheler, L. T.; Rayner, M. M.; Winslow, D. L. Antimicrob. Agents
Chemother. 1993, 37, 2606-2611.
(
7) For reviews see: Lin, J. H. AdV. Drug DeliVery ReV. 1997, 27(2, 3),
7
2
74-778. Vacca, J. P.; Condra, J. H. Drug DiscoVery Today 1997, 2(7),
61-272.
(8) (a) Otto, M. J.; Garber, S.; Winslow, D. L.; Reid, C. D.; Aldrich, P.;
Jadhav, P. K.; Patterson, C. E.; Hodge, C. N.; Cheng, Y. S. E. Proc. Natl.
Acad. Sci. U.S.A. 1993, 90, 7543-7547. (b) Reference 6a. (c) Winslow, D.
L.; Anton, E. D.; Horlick, R. A.; Zagursky, R. J.; Tritch, R. J.; Scarnati,
H.; Ackerman, K.; Bacheler, L. T. Biochem. Biophys. Res. Commun. 1994,
2
05, 1651-1657. (d) Winslow, D. L.; Stack, S.; King, R.; Scarnati, H.;
Bincsik, A.; Otto, M. J. AIDS Res. Hum. RetroViruses 1995, 11, 107-113.
e) King, R. W.; Garber, S.; Winslow, D. L.; Reid, C.; Bachelor, L. T.;
Anton, E.; Otto, M. J. AntiViral Chem. Chemother. 1995, 6, 80-88.
9) (a) Ala, P.; Huston, E.; Klabe, R.; McCabe, D.; Duke, J.; Rizzo, C.;
(
(
Korant, B.; DeLoskey, R.; Lam, P. Y. S.; Hodge, C. N.; Chang, C.-H.
Biochemistry 1997, 36, 6(7), 1573-1580. (b) Jadhav, P. K.; Ala, P.;
Woerner, F. J.; Chang, C.-H.; Garber, S. S.; Anton, E. D.; Bacheler, L. T.
J. Med. Chem. 1997, 40, 181-191.
(
10) (a) Lybrand, T. P.; Kollman, P. A. J. Chem. Phys. 1985, 83, 2923-
2
933. (b) Kuntz, I. D. Science 1992, 257(5073), 1078-1982. (c) Crippen,
G. M. J. Comput. Chem. 1995, 16, 486-500. (d) Ajay, Murcko, M. J. Med.
Chem. 1995, 38, 4953-4967.
(
11) (a) Cho, S. J.; Tropsha, A. J. Med. Chem. 1995, 38, 1060-6. (b)
Klebe, G.; Abraham, U. J. Med. Chem. 1993, 36, 70-80. (c) Yamada, M.;
Itai, A. Chem. Pharm. Bull. 1993, 41, 1203-1205. (d) Cruciani, G.; Watson,
K. A. J. Med. Chem. 1994, 37, 2589-2601.

4572 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
Hodge et al.
Table 1
I
a
entry
R
1
R
2
X
Ki, nm
1
1
1
1
1
1
1
1
1
1
1
A
B
C
D
E
F
G
H
H
methyl
H
methyl
(S)-OH
(S)-OH
(S)-OH
(R)-OH
(S)-OH
(S)-OH
(S)-OH
(S)-OH
(S)-OH
(R)-OAc
(S)-OH
4500
5700
1.9
cyclopropylmethyl
cyclopropylmethyl
(2-naphthyl)methyl
4-fluorobenzyl
(2-naphthyl)methyl
3-aminobenzyl
3-aminobenzyl‚(HCl)
H
cyclopropylmethyl
cyclopropylmethyl
H
4-fluorobenzyl
2-naphthylmethyl
3-aminobenzyl
3-aminobenzyl‚(HCl)
H
1.1
2.8
1.4
0.23
0.25
03.25
>10
4.7
H‚HCl
I
J
2
2
allyl
allyl
a
Inhibition constant vs HIV-1 PR (see refs 2 and 5).
not attain a unique low-energy geometry that overlays well with
the target shape, and considerable efficiency is added to the
discovery process if these compounds are screened out prior to
energy conformations of cyclic urea HIVPR inhibitors; (b) detail
the synthesis, X-ray, and H NMR structures of a series of these
1
structures; and (c) compare the calculated structures with
experimental conformations of both enzyme-bound and free
cyclic ureas. Our results indicate that, with some caution, the
force fields and protocols we describe are useful as one of
several necessary steps in estimating ligand binding affinities.23
In addition, our previously reported structures of several cyclic
ureas bound to HIV protease are or will be available to the public
via the Brookhaven Protein Data Bank: complexes related to
this study include XK263, DMP450, XK216, and DMP323
bound to wild type and several mutants (refs 2 and 5). These
combined data on affinity constants, calculated structures, and
experimental structures of free ligands and complexes are
proposed as useful known controls for testing methods of ligand-
protein docking and affinity prediction.
19
synthesis. If desired, the ligand can then be evaluated further
by modeling other factors contributing to enzyme affinity, such
as the enthalpy of interaction with the receptor and of receptor
reorganization on ligand binding; the free energy of desolvation
of ligand; the free energy of de- or resolvation of the receptor
site, if the new ligand is of different size than the reference
ligand; and the entropic cost of restricting enzyme and ligand
to their bound conformations.20 If net energetic estimates are
favorable relative to the reference ligand, then it is prioritized
as a synthetic target based on these considerations and other
factors such as synthetic accessibility and the likelihood of
improved physical properties. But the key is that the initial
shape evaluation is fast relative to synthesis and useful for crude
screening of diverse novel structures for steric fit.
In all of the methods described above, irrespective of whether
an explicit (e.g., crystal coordinates) or an implicit (e.g.,
CoMFA) receptor field is evaluated, an accurate evaluation of
low-energy ligand conformations is required. This in turn
requires a method of searching conformational space and
evaluating the energies of stable conformations and is a slow-
or rate-determining step in many procedures that take ligand
conformational flexibility and internal strain energy into account.
The importance of ligand flexibility in determining binding free
Methods
Chemistry. The synthesis of the protected precursors to
substituted cyclic ureas have been published.
5a,b
The compounds
used in this study, shown in Table 1, were synthesized as
described in the Experimental Section: treatment of urea 7 with
various alkylating agents in the presence of sodium hydride in
dry dimethylformamide, followed by aqueous workup, removal
of the protecting groups, and purification yielded 1C and 1E-
1
G. Oxidation of 1A to the hydroxyketone and reduction to
energy is addressed in an recent study by Vajda and co-
workers.21 A more detailed discussion of other methods of
the R,S,R,R diol provided 1D; the absolute stereochemistry is
confirmed by the X-ray crystal structure (see Supporting
Information). Details of the synthesis will be described in a
subsequent manuscript. 1H and 1I were prepared as described
in ref 5a, and 1B was prepared as described in ref 5b.
Single-Crystal X-ray Analysis. Structural characterization
of compounds 1B-1I and 1H‚(HCl)2 were carried out using
determining ligand conformations is provided below.
The purpose of this report is to (a) describe parameters for a
22
quenched dynamics simulation that rapidly generates low-
(18) A fundamental assumption in implicit in this approach: structures
that can obtain the binding geometry only at the expense of significant
internal strain energy are not of primary interest as targets. Several arguments
can be made to counter this assumption: (a) there are unpublished
indications that many ligands in the Brookhaven Protein Data Bank appear
to retain significant internal strain energy on binding; (b) the possibility
that the net effect of binding to the receptor may be to stabilize ligand
conformations that are less favorable in the unbound state; and (c) a
preorganized ligand may not have a kinetic pathway to bind to its target.
However, in the prospectiVe design process, in which many good ideas
must be prioritized into a few synthetic targets, it is a reasonable starting
point to expect an energetic advantage for a ligand whose low energy
conformation in the ambient milieu (in our case, aqueous pH 5.5 buffer) is
very close to its optimal bound conformation in the target receptor.
(22) (a) Al-Obeidi, F.; Hadley, M. E.; Pettitt, B. M.; Hruby, V. J. J. Am.
Chem. Soc. 1989, 111, 3413-3416. (b) O’Connor, S. D.; Smith, P. E.; Al-
Obeidi, F.; Pettitt, B. M. J. Med. Chem. 1992, 35, 2870-2881.
(23) During preparation of this manuscript Hulten et al. reported on the
behavior of DMP323 and some other P1/P1′ cyclic urea and cyclic
sulfonamide analogues in molecular dynamics simulations in water and in
the presence of HIV protease. A full conformational search was not
undertaken nor are free ligand geometries analyzed experimentally, which
is the focus of our research. The results do show weak correlation of
calculated and experimental affinities to HIV-1 protease using the GROMOS
force field. Hulten, J.; Bonham, N. M.; Nillroth, U.; Hansson, T.; Zuccarello,
G.; Bouzide, A.; Aaqvist, J.; Classon, B.; Danielson, H.; Karlen, A.;
Kvarnstrom, I.; Samuelsson, B.; Hallberg, A. J. Med. Chem. 1997, 40, 0(6),
885-897.
(
(
(
19) Yamada, M.; Itai, A. Chem. Pharm. Bull. 1993, 41, 1203-1205.
20) Ajay; Murcko, M. A. J. Med. Chem. 1995, 38, 4953-67.
21) Vajda, S.; Wheng, Z.; Rosenfeld, R. Biochemistry 1994, 33, 13977-
1
3988.

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4573
standard X-ray crystallographic techniques. Complete reports
including crystallization conditions, atomic coordinates, thermal
parameters, and interatomic distances and angles are available
as Supporting Information.
Solution Structure Analysis. NMR spectra were taken on
a solution containing approximately 30 mg of sample dissolved
in 0.8 mL of the appropriate solvent. Proton and carbon
chemical shifts were referenced to external TSP (TSP ) sodium
need to be evaluated in our system. More accurate semiem-
pirical AM1 optimization in the gas phase is reasonably fast
for single conformations, but multiple conformations would need
2
7
to be sampled. Metropolis Monte Carlo (MC) and distance
2
8
geometry with penalty functions, followed by energy mini-
mization, would be as accurate as the method employed here;
for these latter two and quenched dynamics the slow (minimiza-
tion) step would be identical, so speed would depend on the
number of structures generated that minimized to the same
family (a deterministic systematic search algorithm that is
optimized to reduce the number of minimizations was reported
recently in ref 26c; the speed of this method on small ring
systems has not yet been described). Another recent report
describes an alternating Monte Carlo random sampling and
stochastic dynamics (SD) simulation strategy for conformational
2
,2,3,3-d4-trimethylsilylpropionate, D2O solvent) or the relevant
1
organic solvent peaks. H chemical shift and coupling constant
data were derived from spin simulation calculations when
needed. All spectra were obtained at 30 °C.
1H and 13C spectra were taken on either a Varian VXR-400S
or a Varian Unity-400 NMR spectrometer (Palo Alto, CA)
operating at 399.95 and 100.59 MHz, respectively. A standard
29
13
searching that appears to be both fast and accurate and allows
the calculation of free energies from ensembles of conformations
as well as the use of a continuum solvation model. The
comparison of a quenched dynamics protocol with mixed MC/
SD has not been carried out to our knowledge. Other methods
have been compared and discussed in refs 24a-b. A promising
algorithm for the complete analysis of conformational free
energies, termed “mining minima”, was also published re-
switchable probe was used for the one-dimensional C; all other
data were acquired using an inverse detection probe equipped
with a z-gradient coil. 1H spectrum were recorded with a digital
resolution of 0.25 Hz/pt (AT ) 4.06s), a tip angle of 30°, and
13
a relaxation delay of 2.0 s. C spectra were recorded with a
digital resolution of 0.76 Hz/pt (AT ) 1.31 s), a 60° tip angle,
1
13
and a 4.0 s relaxation delay. H- C correlation spectra were
1
obtained using a value of 140 Hz for JC-H and 7 Hz for
3
0
3
cently.
JC-H. NOESY spectra were acquired using a 2.0 s presaturation
pulse, a 3.0 s predelay, and a 1.0 s mixing time.
In our work, the Biosym consistent valence force field,
3
1
Ki Determination. Inhibition constants of compounds 1A-
I were determined as described earlier (ref 6a). This selection
cff91, was used for molecular dynamics with standard atom
parameters (parameters and input files are available as Sup-
porting Information). This force field has succeeded in
reproducing experimental geometries of small organic molecules
1
of compounds is of interest as the inhibition constants span more
than four logs, and it may provide a useful dataset for
benchmarking affinity prediction methods (see Discussion).
Conformational Search. Many routines are available for
3
2
in several studies. Comparative studies with other empirical
33
and semiempirical methods have also been reported. The
quenched dynamics simulation was run using Discover 2.9 in
stand-alone mode on a Silicon Graphics Challenge 8 processor
R4400 server as follows: time step of 1 fs, dielectric ꢀ, k
femtoseconds of simulation at T °C, minimization of every
using empirical force fields to find low-energy conformations
of drug-sized molecules24 (during this discussion “energy” will
refer to the calculated internal strain energy of a single
conformation in the gas phase). The method we employed has
been described as high-temperature quenched dynamics by Pettit
1
000th frame using 200 steps of steepest descents to RMSD of
2
2
<1.0 kcal/Å, 500 steps of conjugate gradients to RMSD <0.1
kcal/Å, and 1000 steps of VA90A (modified Newton-Rapheson)
to RMSD <0.01 kcal/Å. The minimized structure was archived,
and co-workers and employs molecular dynamic simulations
at temperatures high enough to overcome conformational
barriers and minimization at fixed intervals along the trajectory
to yield representative low energy structures. Due to the limited
degrees of freedom of the ligands, techniques to avoid local
minima such as simulated annealing are not necessary. The
method is rapid and, as described here and elsewhere, appears
to provide broad sampling of conformational space provided
(26) (a) Smellie, A.; Kahn, S. D.; Teig, S. L. J. Chem. Inf. Comput. Sci.
1
1
995, 35, 285-294. (b) Leach, A. R.; Kuntz, I. D. J. Comput. Chem. 1992,
3, 730-748. (c) Saunders: M.; Houk, K. N.; Wu, Y.-D.; Still, W. C.;
Lipton, M. Chang, G.; Guida, W. C. J. Am. Chem. Soc. 1990, 112, 1419-
1
427.
(27) (a) Beveridge, D. L.; Mezei, M.; Mehrotra, P. K.; Marchese, F. T.;
25
that the conformational barriers are lower than kT. This study
does not attempt to evaluate the various methods or compare
them to quenched dynamics; however, there are clearly restric-
tions on which protocols would perform at the level of accuracy
required in a time frame consistent with screening reasonable
numbers of candidate ligands. The requirements we had were
speed, the ability to handle diverse functional groups without
excessive parameter development, and sufficient accuracy to
generate the experimentally observed conformations as among
the lowest energy calculated conformations. Systematic search
methods26 with even fairly coarse dihedral grids are prohibitively
slow since eight significant dihedrals as well as ring flexibility
Thirumalai, V.; Ravi-Shanker, G. Ann. N.Y. Acad. Sci. 1981, 367, 108-
131. (b) Chang, G.; Guida, W. C.; Still, W. C. J. Am. Chem. Soc. 1989,
1
1
1
11, 4379-4386. (c) Kolossvary, I.; Guida, W. C. J. Comput. Chem. 1993,
4, 691-698. (d) Kolossvary, I.; Guida, W. C. J. Am. Chem. Soc. 1993,
15, 2107-19. (e) Liwo, A.; Tempczyk, A.; Oldziej, S.; Shendrovich, M.
D.; Hruby, V. J.; Talluri, S. Biopolymers 1996, 38, 157-175.
(28) Blaney, J. M.; Dixon, J. S. ReV. Comput. Chem. 1994, 5, 299-335.
(29) (a) McDonald, D. Q.; Still, W. C. J. Am. Chem. Soc. 1994, 116,
1
1
1550-11553. (b) McDonald, D. Q.; Still, W. C. J. Org. Chem. 1996, 61,
385-1391.
(30) Head, M. H.; Given, J. A.; Gilson, M. K. J. Phys. Chem. A 1997,
01(8), 1609-1618.
1
(31) (a) Rizo, J.; Koerber, S. C.; Bienstock, R. J.; Rivier, J.; Hagler, A.
T.; Gierasch, L. M. J. Am. Chem. Soc. 1992, 114, 2852-2859. (b) Kitson,
D. H.; Avbelj, F.; Moult, J.; Nguyen, D. T.; Mertz, J. E.; Hadzi, D.; Hagler,
A. T. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 8920-8924. The Biosym
suite of programs is available from MSI, San Diego, CA.
(32) (a) Ross, R. B.; Hockswender, T. R.; Wilson, C. A. Comput. Polym.
Sci. 1995, 5, 203-212. (b) Liang, C.; Yan, L.; Hill, J. R.; Ewig, C. S.;
Stouch, T. R.; Hagler, A. T. J. Comput. Chem. 1995, 16, 883-897. (c)
Lee, C. H.; Zimmerman, S. S. J. Biomol. Struct. Dyn. 1995, 13, 201-218.
(d) Ramek, M. Int. J. Quantum Chem., Quantum Biol. Symp. 1995, 22,
75-81.
(24) Discussions with leading references can be found in (a) Gundertofte,
K.; Liljefors, T.; Norrby, P.-o.; Pettersson, I. J. Comput. Chem. 1996, 17,
4
29-449. (b) Treasurywala, A. M.; Jaeger, E. P.; Peterson, M. J. Comput.
Chem. 1996, 17, 1171-1182. (c) Ngo, J. T.; Karplus, M. J. Am. Chem.
Soc. 1997, 119, 5657-5667.
(25) (a) Cheng, Y. K.; Pettitt, B. M. J. Am. Chem. Soc. 1992, 114, 4465-
4
1
1
473. (b) Burgess, K.; Ho, K.-K.; Pettitt, B. M. J. Am. Chem. Soc. 1994,
16, 799-800. (c) Burgess, K.; Ho, K.-K.; Pal, B. J. Am. Chem. Soc. 1995,
17, 3808-3819. (d) Taber, D. F.; Christos, T. E.; Hodge, C. N. J. Org.
(33) (a) Knopp, B.; Jung, B.; Wortmann, F. J. Macromol. Theory Simul.
1996, 5, 947-956. (b) Rudnicki, W. R.; Lesyng, B. Comput. Chem. 1995,
19, 253-258.
Chem. 1996, 61, 2081-2084.

4574 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
Hodge et al.
Table 2. Effect of Frame Length, k, on Search Efficiency and
Conformational Sampling (m ) 200 Frames, T ) 2500 °C, ꢀ ) 1)
simulation time,
fs per frame
lowest energy
observed, kcal
CPU time,
mina
1
1
000
00
-3.8
-3.8
-3.3
0.19
71
40
38
43
5
2
0
5
a
Four processors on eight processor Challenge, R4400 CPU.
and the dynamics trajectory continued at high temperature from
001th frame, i.e., from the last unminimized conformation. This
1
Figure 1. Numbering of hydrogen and carbon atoms.
was repeated m times, for a total simulation time of k*m. The
2
bis-N-allyl substituted cyclic urea XK216 (1J, Table 1), not a
cant effect on the ranking of conformations by energy; all of
member of the experimental versus calculated comparison set,
was used to determine the effect of parameter variation on
simulation performance as follows.
Effect of Parameter Variation on Low-Energy Search of
XK216. Parameters ꢀ, k, m, and T were examined to select
conditions for optimum efficiency and breadth of conformational
search. The same “random” seed value was used in all
comparative runs.
A search of XK216 with ꢀ ) 1 and T ) 2500 °C was run,
minimizing every 1 ps (k ) 1000). After 21 ps the lowest
energy observed was 3.3 kcal; after 130 ps the lowest energy
was 3.8 kcal. No lower energy structures were observed after
the calculations discussed below were run with ꢀ ) 1, 40, and
8
0, and no significant change in the energetic ordering of
conformations was observed. AM1 calculations (data not
shown) and experimental data (see below) also suggest that
solvent does not play a significant role in determining confor-
mation in this series of compounds.
Effect of Starting Structure on Low-Energy Conforma-
tional Search. Varying the starting structure by selecting
frames randomly from the trajectory file and resubmitting to
the above conditions resulted in identical low-energy conforma-
tions being generated, not necessarily within the same number
of frames but always within the standard 200 ps time. The
longest required time was 186 picoseconds, indicating the
importance of adequate simulation times. These results suggest
that longer simulation times should be employed if structures
are being evaluated that vary significantly from the reference
structures, as opposed to the homologous series reported here.
Inverting the configuration at each of the chiral ring carbons
of XK216, to yield the S,R,R,S ax,eq,eq,ax molecule, and
repeating the simulation also resulted in the set of low-energy
conformations shown above, again suggesting limited depen-
dence of results on initial conditions provided adequate tem-
peratures and simulation times are used. The simulation does
not invert the chiral centers.
4
ns, indicating either complete sampling or repeated searching
of the same regions of conformational space. Inspection of the
high-temperature trajectory files show extreme contortion of
bond angles and lengths, suggesting that the latter possibility
is unlikely. A value of m ) 200 was therefore chosen for
standard searching. Extended searches of several of the
compounds shown in Table 1 also indicated complete search
within 150 ps.
The effect of the simulation time per frame, k, is shown in
Table 2. Based on these results standard searches were run with
k ) 1000 fs to ensure complete search. The last three entries
demonstrate, as expected, that minimization is considerably
slower than the dynamics calculations.
The most efficient search occurred at the highest temperature
examined (2500 °C), but the minimum-energy structure was
located down to 1000 °C within 200 ps. At 600 °C and below,
the minimum energy structure was not found even after 2 ns
simulation. Following this study simulations were run at 2500
Results
X-ray Crystal Structures of Cyclic Ureas. Free Ligands.
To obtain information on the preferred low energy conforma-
tions of cyclic ureas in the solid state, single-crystal X-ray
structures of the compounds shown in Table 1 were obtained
as described above. These were chosen as a congeneric series
that represent variations in ring substitution and stereochemistry
and for which all relevant data were available, including the Ki
values for enzyme inhibition. In the case of the urea with no
nitrogen substituents, 1A, we were unable to obtain an adequate
crystal and report the structure of the acetate derivative 1I, which
according to NMR and calculated structures has a similar
conformation to 1A. The numbering scheme used in the
following discussions is shown in Figure 1.
The crystal structures of 1B-1I are shown in Figure 2. With
the exception of 1I, all of the derivatives crystallize with the
ring conformation observed in the enzyme‚CU complex (shown
in Figure 3). The 4- and 7-benzyl groups project axially from
the chairlike seven-membered ring, and the hydroxyls are
equatorial, except for 1D, in which the stereochemistry of one
hydroxyl is inverted. 1D is also the only structure, including
1I, in which the four atoms of the urea group are out of plane
(169°); the others are within 2-3° of a common plane.
However, to provide the twist angle of the chair conformation,
the O-C(2)-N(1)-C(7) and O-C(2)-N(3)-C(4) dihedrals
°
C; in the case of cyclopropyl-substituted compounds, energy
conservation failures occurred and simulations were carried out
at 1500 °C.
These data point out that very high temperatures are needed
in the quenched dynamics protocol in order to locate confor-
mational minima of cyclic ureas, since temperatures below 1000
°
C were unable to locate the lowest energy structure. In some
cases, simulation temperatures below 1000 °C were insufficient
for chair to chair conformational inversion.
The effect of the dielectric on the conformational search was
examined by running the simulation conditions shown in Table
2, with k ) 1000 fs, and at ꢀ ) 1, 40, 80 and a distant dependent
dielectric, 4*r. The same families of structures were observed
regardless of dielectric value, but the grouping of low-energy
structures differed as well as the absolute energy values. For
the test system described here the “best” clustering, i.e., the
greatest spread between the lowest and those close in energy
was observed with ꢀ ) 80; this was chosen for subsequent
simulations. This choice is consistent with the dielectric of the
aqueous medium in which the assay takes place but less so with
crystallization conditions. However, it does not have a signifi-

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4575
range from 10 to 20° when both nitrogens are substituted. Thus
the exocyclic angles are closer to optimum (a search of the
Cambridge Crystallographic Database shows that the most
common dihedrals are between 0 and 10°) since no relief of
ring strain compensates for the deviation. The same dihedrals
are 8.0 and 8.7° for the substituted and unsubstituted nitrogens,
respectively, of 1E, and are 6.4 and -0.6 for the unsubstituted
1
I.
Thus with the exception of 1I, the effect of the nitrogen
substitution and hydroxyl inversion on ring conformation is
fairly small, and all of the compounds are preorganized in the
manner in which they were originally conceived. As discussed
2
in the original report of cyclic urea HIV protease inhibitors
and is described in more detail in a more recent study,5b the
alternate chair conformation occurs in 1I to place the hydroxyls
axial and the benzylic groups equatorial.
The 4- and 7-benzylic groups of all of the N,N-disubstituted
crystal structures are remarkably uniform in orientation, showing
the clockwise (looking down the OdC bond axis) screw or
propellor shape that is observed on complexation with enzyme.
The phenyl rings of monosubstituted 1E splay outward but still
have the same approximate torsions, while the unsubstituted 1I
projects the phenyls “downward”, away from the carbonyl, a
rotation of 180° relative to all of the other structures. A possible
explanation is that the latter configuration is more favorable
when the hydroxyls are axial, to avoid steric contact with the
benzylic hydrogens; additionally when three or four hydrocarbon
groups are arrayed around the central ring, (presumably)
favorable van der Waals contacts can occur with neighboring
substituents by placing the benzylic groups “upward” and the
urea substituents “downward”.
In fact a closely packed configuration is observed for the side
chains in all of the structures. The urea substituents are tightly
interdigitated with the benzylic groups in all cases but one,
although several different angles are sterically allowed. The
compounds with aromatic rings in the P2/P2′ position show a
preference for a dihedral angle that allows edge to edge, rather
than edge to face, contacts between the aryl groups. The only
exception is the anilinium ring of DMP450, in which the
benzylic torsion angles turn the charged nitrogens upward,
presumably to interact with the extensively hydrogen-bonded
water that is observed within and across the crystalline array
(five ordered waters per unit cell in addition to the two chloride
ions; see Supporting Information). Forces affecting crystal
packing may also play a role in the observed confor-
mationsshydrogen bonds between the hydroxyls and carbonyls
of a neighboring molecule are observed in all of the structures.
However, in nonpolar, polar aprotic, and aqueous solution the
same geometry appears to occur (see below).
NMR Solution Structure. To gain insight into the preferred
aqueous conformation of a water-soluble cyclic urea, spectra
of the bis-methanesulfonate of DMP450 in deuterium oxide were
obtained, and chemical shifts, coupling constants, and NOE
cross-peaks were assigned from 2D NOESY and COSY
experiments (see Supporting Information). The structures
generated from the quenched dynamics simulation, above, were
used to identify conformations consistent with the distances and
angles observed in the NMR spectra. Three coupling constants
and a single NOE were sufficient to unambiguously select a
closely related family of structures and rule out all other
conformations.
Figure 2. Stereoviews of the crystal structures of 1B-1I, viewed
approximately along the axis of the hydroxyl-bearing ring carbons.
ORTEP diagrams are available as Supporting Information.
range from -135 to -149°. That this angle arises in part due
to strain relief in forming the optimum chairlike ring is supported
by the observation that O-C(2)-N(1)-E* and O-C(2)-N(3)-
E* dihedrals, where E* is the exocyclic nitrogen substituent,
The 1D spectrum shows a single resonance for each proton
and its symmetry-related counterpart about the C2 axis (see
Supporting Information), suggesting a symmetrical structure on

4576 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
Hodge et al.
Figure 3. Stereoview of DMP323 in the active site of HIV-1 protease (represented as a ribbon structure). For details on the interactions between
cyclic urea and protein, see refs 2 and 5.
H4 and of 160-180° for H8a-C8-C4-H4.34 Of the 200
conformations generated in the QD analysis described above,
127 were eliminated by these requirements. A conformation
consistent with these requirements (structure 168 in the QD
protocol) is shown in Figure 4. The benzylic proton that is
shifted upfield (H8B in Figure 8) relative to its geminal
counterpart (H8A) also has the larger observed coupling constant,
which supports the assigned structure, since H8B projects into
the deshielding cone of the P2 aromatic ring.
We also assumed that since one of the N-benzyl hydrogens,
H11a or H11b, forms a strong cross-peak with H4, but the other
does not, then a conformation with a distance difference of g1
Å is likely (Figure 4, bottom). This value is proposed since,
for example, the distance between H4 and its vicinal partner H5
can be no more than 2.5 Å, yet it gives rise to a weaker NOE
than it does through space to one of the N-benzylic hydrogens.
The two benzylic protons are in similar chemical surroundings,
so it is reasonable to assume that the lower limit of the
Table 3. Coupling Constants Used in Determining DMP450
Solution Conformation (See Figure 4)
proton
∂, ppm
3.74
3.93
J, Hz
H
4
12.1, 2.0, 1.3
1.3
13.6, 12.1, 2.0
H
H
5
8a, H8b
2.78, 3.03
the NMR time scale. In the following discussion, referring to
a single proton will include its symmetry partner, e.g., referring
to H8a will be understood to mean H8a and H9a. Also, P1 and
P1′ are used to refer to the benzyl groups attached to seven-
membered ring carbons C4 and C7 and P2 and P2′ refer to the
substituents on N1 and N3, respectively. These terms reflect
the protease binding pockets that are occupied by the cyclic
urea substituents in the protein crystal structure.2 The coupling
constants used for conformational assignment are shown in
Table 6. The modified Karplus equation allows a dihedral angle
of between 60 and 90° for H4-C4-C5-H5 and H8a-C8-C4-
Table 4. Comparison of Calculated and Experimental Conformations
RMSDa
entry
R
1
R
2
X
all atomsb
ring + 1 atomsc
1
1
1
1
1
1
1
1
1
B
C
D
E
F
G
H
CH
CH
CH
CH
3
2
2
2
CH
CH
CH
H
3
2
2
(S)-OH
(S)-OH
(R)-OH
(S)-OH
(S)-OH
(S)-OH
(S)-OH
(S)-OH
(R)-OAc
0.28
0.49
0.49
1.7
0.51
3.5
1.7
1.43
3.5
0.13
0.22
0.13
0.13
0.16
0.12
0.11
0.09
0.39
-cyclopropyl
-cyclopropyl
-2-naphthyl
-cyclopropyl
-cyclopropyl
4-fluorobenzyl
CH -2-naphthyl
4-fluorobenzyl
CH -2-naphthyl
2
2
3-aminobenzyl
3-aminobenzyl‚(HCl)
H
3-aminobenzyl
3-aminobenzyl‚(HCl)
H
H‚(HCl)
I
2
2
2
a
RMS difference (Å) between lowest energy calculated structure and crystal structure. b All heavy atoms. c Atoms in ring plus directly attached
heavy atoms.
Table 5. Energy Distribution of Calculated Conformations
b
b
unique, <1 kcal
unique, <5 kcal
total range,a
kcal
4a, 5e, 6e,
7a chair
4e, 5a, 6a,
7e chairc
4a, 5e,
4e, 5a, 6a,
7e chairc
c
6e, 7a chairc
entry
boat
boat
1
1
1
1
1
1
1
1
1
B
C
D
E
F
G
H
20
21
19
32
18
21
23
20
28
1
4
6
2
4
2
5
3
2
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
10
2
10
17
8
12
17
14
13
1
0
0
0
0
0
0
0
0
3
2
4
0
d
12, 4
2
2
0
0
8, 11
H‚(HCl)
I
2
e
a
b
Difference in internal strain energies between highest and lowest energy conformations from QD protocol, kcal. Number of unique conformers
within 1 or 5 kcal of lowest energy observed (see text). a ) axial, e ) equatorial. d P1′ equatorial (12); P1 equatorial (4). Acetate equatorial (8)
acetate axial (11).
c
f

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4577
Figure 4. Top: Calculated structure of 1H‚(HCl)
4
2
consistent with NMR constraints showing the dihedral angle H
4
-C
4
-C
5
-H
).
5
and H8a-C
8
-C -
4
H
. Not printed for clarity is the dihedral angle H8a-C
8
-C -H (174.6°). Bottom: the distances d(H11a-H ) and (H11b-H
4
4
4
4
N-benzylic proton that does not give rise to a cross-peak with
H4 is 3 Å, and the upper limit of the proton that does give rise
to a cross-peak is 2.5 Å. The upper limit proposed is consistent
with recent studies on small molecules;35 the validity of using
unobserved NOEs in conformational assignments is also ad-
dressed in these references. This requirement eliminated 41
more conformations. The remaining conformations have identi-
cal ring geometry, that is, axial/equatorial/equatorial/axial, and
differ only by dihedral angles defined by the benzylic carbons.
These conformations fall into three families, where members
of each family differ from other family members by less than
averages do not appear in the QD analysis, and the observed
low-energy conformations are sufficient to explain the observed
NMR.
Thus the NMR studies are consistent with the average
conformation of DMP450 in water being close to that observed
in the free crystal structure and the lowest energy calculated
structure (see below). Data in organic solvents are not
conclusive but are consistent with the axial,equatorial,equato-
rial,axialconformation for N,N-disubstituted cyclic ureas (see
Experimental Section). The NMR solution structure of DMP323
complexed to HIV-1 PR also displays the axial,equatorial,-
equatorial,axial conformation.
0
.1 Å. All of the structures are consistent with the other
observed cross-peaks in NOESY spectrum, although these peaks
were not necessary for structural assignment. Structure 168 is
most consistent with the pronounced upfield shift observed for
one of the N-benzylic protons (3.30 vs 4.31 ppm for H11a vs
H11b), since it has a strong interaction with the shielding region
of the P1 aromatic ring. Structure 168 also has very similar
ring and P1 dihedrals to those observed in the crystal structure
and has the lowest calculated energy.
It is possible that low-energy conformations not identified
in the QD analysis would also be consistent with the observed
NOEs and coupling constants. Therefore the three angles and
the single distance were used as restraints in a high-temperature
constrained dynamics calculation (not shown); all of the
structures generated minimized to the same families found in
the original calculation. Thus, provided the force field employed
is not grossly inaccurate, it is likely that no other stable
conformations meet the simple NMR constraint requirements.
Other force fields have been used in calculating low-energy
cyclic urea conformations (data not shown) and also yield the
axial,equatorial,equatorial,axial conformation as most stable for
fully substituted cyclic ureas. It is also reasonable to suggest
that the distances and angles used in the assignment represent
averages of values that lie outside the assigned ranges; again,
however, low-energy conformations that would give rise to such
Conformations of Cyclic Ureas Bound to HIV-Protease.
2,5
Published crystallographic analyses have shown that CUs with
large (> three atoms) substituents on both nitrogens bind to
HIV-1 protease with the ring in the axial,equatorial,equatorial,-
axial chair conformations. The NMR solution structure of
DMP323 complexed with HIV-1 PR is consistent with the same
4
average conformation. With the exception of the mono- and
unsubstituted analogues, 1E and 1I, the same ring conformation
is observed in the free ligands. The P1/P1′ benzyl torsion angles
differ slightly, and the P2/P2′ torsions differ significantly in
some cases (1F, 1H‚(HCl)2).
Calculated Conformations. To determine our ability to
computationally generate a priori the experimentally observed
low-energy conformations, the cyclic ureas in Table 1 were
analyzed using the optimized quenched dynamics conditions
described above. A summary of the single lowest-energy
structures and their RMS difference from the experimental
conformations is shown in Table 4. The column denoted
“
Ring+1 RMSD” refers to the RMS deviation of ring atoms as
well as the exocyclic atoms attached to ring atomssi.e., a total
of 14 heavy atomss and “all atoms” refers to all heavy atoms.
The effects of symmetry are taken into account. Figure 5
summarizes the lowest energy conformation of each molecule
observed in the QD simulation.
As described in the methods section, the low-energy confor-
mations are calculated in the gas phase but with a dielectric ꢀ
(
34) Altona, C.; Franke, R.; de Haan, R.; Ippel, J. H.; Daalmans, G. J.;
Hoekzema, W. A. J. A.; van Wijk, J. Magn. Reson. Chem. 1994, 32, 670-
78.
35) (a) Brueschweiler, R.; Blackledge, M.; Ernst, R. R. J. Biomolecular
6
)
80, and no effort is made to reproduce the intermolecular
contacts with solvent or neighboring molecules observed in the
crystal structure. Therefore the calculated and observed struc-
(
NMR 1991, 1, 3-11. (b) Bean, J. W.; Kopple, K. D.; Peischoff, C. D. J. J.
Am. Chem. Soc. 1992, 114, 5328-5334.

4578 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
Hodge et al.
lecular conformers. However, in many cases the structures
match fairly closely, suggesting that crystal packing has limited
effect at least on ring conformational preference (or, less likely,
that inaccuracies in the force field balance out the missing
intermolecular effects).
Lowest Energy Conformation. The low-energy QD-gener-
ated ring conformations for all of the disubstituted cyclic ureas
are almost identical to the crystallographically observed free
ligand conformations (Figure 5 and Table 4). The main
variation between the observed and calculated structures occurs
in the dihedral angles of the nitrogen substituents, and as the
group increases in size, the relative weighting of small changes
increases, causing larger RMSD values for all atoms. Com-
pound 1B has a very small all-atom RMSD, as even the small
methyl group on nitrogen substituents appears to enforce the
P1/P1′ and ring geometry. The bis-N-cyclopropylmethyl sub-
stituted compounds 1C and 1D display four and two distinct
rotamers of equivalent energy in the calculated structures,
respectively; these same atoms show high thermal motion in
the X-ray structures (see Supporting Information), supporting
the validity of several degenerate conformations. A 180°
rotamer of C4-C8 is observed in the lowest 1 kcal only in 1D,
although this variant appears in other calculated structures at
higher energy.
Mono-N-2-naphthylmethyl cyclic urea 1E crystallizes with
the plane of the naphthyl ring tucked inward between the two
phenyl rings (Figure 2), similar to the geometry of P2 benzyl
rings in complex with the HIVPR enzyme (see below), whereas
the two calculated conformations in the lowest 1 kcal include
a twist boat as well as the usual chair. (NMR NOESY
experiments also suggest a twist boat for this compound in
CDCl3: Lim, M. personal communication.) The calculated chair
conformer of 1E has the naphthalene ring turned up and outward
to expose its edge to “solvent” and its face to the neighboring
phenyl ring. All of the calculated structures that contain an
aromatic nitrogen substituent show a preference for this upturned
ring, in which an face-edge, rather than an edge-edge,
interaction is made with the P1 phenyl rings. This orientation
differs from some of the experimental structures and may be in
part an artifact of the force field, which will not capture the
details of π-π interactions. However, one of the benzyl rings
of bis-4-fluorobenzyl derivative 1F, and both of the rings of
the anilinium derivative 1H (but not aniline derivative 1G)
exhibit this preference in the crystal structure, and the solution
structure of 1H appears to be consistent with the calculated and
crystallographic conformation. In all of the bis-P2-aryl (1F,
1G, 1H) cyclic ureas analyzed by quenched dynamics, between
three and five equienergetic benzyl rotamers are observed; in
these cases thermal motion is not observed in the crystal
structures. Two possible explanations come to mind: the force
field does not accurately balance the interaction energies with
the conformational strain energy, or the hydrogen bonding
between neighboring molecules observed crystallographically,
but neglected in the simulation, stabilizes the observed confor-
mation. In crystal structures of cyclic ureas co-complexed with
HIV-1 PR, the hydrogen bonds from amino acid backbone
nitrogens to the cyclic urea carbonyl also clearly favor the
conformation in which the aromatic groups are tucked between
the P1/P1′ phenyl rings and away from steric contact with other
residues (Figure 3). The preferred arrangement of the four
phenyl rings appears to be due to a delicate stereoelectronic
balance that is reproduced in part, but not completely, by the
force field. Thus observed differences between the calculated
and observed benzyl rotamers appear to be due to small
Figure 5. Stereoviews of lowest energy conformation of cyclic ureas
from quenched dynamics simulation viewed approximately along the
axis of the hydroxyl-bearing ring carbons (see also Table 5).
tures are not expected to be identical if these intermolecular
forces are significant and in opposition to preferred intramo-

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4579
inaccuracies in the force field’s treatment of aryl-aryl interac-
tions as well as to neglect of the crystal environment.
N-substituted cyclic ureas with the procedure described here
and selected only the lowest energy conformation, the result
would be a close match with both the free and (where available)
bound crystallographic conformation in all important geometric
parameters except the P2/P2′ benzylic dihedrals, which have
several low-lying minima. Although not the subject of this
paper, these calculated conformations can easily be docked and
minimized in the active site to match the bound ligand geometry
The QD-generated conformations of 1I place the equatorial
alcohols, axial alcohols, and a boat conformer at approximately
equal energy, whereas the crystal structure shows a well-resolved
chair conformer with axial alcohols (Figure 5i). Additionally,
the P1/P1′ benzyl groups in the X-ray structure are oriented
almost 180° from the calculated chair conformer. The QD
trajectory in fact contains the experimentally observed confor-
mation at 123 ps, with a relative energy only 1.6 kcal higher
than lowest energy structure. Unlike the N-substituted cyclic
ureas, the QD trajectory for 1I contains many distinct conform-
ers that are close in energy, due to the increased steric freedom
of the P1/P1′ benzyl groups. NMR in organic solvent shows
coupling constants consistent with either the eq,ax,ax,eq or ax,-
eq,eq,ax chair conformation, but no conclusions on the orienta-
tion of the benzyl dihedral can be drawn (see Experimental
Section). It should also be pointed out that the crystal structure
of 1I shows intermolecular hydrogen bonds between the urea
nitrogens and carbonyl, so the poorer match between gas-phase
calculations and the crystal structure is not surprising.
Conformations Observed in Lowest 5 kcal. Table 5
summarizes the energetic distribution of the two hundred
minimum energy conformations from each simulation, and their
distribution into +1 and +5 kcal energy bands. On average,
each conformation was visited about two times during the 200
ps simulation. The second column indicates the difference
between the highest and lowest energy conformers found in the
simulation. The number of unique conformers in the lowest 1
kcal is shown in the third column. To find unique conforma-
tions, the full set of 200 were ranked by total strain energy and
those within 1 kcal of the minimum were clustered based on
the 14 atoms that make up the ring and their non-hydrogen
attachments. Families were formed of clusters that differed from
each other by more than 0.3 Å RMSD, which grouped ring
conformers into chair, boats, twist boats, etc., uniquely. Within
families, a conformation was considered unique if any of the
side chain dihedrals differed from the others by >10°: that is,
if all of the dihedrals were identical within (10° of another
member of the family, a conformer was discarded; if a single
dihedral differed by >10° (see ref 26c) the conformer was
considered unique. Compounds 1B and 1I therefore have four
dihedrals, compound 1E has six, and the others have eight for
the purpose of this clustering.
1
7
within a few tenths of an angstrom.
Perhaps just as important to both ligand design and the
prediction of binding affinity is the observation that other,
significantly different cyclic urea conformations are very close
in energy to the enzyme-bound conformation, as seen in Table
5 and Figure 5. These results indicate that the initial confor-
2
,5b
mational analyses
are insufficient to allow detailed under-
standing of binding affinity, since the entropic contributions of
other forms would be significant at ambient temperature. This
in turn suggests that further increases in potency may be possible
with increased conformational rigidity.
The observation that the calculated structures as well as the
experimental structures of the free inhibitor in both crystalline
and aqueous environments have stable conformations that are
very close to enzyme-bound conformations supports the notion2
that the potency of cyclic ureas relative to their acyclic
counterparts arises in part from preorganization of the side
chains, the diol, and the water-mimicking carbonyl for inter-
action with the respective enzyme residues.
The quenched dynamics method described here is able to
sample conformational space efficiently and thoroughly for
molecules such as cyclic ureas; although fairly simple structur-
ally, the eight significant dihedral angles and the presence of a
flexible ring cause difficulty in full conformational analysis by
other means. Regarding the use of quenched dynamics as a
method of prescreening putative ligands for the ability to attain
the shape of a reference ligand: a clear advantage is that initially
no simulation of enzyme is carried out; only conformational
searching of the novel ligand and comparison to the reference
ligand are required. The exponential scaling of computational
time with increasing numbers of atoms that is characteristic of
molecular mechanics is thus avoided, and any candidate ligand
for which trustworthy parameters are available can be evaluated
in a matter of minutes. The method obviously suffers from the
fact that regions of the enzyme that do not interact with the
reference ligand are not explored; also, potentially favorable
adjustments in the enzyme active site that could increase
inhibitor affinity are neglected. A more sophisticated analysis
is needed for targets that pass the prescreening requirements. It
should be noted, however, that attempts to even qualitatively
rank order ligands in order of binding affinity using computa-
tional techniques has met with questionable success, even among
With the exception of unsubstituted 1I, all of the conforma-
tions could be classed as pseudo-chair or pseudo-boat; bond
angles in each geometry were similar across all of the CUs
studied. The number of conformers with each ring geometry
is shown in the fourth and fifth columns of Table 5. Only 1E
and 1I have boats among the lowest 1 kcal (Figure 5D and I).
The same analysis was carried out for conformations occurring
within the lowest 5 kcal (columns 6-8 of Table 5). At the
higher energy the boat conformations begin to be more
populatedsinterestingly, the R,S,R,R isomer 1D does not adopt
a boat conformation within this energy band, whereas 1C,
identical except for the stereochemistry of one hydroxyl, has a
number of low-energy boat-side chain combinations.
3
6
highly congeneric series.
Thus in some cases, the crude
assessment provided by matching low energy conformations of
a prospective ligand to the known conformation of a protein-
bound ligand may be the most accurate information that a
chemist will have access to in a useful time frame. It is also
probably safe to say that matching a prospective ligand to a
bound conformation without carrying out a complete confor-
mational analysis will often yield poor results if the ligands are
flexible.
Discussion
For the purposes of molecular design, the value of accurately
reproducing free ligand low-energy conformations lies in being
able to prospectively, as opposed to retrospectively, select
ligands that are able to adopt a complementary conformation
to the active site of a receptor of interest. If one analyzed
(36) (a) Head, R. D.; Smythe, M. L.; Oprea, T. I.; Waller, C. L.; Green,
S. M.; Marshall, G. R. J. Am. Chem. Soc. 1996, 118, 3959-3969. (b)
Reference 20. (c) Kauvar, L. M.; Higgins, D. L.; Villar, H. O.; Sportsman,
J. R.; Engqvist-Goldstein, A.; Bukar, R.; Bauer, K. E.; Dilley, H.; Rocke,
D. M. Chem. Biol. 1995, 2, 107-118. (d) Welch, W.; Ruppert, J.; Jain, A.
N. Chem. Biol. 1996, 3, 449-462.

4580 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
Hodge et al.
This report expands the set of complex organic molecules
and the solution was heated at 70 °C for 5-7 h or until TLC (10%
ethyl acetate/hexane) indicated loss of starting material. The reaction
was cooled, quenched with methanol (1-2 mL), and partitioned
between ethyl acetate (80 mL) and water (70 mL), and the upper layer
removed. The lower layer was washed with ethyl acetate (2 × 30 mL).
The organic extracts were combined and washed with water (2 × 50
that have been searched for conformational minima rapidly and
successfully (if the benchmark for success is finding experi-
mentally observed conformations). Subtle differences with
experiment are noted, particularly in the edge-interactions of
phenyl rings. The calculated lowest energy structure of
nitrogen-unsubstituted cyclic urea 1I is significantly different
from its X-ray crystal structure, but the correct structure does
appear at 1.6 kcal higher in energy, and the discrepancy is not
difficult to rationalize (see Results section). The structures
generated by the quenched dynamics protocol are also useful
as a simple means of identifying conformations consistent with
observed NMR NOEs and coupling constants. Estimation of
the internal strain energy imposed on cyclic ureas after binding
to HIV1-PR will be reported separately.
Finally, these results add to the number of cyclic ureas
reported in the literature for which experimentally determined
free and enzyme-bound conformations and enzyme affinity
constants are known. We have found these data to be useful
in evaluating the performance of various conformational analy-
sis, ligand-protein affinity prediction, and ligand-protein
docking methods developed in our laboratories and elsewhere.
The highly specific binding mode, the absence of unusual
functionality, and the rigidity of these inhibitors make them
particularly tractable for this purpose; these studies as well as
other fully characterized congeneric series of interest will be
reported shortly. Others in the chemistry community may find
these data equally valuable in benchmarking their methods.
4
mL) and brine (50 mL), dried over MgSO , and concentrated to a
residue. The residue was chromatographed on silica gel and eluted
with 10-15% ethyl acetate/hexane to afford the bis-cyclopropylmethyl
1
cyclic urea (SEM-protected) as a colorless oil (710 mg, 86%). H NMR
(300 MHz, CDCl ): 7.51-7.4 (m, 10H, Ph); 4.8-5 (d, 4H, OCH2O);
3
4.08(s, 2H); 3.5-3.9 (m, 14H); 0.8-0.9 (m, 4H). MS: (CDI) 696 (M
+ 1, 100%).
B. Deprotection. The above product was hydrolyzed in methanol
(10 mL) and 1.0 M HCl in ether (10 mL) at room temperature under
2
N for 12 h. The solvent was evaporated, and the residue was
chromatographed with 10-15% ethyl acetate/dichloromethane eluent
to afford 377 mg (86%) bis-alkylated cyclic urea 1C: white solid, mp
2
(
7
)
10-212 °C. NMR (300 MHz, CDCl3): 7.0-7.4 (m, 10H); 4.8-5
q, J ) 7.5 Hz, 4H); 3.9 (s, 2H); 3.8 (m, 2H); 3.6 (m, 4H); 3.5 (q, J )
.5 Hz, 2H); 3.1 (t, J ) 11 Hz, 2H); 3.0 (t, J ) 11 Hz, 2H); 1.9 (q, J
7.5 Hz, 2H); 0.8-1.0 (m, 8H); 0.2-0.4 (m, 4H). MS. (CDI) 435
(M + 1, 100%). HRMS: calcd, 435.2647; found, 435.2636. Anal.
Calcd for C27 : C, 74.62, H, 7.89; N, 6.44. Found: C, 74.46;
34 2 3
H N O
H, 7.81; N, 6.39. Crystals for X-ray analysis from toluene/CH
colorless irregular cubes.
2 2
Cl ,
[4R-(4a,5a,6b,7b)]-Hexahydro-5-hydroxy-6-acetoxy-4,7-bis-
phenylmethyl)-2H-1,3-diazepin-2-one (1A). MEM-protected urea
(
5b
IIB was deprotected using the procedure described above to yield
urea 1A: mp 172-174 °C. MS(CDI) 327(M + 1, 100%). HRMS
calcd: 327.1701; found: 327.1713. Anal. Calcd for C19
9.92; N, 8.58; H, 6.79. Found: C, 69.39; N, 8.47; H, 6.73.
4R-(4a,5a,6b,7b)]Hexahydro-5,6-bis(hydroxy)-1,3-bis[(4-fluo-
rophenyl)methyl)-4,7-bis(phenylmethyl)-2H-1,3-diazepin-2-one
XL472, 1F). Following the two-step general procedure described
22 2 3
H N O : C,
6
Experimental Section
[
Protease Inhibition Assays. Values for inhibition constant, Ki, were
determined with a fluorescent peptide substrate at pH 5.5 with 1.0 M
NaCl as described previously (ref 6a).
(
above, N,N-bis(4-fluorobenzyl) cyclic urea 1F was obtained from MEM-
protected urea IIA using 4-fluorobenzyl bromide as the alkylating agent
to give colorless crystals from butyl ether: mp 133 °C. MS (CDI):
O
5
83.2 (M + H, 99%). HRMS calcd, 583.2772; found: 583.2767. Anal.
Calcd for C36 : C, 74.21; H, 6.23; N, 4.81. Found: C, 73.82;
H, 6.21; N, 4.74.
4R-(4a,5a,6b,7b)]Hexahydro-5,6-bis(hydroxy)-1,3-bis(2-naphth-
HN
RO
NH
OR
36 2 3 2
H N O F
[
ylmethyl)-4,7-bis(phenylmethyl)-2H-1,3-diazepin-2-one (XK263, 1G).
Following the two-step general procedure described above, N,N-bis-
II
IIA: R = [(2-trimethylsilyl)methoxy]ethoxy (SEM)
IIB: R = 2-methoxyethoxy (MEM)
(2-naphthylmethyl) cyclic urea 1G was obtained from SEM-protected
urea IIA using 2-bromomethylnaphthalene as the alkylating agent:
white solid, 202-204 °C. NMR (300 MHz, CDCl3): 7.0-8.0 (m,
14H); 5.1 (d, 2H); 3.6 (m, 4H, CH); 3.2 (d, 2H); 3.1 (d, 4H); 2.2 (s,
2H). MS (CDI) m/z 607 (M + 1, 100%). HRMS calcd, 607.2961;
Chemistry. All procedures were carried out under inert gas in oven-
dried glassware unless otherwise indicated. Proton NMR spectra were
obtained on VXR or Unity 300 or 400 MHz instruments (Varian
Instruments, Palo Alto) with TMS as an internal reference standard.
Melting points were determined on a Mettler SP61 apparatus and are
uncorrected. Elemental analyses were performed by Quantitative
Technologies, Inc., Bound Brook, NJ. High-resolution mass spectra
were carried out on a VG 70-VSE instrument with NH3 chemical
ionization. Thin layer and column chromatography were carried out
on plates or silica gel from E. Merck, Darmstadt, FRG. Separation of
optical isomers was performed using supercritical fluid chromatography
with a Chiracel OD (Daicel Chemical Ind. Ltd.) and 20% methanol
modified CO2 mobile phase. Optical rotations were obtained on a
Perkin-Elmer 241 polarimeter. Solvents and reagents were obtained
from commercial vendors in the appropriate grade and used without
further purification unless otherwise indicated.
found, 607.2960. Anal. Calcd for C41
4.62. Found: C, 80.85; H, 6.20; N, 4.54. Crystals for X-ray analysis
from CH OH, colorless hexagonal rods.
38 2 3
H N O : C, 81.16; H, 6.31; N,
3
[4R-(4a,5a,6b,7b)]Hexahydro-5,6-bis(hydroxy)-1-(2-naphthyl-
methyl)-4,7-bis(phenylmethyl)-2H-1,3-diazepin-2-one (XK291, 1E).
MEM-protected urea IIB (3.55 g, 7.06 mmol) in 20 mL DMF of was
added to a flask containing NaH (60% in oil) (1.78 g, 44.5 mmol) that
had been washed with dry hexane (2 × 10 mL) in 8 mL of DMF. The
mixture was stirred at room temperature under N2 for 10 min.
2-Bromomethylnaphthalene (2.01 g, 8.9 mmol) was added, and the
solution was heated to 40 °C for 4h. The solution was quenched with
methanol (5-8 mL), partitioned between ethyl acetate (180 mL), and
water (170 mL) and the organic layer was removed. The aqueous layer
was washed with ethyl acetate (2 × 80 mL). The organic extracts
were combined and washed with water (2 × 80 mL) and brine (80
[
4R-(4a,5a,6b,7b)]Hexahydro-5,6-bis(hydroxy)-1,3-bis(cyclo-
propylmethyl)-4,7-bis(phenylmethyl)2H-1,3-diazepin-2-one (1C). A.
4
mL), dried over MgSO , and concentrated to a residue. The residue
5
a
Alkylation. SEM-protected urea IIA (700 mg, 1.19 mmol) in 2 mL
of DMF was added to a flask containing NaH (11.9 mmol) that had
been washed with dry hexane (2 × 10 mL) in 8 mL of DMF. The
was chromatographed on silica gel and eluted with 10-40% ethyl
acetate/hexane to separate the protected bis-substituted (1.4 g, 25%)
and monosubstituted (0.6 g, 14%) cyclic ureas as colorless oils. The
monosubstituted naphthalene (90 mg, 0.14 mmol) was hydrolyzed in
methanol (10 mL) and 4.0 M HCl in dioxane (10 mL) by stirring at
2
mixture was stirred at room temperature under N for 10 min.
Bromomethylcyclopropane (Aldrich, 0.81 mL, 9.33 mmol) was added,

Cyclic Urea HIV Protease Inhibitors
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4581
room temperature under N
and the residue was chromatographed on silica gel and eluted with
0% ethyl acetate/dichloromethane to afford mono-N-naphthylmethyl
2
for 12 h. The solvent was evaporated,
Supporting Information Available: X-ray crystal data,
descriptions of data collection, treatment, solution, and refine-
ment and tables of fractional coordinates, isotropic and aniso-
tropic thermal parameters for compounds 1B-1I; input file for
4
1
cyclic urea 1E: white solid, 49.5 mg, (76%); mp 210-213 °C.
H
NMR (300 MHz, CDCl ): 7.2-7.8 (m, 17H); 5.1 (d, J ) 15 Hz, 2H);
3
1
13
the quenched dynamics protocol; and H, C, hsqc and COSY
3
1
.9 (m, 1H); 3.6 (m, 2H); 3.4 (m, 1H); 3.1 (m, 4H); 2.9 (d, J ) 15 Hz,
H); 2.7 (br s, 1H); 2.4 (br s, 1H). MS (CDI) 467 (M + 1, 100%).
1
spectra of the bis-methanesulfonate salt of H (79 pages, print/
HRMS calcd: 467.2335; found: 467.2330. Anal. Calcd for
PDF). See any current masthead page for ordering and web
access instructions.
1
C
30
H N O
30 2 3
‚ /
.30; N, 5.82. Crystals for X-ray analysis from CH
colorless parallelepipeds.
H
2 2
O: C, 75.76; H, 6.57; N, 5.89. Found: C, 75.68; H,
6
3
OH/hexane,
JA972357H