Computer-assisted design of selective imidazole inhibitors for cytochrome P450 enzymes

Article abstract of DOI:10.1021/jm030608t

A modified version of the DOCK program has been used to predict inhibitors for cytochrome P450cam and its L244A mutant. A library of azole compounds was designed in silico and screened for binding to wild-type P450cam. Lead compounds were synthesized and found to inhibit wild-type P450cam. To test our approach to designing ligands that discriminate between closely related sites, the azole library was DOCKed into both the active sites of wild-type P450cam and its L244A mutant. The L244A active site is predicted to be slightly larger than that of wild-type P450cam. Ligands predicted to have a high affinity for the mutant alone were synthesized and assayed with the recombinant enzymes. All of the compounds showed inhibition of the L244A enzyme (IC₅₀ = 6-40 μM), and the compounds that were predicted to be too large to bind to the wild-type showed poor inhibition (IC₅₀ ≥ mM). The binding mode was shown to be similar to that predicted by our modified version of DOCK by spectroscopic analysis. A discrepancy between the IC₅₀ values and spectroscopic K_s values indicates that the spectroscopic binding constants do not accurately estimate inhibitory activity. This study, the first report of computer-assisted ligand (drug) design for P450 enzymes in which the coordination bond between imidazole and the heme is explicitly considered in structural modeling, opens a promising design avenue because azole compounds are widely used as P450 enzyme inhibitors and drugs.

Full text of DOI:10.1021/jm030608t

3572
J . Med. Chem. 2004, 47, 3572-3579
Com p u ter -Assisted Design of Selective Im id a zole In h ibitor s for Cytoch r om e
P 450 En zym es
Andreas Verras, Irwin D. Kuntz, and Paul R. Ortiz de Montellano*
Department of Pharmaceutical Chemistry, University of California, 600 16th Street, San Francisco, California 94143-2280
Received December 12, 2003
A modified version of the DOCK program has been used to predict inhibitors for cytochrome
P450cam and its L244A mutant. A library of azole compounds was designed in silico and
screened for binding to wild-type P450cam. Lead compounds were synthesized and found to
inhibit wild-type P450cam. To test our approach to designing ligands that discriminate between
closely related sites, the azole library was DOCKed into both the active sites of wild-type
P450cam and its L244A mutant. The L244A active site is predicted to be slightly larger than
that of wild-type P450cam. Ligands predicted to have a high affinity for the mutant alone
were synthesized and assayed with the recombinant enzymes. All of the compounds showed
inhibition of the L244A enzyme (IC50 ) 6-40 µM), and the compounds that were predicted to
be too large to bind to the wild-type showed poor inhibition (IC50 g 1 mM). The binding mode
was shown to be similar to that predicted by our modified version of DOCK by spectroscopic
analysis. A discrepancy between the IC50 values and spectroscopic Ks values indicates that the
spectroscopic binding constants do not accurately estimate inhibitory activity. This study, the
first report of computer-assisted ligand (drug) design for P450 enzymes in which the coordination
bond between imidazole and the heme is explicitly considered in structural modeling, opens a
promising design avenue because azole compounds are widely used as P450 enzyme inhibitors
and drugs.
In tr od u ction
structural data, computational methods that exploit this
information and methods that are easily translatable
across various P450 enzymes must be developed.
We report application of computational design meth-
ods for the generation of high-affinity inhibitors for
cytochrome P450cam. To determine the ability of the
approach to discriminate between closely related active
sites, we have expressed the L244A mutant, which only
differs from the parent enzyme by three carbon atoms,
and then generated inhibitors that preferentially target
the mutant. While P450cam is a bacterial enzyme of
little interest as a drug target, it has been well char-
acterized and represents an ideal system in which to
test and refine the necessary approaches. The prediction
of inhibitors for P450 enzymes presents a particular
challenge because of the relatively low specificity of
P450 active sites. In previous efforts, we have made
some progress in predicting the substrate specificity of
P450cam.8 However, this is the first instance in which
such methods have been applied to the design of
potential inhibitors of a structurally characterized P450
enzyme.
Cytochrome P450 enzymes, which catalyze the mono-
oxygenation of a broad diversity of substrates, play
essential roles in both the metabolism of xenobiotics and
the biosynthesis and catabolism of endogenous lipophilic
factors.1 Inhibition of P450 enzymes, and therefore of
the synthesis of endogenous factors, is a promising
avenue for the development of therapeutic agents. Two
examples are the inhibition of estradiol synthesis by
aromatase in estrogen-dependent breast cancer and the
possible control of retinoic acid levels through inhibition
of P450-dependent retinoic acid metabolism.2,3 The
inhibition of P450-dependent xenobiotic metabolism is
also potentially desirable in situations in which xeno-
biotics are converted to toxic or carcinogenic metabo-
lites.4 In a different context, the rapid selection of drug
candidates that minimize the potential for P450-based
drug-drug interactions is becoming increasingly im-
portant in the pharmaceutical industry.
Conventional drug design approaches have led to the
development of a range of P450 inhibitors that are
currently used in the clinic, primarily as antifungal
agents. These agents include miconazole, ketoconazole,
and fluconazole. It is likely that structure-based drug
design methods will soon be applicable to the design of
P450-targeted agents, as X-ray crystallography is begin-
ning to yield structural information on relevant en-
zymes. In recent years, the structures of Mycobacterium
tuberculosis sterol 14-demethylase, membrane-bound
mammalian P450 enzymes, and even human P450
enzymes have been reported.5-7 With this influx of
Resu lts a n d Discu ssion
Vir tu a l Libr a r y Design a n d a Mod ified DOCK-
in g Algor ith m . An initial virtual library of ligands was
generated to screen for potential inhibitors for cyto-
chrome P450cam. The choice of an imidazole scaffold
was based on the knowledge that imidazoles and other
aromatic nitrogen heterocycles that coordinate to the
iron offer an effective route to inhibition of P450
enzymes.9 Past research into heme binding imidazoles
indicates that substitution is limited to two positions.10
Thus, variation of the substituents at the 1 and 5
* To whom correspondence should be addressed. Phone: (415) 502-
4728. Fax: (415) 502-4728. E-mail, ortiz@cgl.ucsf.edu.
10.1021/jm030608t CCC: $27.50 © 2004 American Chemical Society
Published on Web 05/29/2004

P450 Imidazole Ligand Design
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14 3573
F igu r e 2. CovDOCK imidazole orientations, showing the
number 1 (blue), 50 (green), and 150 (red) ranked compounds
when CovDOCKed against wild-type P450cam. The side chains
have been stripped to better illustrate the placement of the
imidazole relative to the heme.
F igu r e 1. Covalent DOCK parameters. R is the distance from
the heme iron to the imidazole N3. Θ1 and Θ2 are the angles
as measured from nitrogens on the heme to the heme iron to
imidazole N3. Light-faced atoms have their van der Waals
contributions negated.
distance. This further decreases the effective size of the
ligands and their overall CovDOCK score. A breakdown
of the score contributions taken as absolute values to
show their effective contributions to the overall Cov-
DOCK score is illustrated in Figure 3. CovDOCK is able
to return expected binding modes but does not dominate
the score while doing so.
positions allows for the introduction of active site
contacts and isoform specificity, whereas substitutions
at the 2 or 4 positions introduce steric clashes with the
heme moiety that preclude coordination of the imidazole
to the heme iron atom. The nitrogen at position 3 of the
imidazole must remain unsubstituted because it is the
atom that actually coordinates to the iron (Figure 1).
Substituents for the virtual library were taken from the
ACD as described in the Experimental Section. A
monosubstituted imidazole library was then constructed
with the substituent exclusively in the 1N position
because previous work suggested that a penalty exists
for desolvating the unsubstituted 1N position.11 How-
ever, the greatest library diversity was provided by the
1,5-disubstituted structures, which comprised 2550 of
the total of 3508 structures.
Because DOCK does not recognize the coordinate
imidazole-iron bond, initial attempts to DOCK the
library produced unreasonable inhibitor orientations.
Even when the ligand was preoriented correctly, the
correct orientation was lost during minimization. To
overcome this problem, Covalent DOCK (CovDOCK)
was used. By introduction of a penalty for long imidazole
N3-heme iron distances or improper angles, a binding
mode was achieved that more correctly mimicked the
binding observed in crystal structures of imidazole-
P450 complexes (Figure 2). The binding of imidazole-
P450 complexes diverges to some extent from any fixed
distance and angle,11-13 so the harmonic restraints were
deliberately set to low values to allow for this variation.
The best ligand among the initial hits had a Cov-
DOCK score of -15.9, compared to the value of -30
expected for a strong ligand hit. The relatively low
scores, however, were not due to the harmonic penalties
used in CovDOCK but were a consequence of the small
active site and the lack of electrostatic interactions. It
is also to be noted that the van der Waals’ interactions
of three imidazole atoms (C2, N3, and C4) are ignored
in order to allow for a quasi-covalent imidazole-iron
In h ibition of P 450ca m . As might be expected, the
hits pulled from the library were primarily small
ligands. The 200 top-scoring hits were clustered by
Daylight fingerprints (Daylight Company, Santa Fe,
NM), which effectively arranges them by functionality
and not by 3D shape or conformation. A preference for
monosubstituted imidazoles is immediately apparent in
the preferred structure list despite the fact that mono-
substituted ligands represent less than one-third of the
total virtual library. Representatives of several attrac-
tive clusters from among the top 200 ranked ligands
were selected for experimental evaluation and were
synthesized. Ligands were chosen for synthesis by
manually inspecting the top 200 ranked compounds and
selecting ligands that had good binding modes and
maintained an imidazole-heme coordinate character
similar to that found in the crystallographic data. The
ability of the compounds to inhibit the oxidation of
camphor to 5-exo-hydroxycamphor was measured by gas
chromatographic analysis of the product to starting
material ratio. All of the monosubstituted compounds
showed good inhibition of wild-type P450cam; most were
in the low micromolar range (Table 1). The monosub-
stituted compounds were then assayed against the
L244A mutant, and the results, as expected, indicated
a preference for the wild-type. The results are shown
on a log scale to better illustrate the differences in
inhibitory activity between the wild-type and mutant
(Figure 4). There is more space created by the L244A
mutation that is not exploited by the monosubstituted
compounds. While being able to generate high-affinity
inhibitors in a first round synthesis is promising in

3574 J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14
Verras et al.
F igu r e 3. Covalent DOCK score breakdown indicated by Covalent DOCK score contribution to the overall DOCK score. The
absolute values of the score components are displayed to show relative magnitude.
F igu r e 4. Monosubstituted IC50 selectivity indicated by monosubstituted compound IC50 values assayed against the wild-type
and the L244A mutant.
Ta ble 1. Monosubstituted IC50 Values and CovDOCK Scoresa
P450cam. The L244A mutation was made in the 1PHD
PDB structure using Sybyl, and a grid was generated
for CovDOCKing without minimization. The original
virtual library was then covalently DOCKed into this
L244A active site. In an effort to generate selective
inhibitors, the best hits against the mutant were taken
and CovDOCKed into wild-type P450cam. Large differ-
ences in the CovDOCK score (>50) were assumed to
represent active site clashes and thus a potential poor
fit. By comparing the CovDOCK results between the two
active sites, we hoped to find the characteristics of
ligands that would inhibit the L244A isozyme selec-
tively. Gschwend and Kuntz14 used a similar approach
before, predicting selectivity by computational screening
between two strains of dihydrofolate reductase. Table
2 illustrates the scores for several of our compounds
against the L244A mutant and the differences between
a
Monosubstituted ligands synthesized and their IC50 values and
the scores for the two forms of the enzyme. A major
difference between compounds predicted to be inhibitors
CovDOCK scores against wild-type P450cam.
for the L244A vs the wild-type is quickly apparent in
the second substituent on the imidazole at the 5
position. While these disubstituted compounds comprise
two-thirds of the library, they were not predicted to fit
into the wild-type, and when they were redocked into
the wild-type, they were found to have high scores due
itself, we were particularly interested in generating
compounds selective for the L244A mutant by rational
design.
We used CovDOCK to search for inhibitors predicted
to exhibit a high affinity for the L244A mutant and,
simultaneously, a mitigated affinity for wild-type

P450 Imidazole Ligand Design
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14 3575
Ta ble 2. Second-Round CovDOCK Resultsa
F igu r e 5. Difference spectrum of 1-cyclopentyl-5-propylimi-
dazole. This type II shift is seen with all the inhibitors.
a
Disubstituted compounds rank against the L244A mutant as
Sp ectr oscop ic An a lysis. Binding of the compounds
was evaluated spectrophotometrically. To ensure a
similar starting state, a difference spectrum was taken
in which inhibitors were titrated into solution in the
absence of camphor. All of the inhibitors exhibited a type
II spectral shift in which the water-coordinated Soret
maximum at 417 nm disappeared while the 424 nm
imidazole-bound peak intensified. This shift is consis-
tent with binding of a primary or aromatic nitrogen as
the sixth ligand to the iron.15 Figure 5, which illustrates
the difference spectra obtained with 1-cyclopentyl-5-
propylimidazole, is typical of all of the spectra used to
determine the Ks values. A clear isosbestic point is
observed, and the transition to 424 nm is saturatable.
The spectroscopic data agree with the binding mode
predicted by CovDOCK.
scored by CovDOCK and the difference in score when the same
compound is covalently DOCKed back into the wild-type enzyme.
Ta ble 3. Disubstituted IC50 Valuesa
The L244A P450cam is very similar to wild-type
P450cam spectroscopically: both have a 450 nm peak
in the reduced CO-bound state, and both enzymes
absorb in the ferric state at the same wavelength, both
in the high spin (390 nm, no sixth ligand) and low spin
(416 nm, water-coordinated) states. Furthermore, the
affinity for camphor, the natural substrate, remains
unchanged (wild-type Ks ) 1.3 ( 0.2 µM, L244A Ks )
1.5 ( 0.2 µM). The L244A mutant, as previously
reported, is catalytically viable and, though it turns over
camphor more slowly, exhibits a degree of uncoupling
nearly identical to that of the wild-type enzyme.16 This
means that the NADH and O2 consumed by the enzyme
are used to the same extent in hydroxylation of camphor
rather than in the production of H2O2 or H2O. Never-
theless, an important difference was found in the
affinity of the two forms of the enzyme for imidazole.
The Ks of wild-type P450cam for unsubstituted imida-
zole is 7.5 ( 1.4 µM, whereas that of the L244A mutant
is 304 ( 49 µM. This 43-fold difference was unexpected
because the spectroscopic, substrate binding, catalytic,
and uncoupling data suggest that the wild-type and
mutant active sites are very similar. The loss of affinity
for imidazole presumably reflects active site changes
such as an increase in the active site water content,
possible changes in the positioning of water molecules,
and small perturbations of active site residues that do
not alter the protein spectrum or its natural affinity for
camphor. Reports of azole resistance among fungal
strains are increasing,17,18 and the L244A mutant may
provide a useful model in this regard. Furthermore, the
data indicate that our disubstituted inhibitors are able
to compensate for the imidazole scaffold’s mitigated
affinity for the L244A active site.
a
Disubstituted compounds synthesized and their inhibition of
L244A and wild-type P450cam enzymes.
to steric clashes. Notably, the second substituent on the
ligands predicted to bind to the L244A is generally
small, in agreement with the small active site change
due to mutating a leucine to an alanine. On the basis
of these results, the disubstituted compounds were
predicted to be selective inhibitors and five ligands were
chosen for synthesis. The criteria for selection included
good imidazole-heme coordination, a good CovDOCK
score against the L244A mutant, and a poor score
against the wild-type.
The disubstituted compounds, as predicted, have good
affinity for the L244A protein (Table 3). When assayed
against the wild-type, they show almost no inhibition,
and even at a 1 mM concentration none of the disub-
stituted compounds completely inhibited catalytic activ-
ity. These results are very promising because they
suggest that we were able to predictively identify
ligands complementary to the L244A active site but
slightly too large to be able to inhibit the wild-type.

3576 J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14
Verras et al.
F igu r e 6. Monosubstituted compound Ks values against wild-type P450cam and its L244A mutant.
The exact nature of the conformational changes, and
the reason for the low inhibitory activity of some
compounds that have low spectroscopic binding con-
stants (Ks values), remain to be defined. One possibility
is that the large inhibitors require a slow conformational
change and are thus slow-binding, reversible inhibitors
that adhere to the following kinetic model:19
E + I f EI f EI*
However, we have not observed the time dependence
in the Ks value determinations implied by this model.
Thus, although conformational changes are clearly
involved in the discrepancy between the Ks and IC50
values, the definition of the precise relationship between
these values requires further exploration.
F igu r e 7. Disubstituted compound Ks values against wild-
type P450cam and its L244A mutant.
Con clu sion s
We have shown that it is possible, using structure-
based design and currently available docking algo-
rithms, to generate ligands of high affinity for both
cytochrome P450cam and, more importantly, its L244A
mutant. The CovDOCK method and azole libraries offer
promise as a facile, modular system easily applicable
to crystallographically characterized P450 enzymes. The
library construction was based on synthetic feasibility
and is not limited to the 3500 compounds actually
included in this study. Indeed, the library has recently
been expanded to 4500 compounds and is completely
amenable to the inclusion of alternative P450-coordinat-
ing scaffolds such as triazoles, pyridines, and other
heterocycles, all of which would significantly increase
the chemical diversity of the library.
On the basis of the IC50 data, we were able to exploit
small differences in the active sites of these two
isozymes and synthesize two distinct groups of ligands
with selectivity for either enzyme. The results are
promising, and with the accelerating pace of determi-
nation of P450 crystal structures and the swift screening
methods made possible by computational advances, one
can envision that dozens, even scores, of enzymes could
be DOCKed in an effort to generate isoform selectivity.
Although P450cam can hydroxylate a variety of non-
natural substrates,15 the enzyme has a single natural
substrate (camphor) and has clearly evolved to optimize
its oxidation. P450cam has generally been thought to
have a relatively rigid active site because crystal
structures of the enzyme with various substrates and
inhibitors show only minor adjustments of the active
The Ks values for the monosubstituted compounds
against both the L244A and wild-type P450 enzymes
are shown in Figure 6. A general trend toward selectiv-
ity is evident and expected given the difference in base
affinities of the scaffolds. Surprisingly, the Ks values for
the disubstituted compounds do not reflect the high
selectivity observed in the IC50 assays (Figure 7). While
the IC50 assays are done in the presence of camphor (250
µM), it is still surprising that such a large discrepancy
exists. In theory, if the disubstituted compounds are
able to coordinate to the iron, they should prevent
electron transfer and catalysis; however, even at inhibi-
tor concentrations up to 1 mM, activity is still seen in
the presence of the disubstituted ligands.
While exploration of the intricacies of this anomaly
is outside the scope of this paper, it is clear that this
observation is related to a requirement of conforma-
tional changes in the P450cam active site. Support for
this argument is provided by measurements of Ks values
of ligands predicted to be too large to fit into either
active site. Thus, 1,5-dicyclohexylimidazole binds to both
the wild-type (90 ( 55) and the L244A mutant (20 ( 4)
even though, on the basis of a rigid model of the protein,
this large, disubstituted compound should be unable to
bind to either the wild-type or L244A active site. Even
more surprisingly ketoconazole was found to bind with
high affinity to both the wild-type (0.5 ( 0.1) and L244A
mutant (0.6 ( 0.1). No crystal structure of P450cam
exists that suggests an active site anywhere large
enough to accommodate this ligand. Thus, a conforma-
tional change must occur.

P450 Imidazole Ligand Design
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14 3577
site cavity.11,20,21 Of course, it is evident that the protein
must undergo substantial conformational changes to
allow the entry of substrates into the active site cavity
because no entry channel is seen in the crystal struc-
ture. The spectroscopic results demonstrate that con-
formational mobility, even in an enzyme such as
P450cam, is an important parameter in identifying
ligands that differentiate between two similar P450
active sites. Also, while spectroscopic assays are fre-
quently the initial test of binding to cytochrome P450
enzymes, our results suggest that they can be mislead-
ing in inhibitor prediction. Studies measuring spectro-
scopic binding to microsomal P450 enzymes and in vivo
activity have shown discrepancies between Ks and IC50
values.22 Our results suggest that discrepancies may
also exist in single enzyme studies and that caution
must be employed when screening ligands for inhibitory
potential by spectroscopic means.
280 °C over 20 min was used. The â-thujone standard eluted
at approximately 6 min, starting material eluted at approxi-
mately 6.2 min, and product appeared at 9.9 min. The
difference between the product to starting material ratio was
then used to calculate turnover. IC50 values were calculated
by the Langmuir isotherm equation:25
IC50
Activity ) 1 + 1 +
(
)
[Inh]
Or ga n ic Syn th esis. Two rounds of synthesis were done.
Initially all the imidazoles synthesized were N-monoalkylated
and were prepared as previously reported.26 The syntheses of
N-benzylimidazole and N-cyclopropylmethylimidazole are de-
scribed here as typical procedures.
N-Ben zylim id a zole. Imidazole (100 mg, 1.4 mmol) was
suspended in 5 mL of acetone under argon in a round-bottom
flask. Dry, powdered KOH (392.7 mg, 7 mmol) was added with
stirring. The solution became beige. After 3 min a 190 µL
aliquot of benzyl bromide (1.57 mmol) was added dropwise,
and a white precipitate appeared. The acetone was removed
on a rotary evaporator, and the precipitate was resuspended
in CH2Cl2. Water was added, the solution pH was lowered to
∼4 with concentrated HCl, and the product was extracted in
the aqueous layer. The pH of the aqueous layer was then
raised to ∼10 with concentrated NaOH, and the product was
extracted with CH2Cl2 and crystallized directly from CH2Cl2.
The purity and identity were judged by TLC (ethyl acetate +
1% triethylamine, Rf ) 0.1), mass spectrometry, and NMR:
17% yield; mp 73-74 °C (reported 72-74 °C); 1H NMR (CDCl3)
δ 7.52 (1H, s, imidazole), 7.08 (1H, s, imidazole), 6.88 (1H, s,
imidazole), 5.18 (2H, s, NCH2), 7.14 benzyl (2H, d, aryl, J )
8.0 Hz), 7.33 (2H, m, aryl), and 7.32 (1H, m,aryl); 13C NMR
(DMSO) δ 137.4, 136.18, 129.83, 128.99, 128.27, 127.28, and
50.8; EI-MS, m/z (low resoln), 159.25 (M + H).
Exp er im en ta l Section
P r otein Exp r ession a n d Bin d in g Assa ys. Cytochrome
P450cam and the P450cam L244A mutant were expressed in
Escherichia coli as previously reported.23 Both proteins were
stored with 100 µM camphor at -70 °C. Camphor was removed
before the proteins were used by twice eluting with 100 mM
sodium phosphate through a PD-10 size exclusion column. The
P450cam thus obtained has a water coordinated to the heme
iron and is in the low-spin state. Spectroscopic binding (Ks)
assays were done in 100 mM potassium phosphate buffer at
37 °C by difference UV spectroscopy. The two cuvettes
contained 0.5 µM protein in 300 µL of 100 mM potassium
phosphate buffer. The potential inhibitors were dissolved in
DMSO with the exception of imidazole, which was dissolved
in water. DMSO never constituted more than 2% of the final
sample volume. An equal amount of DMSO was added to the
reference cuvette, and the shift in absorbance from 416 to 424
nm as increasing concentrations of camphor were titrated into
the cuvette was observed. The difference between 416 and 424
nm values was then recorded and fitted to a binding curve to
obtain Ks values. Ks values were calculated by nonlinear
regression analysis using the equation
1-Meth ylcyclop r op ylim id a zole. Imidazole (100 mg, 1.4
mmol) was suspended and deprotonated as described above.
A 230 µL aliquot of bromomethylcyclopropane (1.57 mmol) was
added dropwise, and the solution became a darker shade of
beige. The reaction was worked up as described above and,
after an acid-base wash, the solution was concentrated by
rotary evaporation to give a yellow oil that was deemed
sufficiently pure by TLC (ethyl acetate + 1% triethylamine,
Rf ) 0.1), mass spectrometry, and NMR for direct study: 50%
1
yield; H NMR (CDCl3) δ 7.55 (1H, s, imidazole), 7.06 (1H, s,
∆Amax[I]
imidazole), 6.99 (1H, s, imidazole), 3.79 (2H, d, NCH2, J ) 6.8
Hz), 1.2 (1H, m, cyclopropyl), 0.68 ppm (2H, q, cyclopropyl, J
) 8.0 Hz), and 0.36 (2H, q, cyclopropyl, J ) 4.2 Hz); 13C NMR
(DMSO) δ 137.37, 128.62, 119.75, 50.85, 12.58, 4.08; EI-MS,
m/z, exptl 122.0843 (M+) (calcd 122.0844, C7H10N2).
∆A )
Ks + [I]
where ∆A is the 416 trough to 424 peak absorbance difference,
∆Amax is the difference in absorption at saturation, and [I] is
the inhibitor or ligand concentration.24
The compounds reported in this study were synthesized
except for imidazole, 1,5-dicyclohexylimidazole, and ketocona-
zole. Imidazole and 1,5-dicyclohexylimidazole were purchased
from Aldrich (St. Louis, MO), and ketoconazole was obtained
from J ansen Pharmaceuticals (Beerse, Belgium).
Ga s Ch r om a togr a p h ic In h ibition Assa y a n d IC50 Ca l-
cu la tion . Catalytic activity was quantitated by gas chromato-
graphic product analysis. The protein was eluted twice through
a PD-10 column with 100 mM sodium phosphate to remove
camphor. Assays were done under the following conditions:
P450cam 5 nM, putidaredoxin 2 µM, putidaredoxin reductase
2 µM, camphor 250 µM, and NADH 1000 µM. The L244A
protein was assayed at a concentration of 20 nM. The total
reaction volume was 150 µL. Varying amounts of inhibitor
were added in DMSO, but the DMSO volume never exceeded
3% of the total reaction volume. All assays were done in 100
mM potassium phosphate in a 20 °C bath for 15 min. The
reactions were then quenched with 20 µL of CH2Cl2 containing
2 mM â-thujone as an internal standard.
The second round of ligands prepared as potential inhibitors
of the P450cam L244A mutant were all 1,5-disubstituted
compounds. These were prepared by a two-step synthesis
utilizing the synthon tosylmethyl isocyanide.27 The yields were
generally low, less than 10%, but the reactions afforded
sufficient product for the spectroscopic and kinetic assays. The
synthesis of 1-benzyl-5-cyclopropylimidazole is given as a
typical procedure.
1-Ben zyl-5-cyclop r op ylim id a zole. To ∼2 mL of dry mo-
lecular sieves was added 10 mL of anhydrous CH2Cl2, followed
by benzylamine (1.46 mL, 13.1 mmol) and subsequently by
dropwise addition of cyclopropylcarboxaldehyde (1 mL, 13.1
mmol). The disappearance of the aldehyde peak and appear-
ance of the imine peak were monitored by infrared spectros-
copy. After 2 h, the reaction mixture was filtered and rotary-
evaporated to give a clear-yellow product that was used as
obtained in the following step.
Dry potassium carbonate (9.55 g, 50 mmol) was added to
tosylmethyl isocyanide (1.99 g, 10 mmol) suspended in 15 mL
of anhydrous MeOH followed by dropwise addition of the imine
product from the previous reaction. The resulting solution was
allowed to stir for 24 h and was then filtered, the filtrate was
subjected to rotary evaporation, and the residue was extracted
thrice with ether (3 × 20 mL). The product was purified by
The CH2Cl2 was then withdrawn directly from the reaction
mixture and injected onto an HP 5980 series II gas chromato-
graph equipped with an Agilent Technologies (Palo Alto, CA)
DB-10 capillary column. A temperature gradient from 40 to

3578 J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14
Verras et al.
flash chromatography on silica with a solvent system consist-
ing initially of 300 mL of 50/50 (v/v) of hexane/ethyl acetate
followed by 80/20 (v/v) of ethyl acetate/methanol. The desired
product is one of the final substances to be eluted because of
the protonated imidazole nitrogen. Finally, the product was
partitioned between acid and base as described above. Rotary
evaporation of the solution then yielded a yellow oil that was
pure by TLC (ethyl acetate/methanol 80/20, Rf ) 0.2) and
NMR.
primary criterion for reactant selection was synthetic feasibil-
ity. A virtual library of 582 primary and secondary alkyl
halides, 378 anilines, 51 aldehydes, and 50 primary amines
was compiled. The fragments were joined to their appropriate
position on the imidazole scaffold using SYBYL. The final
library contained 3508 compounds: 479 N-alkylated imida-
zoles, 376 N-arylimidazoles, and 2550 1,5-disubstituted imi-
dazoles.
Cova len t DOCK. DOCK 4.029 was employed to screen the
library against cytochrome P450cam. Grid coordinates were
taken from the 1PHD imidazole complexed structure deter-
mined by Poulos.11 The charges for the heme were taken from
the semiempirical ZINDO calculations of Gilda Loew.30
The following compounds were made by the appropriate
procedure described above.
1-Met h ylcycloh exylim id a zole: 22% yield; 1H NMR
(CDCl3) δ 7.34 (1H, s, imidazole), 6.97 (1H, s, imidazole), 6.79
(1H, s, imidazole), 3.65 (2H, d, NCH2, J ) 8.2 Hz), 1.5-1.7
ppm (6H, m, cyclohexyl), and 1.05-1.19 (5H, m, cyclohexyl);
13C NMR (DMSO) δ 137.83, 128.11, 119.80, 52.01, 36.10, 29.82,
25.86, 25.09; EI-MS, m/z exptl 164.1313 (M+) (calcd 164.1313,
Covalent DOCK (CovDOCK) (G. Skillman, Ph.D. thesis,
UCSF), originally written to apply to serine proteases, was
modified to model the imidazole-iron coordinate bond. Cov-
DOCK adds harmonic restraints to the DOCK scoring function:
C
10H16N2).
1-(2-Eth ylbu tyl)im id a zole: 56% yield; 1H NMR (CDCl3)
CovDOCK score ) k1(dist)2 +
k2(angle A)2 + k3(angle B)2 (1)
δ 7.42 (1H, s, imidazole), 7.03 (1H, s, imidazole), 6.85 (1H, s,
imidazole), 3.81 (2H, d, NCH2, J ) 8 Hz), 1.63 (1H, septuplet,
2-ethylbutyl), 1.27 (4H, pentuplet, 2-ethylbutyl, J ) 9.4 Hz),
and 0.87 (6H, t, t-CH3, J ) 8.2 Hz); 13C NMR (DMSO) δ 137.63,
128.13, 119.71, 49.04, 36.10, 22.57, 10.34; EI-MS, m/z exptl
152.1338 (M+) (calcd 152.1313, C9H16N2).
k1 ) 50; k2 ) 0.01; k3 ) 0.01
The three restraints used in this work were a distance
restraint from the iron to the imidazole N3 and two orthogonal
angle restraints from nitrogens NA and NB on the heme to
the iron and to the imidazole N3 (Figure 1). The van der Waals
contributions of the nitrogen coordinating the imidazole to the
heme as well as the carbons adjacent to the heme were set to
zero. Removing the van der Waals contributions of these atoms
was necessary to allow for the close contacts observed in crystal
structures of P450 imidazole inhibitor complexes.
1-Cyclop en tyl-5-p r op ylim id a zole: 3% yield; 1H NMR
(CDCl3) δ 7.49 (1H, s, imidazole), 6.79 (1H, s, imidazole), 4.33
(1H, pentuplet, cyclopentyl, J ) 8.0 Hz), 2.52 (2H, t, propyl, J
) 8.0 Hz), 1.6-1.95 (10H, m), and 1.05 (3H, t, propyl, J ) 7.2
Hz); 13C NMR (CDCl3) 133.58, 131.72, 125.5, 55.72, 33.47,
26.27, 23.76, 21.45, 13.8; EI-MS, m/z exptl 178.1469 (M+) (calcd
178.1470, C11H18N2).
1-Meth ylcyclop r op yl-5-p r op ylim id a zole: 11% yield; 1H
NMR (CDCl3) δ 7.65 (1H, s, imidazole), 6.77 (1H, s, imidazole),
3.62 (2H, d, NCH2, J ) 6.8 Hz), 2.53 (2H, t, propyl, J ) 8.0
Hz), 1.76 (2H, sextet, propyl, J ) 7.2 Hz), 1.21 (1H, m,
cyclopropyl), 0.98 (3H, t, propyl, J ) 7.2 Hz), 0.34 (2H, q,
cyclopropyl, J ) 5.2 Hz), and 0.24 (2H, q, cyclopropyl, J ) 4.4
Hz); 13C NMR (CDCl3) δ 136.07, 132.62, 125.70, 49.46, 23.46,
19.27, 14.16, 10.91, 4.14; EI-MS, m/z exptl 164.1313 (M+) (calcd
164.1313, C10H16N2).
1-Meth ylcyclop r op yl-5-bu tylim id a zole: 13% yield; 1H
NMR (CDCl3) δ 7.53 (1H, s, imidazole), 6.77 (1H, s, imidazole),
3.66 (2H, d, NCH2, J ) 6.8 Hz), 2.41 (2H, d, butyl, J ) 7.2
Hz), 1.88 (1H, nonatet, butyl, J ) 6.8 Hz), 1.16 (1H, m,
cyclopropyl), 0.66 (2H, q, cyclopropyl, J ) 8.4 Hz), and 0.33
(2H, q, cyclopropyl, J ) 4.8 Hz); 13C NMR (CDCl3) δ 136.01,
130.32, 126.66, 49.04, 32.97, 27.81, 22.32, 11.12, 4.08; EI-MS,
m/z exptl 178.1470 (M+) (calcd 178.1470, C11H18N2).
1-Ben zyl-5-cyclop r op ylim id a zole: 6% yield; 1H NMR
(CDCl3) δ 7.42 (1H, s, imidazole), 6.84 (1H, s, imidazole), 5.15
(2H, s, methyl), 7.23-7.34 (5H, m, benzyl), 0.91 (1H, m,
cyclopropyl), 0.85 (2H, m, cyclopropyl), and 0.68 (2H, q,
cyclopropyl, J ) 4.4); 13C NMR (CDCl3) δ 137.161, 136.44,
134.24, 128.70, 128.40, 127.74, 125.72, 48.31, 5.51, 4.34; EI-
MS, m/z exptl 198.1153 (M+) (calcd 198.1157, C13H14N2).
1-Cycylop r op yl-5-ter t-bu tylim id a zole: 9% yield; 1H NMR
(CDCl3) δ 7.41 (1H, s, imidazole), 6.77 (1H, s, imidazole), 3.09
(1H, m, cyclopropyl), 2.13 (2H, s, tert-butyl), 1.0 (9H, s, tert-
butyl CH3), and 0.85-1.05 (4H, m, cyclopropyl); 13C NMR
(CDCl3) δ 136.54, 129.62, 127.87, 37.20, 30.78, 29.32, 26.04,
6.25; EI-MS, m/z exptl 178.1467 (M+) (calcd 178.1470, C11H18N2).
Libr a r y Design . Because azole compounds have long been
known for their utility as P450 inhibitors,9 we chose an
imidazole scaffold for this project. To create the virtual library,
the imidazole scaffold was modeled in SYBYL (Tripos Inc., St.
Louis, MO) and substituents were selected from the Available
Chemical Database (ACD) (Molecular Design Systems, San
Leandro, CA) based on potential synthetic routes for creating
1-alkylimidazoles,26 1-arylimidazoles,28 and 1,5-disubstituted
imidazoles.27 The corresponding primary and secondary halide,
aniline, aldehyde, and primary amine reactants were extracted
from the ACD using Merlin (Daylight Company, Santa Fe,
NM) and UCSelect (G. Skillman, Ph.D. thesis, UCSF). Our
Ack n ow led gm en t. We thank Wesley Straub for
help with the synthetic organic chemistry procedures,
Geoffry Skillman for advice and the use of his code, and
Susan Miller for discussions of inhibitor kinetics. This
work was supported by National Institutes of Health
Grants GM25515 and GM56531 and by a BioStar grant
from California.
Refer en ces
(1) Ortiz de Montellano, P. R. In Cytochrome P450: Structure,
Mechanism, and Biochemistry, 2nd ed.; Ortiz de Montellano, P.
R., Ed.; Plenum Press: New York, 1995.
(2) Guengerich, P.; Chun, Y.; Kim, D.; Gillam, E.; Shimada, T.
Cytochrome P450 1B1: a target for inhibition in anticarcino-
genesis strategies. Mutat. Res./ Fundam. Mol. Mech. Mutagen.
2003, 523, 173-182.
(3) Fujii, H.; Sato, T.; Kaneko, S.; Gotoh, O.; Fujii-Kuriyama, Y.;
Osawa, K.; Kato, S.; Hamada, H. Metabolic inactivation of
retinoic acid by a novel P450 differentially expressed in develop-
ing mouse embryos. EMBO J . 1997, 16, 4163-4173.
(4) McLemore, T. L.; Litterst, C. L.; Coudert, B. P.; Liu, M. C.;
Hubbard, W. C.; Adelberg, S.; Czerwinski, M.; McMahon, J . A.;
Eggleston, J . C.; Boyd, M. R. Metabolic activation of 4-ipomeanol
in human lung, primary pulmonary carcinomas, and established
human pulmonary carcinoma cell lines. J . Nat. Cancer Inst.
1990, 82, 1420-1426.
(5) Podust, L. M.; Stojan, J .; Poulos, T. L.; Waterman, M. R.
Substrate recognition sites in 14-sterol demethylase from com-
parative analysis of amino acid sequences and X-ray structure
of Mycobacterium tuberculosis CYP51. J . Inorg. Biochem. 2001,
87, 227-235.
(6) Williams, P. A.; Cosme, J .; Sridhar, V.; J ohnson, E. F.; McRee,
D. E. Mammalian microsomal cytochrome P450 monooxy-
genase: structural adaptations for membrane binding and
functional diversity. Mol. Cell 2000, 5, 121-131.
(7) Williams, P. A.; Cosme, J .; Ward, A.; Angove, H. C.; Vinkovi, D.
M.; J hoti, H. Crystal structure of human cytochrome P450 2C9
with bound warfarin. Nature 2003, 424, 464-468.
(8) Zhang, Z.; Sibbesen, O.; J ohnson, R. A.; Ortiz de Montellano, P.
R. The substrate specificity of cytochrome P450cam. Bioorg. Med.
Chem. 1988, 6, 1501-1508.
(9) Testa, B.; J enner, P. Inhibitors of cytochrome P-450s and their
mechanism of action. Drug Metab. Rev. 1981, 12, 1-117.

P450 Imidazole Ligand Design
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 14 3579
(10) Rogerson, T. D.; Wilkinson, C. F.; Hetarski, K. Steric factors in
the inhibitory interaction of imidazoles with microsomal en-
zymes. Biochem. Pharmacol. 1977, 26, 1039-1042.
(11) Poulos, T. L.; Howard, A. J . Crystal structures of metyrapone-
and phenylimidazole-inhibited complexes of cytochrome P-450cam.
Biochemistry 1987, 26, 8165-8174.
(12) Yano, J . K.; Koo, L. S.; Schuller, D. J .; Li, H.; Ortiz De
Montellano, P. R.; Poulos, T. L. Crystal structure of a thermo-
philic cytochrome P450 from the archaeon Sulfolobus solfactari-
cus. J . Biol. Chem. 2000, 275, 31086-31092.
(21) Raag, R.; Poulos, T. L. Crystal structures of cytochrome P-450cam
complexed with camphane, thiocamphor, and adamantane:
factors controlling P-450 substrate hydroxylation. Biochemistry
1991, 30, 2674-2684.
(22) Lavrijsen, K.; Houdt, J . V.; Thijs, D.; Meuldrmans, W.; Heykants,
J . Induction potential of antifungals containing an imidazole or
triazole moiety. Miconazole and ketoconazole, but not itracona-
zole are able to induce hepatic drug metabolizing enzymes of
male rats at high doses. Biochem. Pharmacol. 1986, 35, 1867-
1878.
(13) Podust, L. M.; Poulos, T. L.; Waterman, M. R. Crystal Structure
of cytochrome P450 14alpha-sterol demethylase (Cyp51) from
Mycobacterium tuberculosis in complex with azole inhibitors.
Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 3068-3073.
(14) Gschwend, D. A.; Sirawaraporn, W.; Santi, D. V.; Kuntz, I. D.
Specificity in structure-based drug design: identification of a
novel, selective inhibitor of Pneumocystis carinii dihydrofolate
reductase. Proteins 1997, 29, 59-67.
(23) Gunsalus, I. C.; Wagner, G. C. In Methods in Enzymology;
Fleischer, S., Packer, L., Eds.; Academic Press: New York, 1978;
Vol. 52, Part C, pp 166-188.
(24) Rao, S.; Wilks, A.; Hamberg, M.; Ortiz de Montellano, P. R. The
lipoxygenase activity of myoglobin. Oxidation of linoleic acid by
the ferryl oxygen rather than protein radical. J . Biol. Chem.
1994, 269, 7210-7216.
(25) Copeland, R. A. Enzymes, 2nd ed.: Wiley-VCH: New York, 2000;
pp 76-109.
(26) Grimmett, M. R. Comprehensive Heterocyclic Chemistry; Katritz-
ky, A. R., Rees, Ch. W., Eds.; Pergamon Press: Oxford, 1984; p
382.
(27) Leusen, A.; Wildeman, J .; Oldenziel, O. Base-induced cyclo-
addition of sulfonylmethyl isocyanides to C,N double bonds.
Synthesis of 1,5-disubstituted and 1,4,5-trisubstituted imidazoles
from aldimines and imidoyl chlorides. J . Org. Chem. 1977, 42,
1153-1159.
(28) J ohnson, A. L.; Kauer, D.; Sharma, D. C.; Dorfman, R. I. The
synthesis of 1-arylimidazoles, a new class of steroid hydroxyl-
ation inhibitors. J . Med. Chem. 1969, 12, 1024-1028.
(29) Ewing, T. J .; Makino, S.; Skillman, A. G.; Kuntz, I. D. DOCK
4.0: search strategies for automated molecular docking of
flexible molecule databases. J . Comput.-Aided Mol. Des. 2001,
15, 411-428.
(30) Collins, J . R.; Camper, D. L.; Loew, G. H. Valproic acid
metabolism by cytochrome P450: a theoretical study of stereo-
electronic modulators of product distribution. J . Am. Chem. Soc.
1991, 113, 2736-2743.
(15) Schenkman, J . B.; Remmer, H.; Estabrook, R. W. Spectral
studies of drug interaction with hepatic microsomal cytochrome.
Mol. Pharmacol. 1967, 3, 113-123.
(16) De Voss, J . J .; Sibbesen, O.; Zhang, Z. P.; Ortiz de Montellano,
P. R. Substrate docking algorithms and prediction of the
substrate specificity of cytochrome P450cam and its L244A
mutant. J . Am. Chem. Soc. 1997, 119, 5489-5498.
(17) Kelly, S. L.; Lamb, D. C.; Cannieux, M.; Greetham, D.; J ackson,
C. J .; Marczylo, T.; Ugochukwu, C.; Kelly, D. E. An old activity
in the cytochrome P450 superfamily (CYP51) and a new story
of drugs and resistance. Biochem. Soc. Trans. 2001, 29, 122-
128.
(18) Kelly, S. L.; Lamb, D. C.; Loeffler, J .; Einsele, H.; Kelly, D. E.
The G464S amino acid substitution in Candida albicans sterol
14alpha-demethylase causes fluconazole resistance in the clinic
through reduced affinity. Biochem. Biophys. Res. Commun. 1999,
262, 174-179.
(19) Morrison, J . F.; Walsh, C. T. The behavior and significance of
slow-binding enzyme inhibitors. Adv. Enzymol. Relat. Areas Mol.
Biol. 1988, 61, 201-301.
(20) Poulos, T. L.; Finzel, B. C.; Howard, A. J . High-resolution crystal
structure of cytochrome P450cam. J . Mol. Biol. 1987, 195, 687-
700.
J M030608T